The Power of Generative AI: A Conversation with Thomas Hadden

Generative AI has been making headlines in recent months, and its potential impact on various industries is vast. In this conversation, we discussed the future of generative AI with Thomas Hadden, a senior manager at Deloitte. The conversation was informative, exciting, and at times, cautious.

As we began our discussion, Thomas emphasized that generative AI is a very powerful tool that can have a significant impact on organizations. He advised managers to be aware of the technology and its potential applications within their company. "Let us know what you're doing," he said, "but we think it's going to be really critical to our future and we'd like you to experiment with it." This sentiment is echoed by many experts in the field, who believe that generative AI has the potential to revolutionize various industries.

One of the key takeaways from Thomas's conversation was the importance of community of practice. He suggested that organizations should create communities where people can share their experiences, learn from each other, and discuss best practices for implementing generative AI. "What works? What doesn't work?" he asked. "Which tools work best for us?" These questions are crucial in understanding the nuances of generative AI and how it can be effectively applied within an organization.

Thomas also highlighted the need for legal considerations when working with generative AI. He emphasized that general counsel and legal staff should be involved in any project related to AI, particularly those that involve generating content or interacting with users. "There's some maybe some uh security issues ETC some ethical issues," he noted. These concerns are becoming increasingly important as AI becomes more ubiquitous in our daily lives.

One of the most exciting aspects of generative AI is its potential to transform the way we interact with technology. Thomas predicted that conversational interfaces will become the norm, replacing traditional command-line interfaces and point-and-click systems. "Almost every online activity," he said, "search or on social media or just you know internet e-commerce" will be mediated by a conversational front end.

However, there are also challenges to overcome when it comes to generating high-quality content with generative AI. Thomas noted that chatbots will require large amounts of high-quality customer service content to function effectively. "Many companies haven't bothered to curate in the past," he observed. This is an area where organizations can improve their preparedness for the rise of conversational interfaces.

As we approached the end of our conversation, Thomas offered some final advice for organizations looking to get ahead of the curve when it comes to generative AI. "The time to move on these things is now," he said. "I mean, the Morgan Stanley people have been working with Open AI now for 18 months and Deloitte has been working with Open AI for 18 months on code generation." This highlights the importance of starting small and experimenting aggressively, rather than waiting for others to develop the technology.

Finally, Thomas emphasized the need to think big about the potential impact of generative AI on an organization. "You can start small," he advised, "but think big about what it's going to do for your organization and how it's going to transform it." This mindset is crucial in harnessing the power of generative AI and staying ahead of the competition.

As our conversation came to a close, Thomas offered some parting words of wisdom. The future of generative AI is exciting and uncertain, but one thing is clear: organizations must be prepared to adapt and evolve in order to stay ahead of the curve. With careful planning, experimentation, and collaboration, it's possible to unlock the full potential of this powerful technology.

"WEBVTTKind: captionsLanguage: enwell you know I think the time to move on these things is now um I mean the the Morgan Stanley people have been working with open AI now for 18 months and um Deloitte has been working with Oakland AI for 18 months on code generation so you don't want to just sit back and wait for things to develop it it may be um a hard thing to be a fast follower on so um start experimenting aggressively think big about what it's going to do for your organization and how it's going to transform it so think big but you start small if you if you need to all right Thomas Davenport it's great to have you on the show thanks for having me you know 10 years ago you go you co-authored the article with DJ Patel in the Harvard Business Review titled data science is the sexiest job of the 21st century you know there's a lot to unpack and how data science has evolved since the article but maybe to set the stage for today's conversation is data science still the sexiest job of the 21st century and if so why yes and if not why not well first I would say um you know I think it was a memorable title but there were many non-sexy aspects of data science um at the time and I remember thinking as I was interviewing the people for the article um I said you know there's a whole lot of data Plumbing going on here that's not sexy at all but um it was obviously sexy in terms of the desirability of you know hiring these people and the labor market and so on um I think that's still the case um it's changed a bit in that um there are many many other sources of data science expertise than sort of the traditional sort of PhD in physics types who were really the only thing around at the time um and um the job has fragmented a fair amount there are these different roles and then um we can do more with citizen data science than we could um at the time certainly um um so you know we realize now these are not unicorns they're pretty good at creating models but now we realize that's not all there is to data science um to successful data science anyway so I don't think you know there's a lot of clarity in organizations about who is a data scientist and who isn't and we haven't really succeeded very much in creating certification for them and skill testing and so on despite you know a few attempts but in general you know it's still a desirable job and um if you have the skills it's great to be a data scientist yeah that's great and you recently co-wrote another article with DJ Patel on is data science still sexy uh and you know data science in a lot of ways you mentioned in the article has become a lot more institutionalized you know we've seen over the past few years a significant increase in investment in data science by organizations across various Industries how would you characterize the current landscape of investment for data science that organizations and leaders are doing and how do you think that shift and that increase has impacted the state of data science adoption today in comparison to when you first wrote the article it's interesting I you know I think data science is much more institutionalized um in some companies it has become you know it's kind of the um preeminent job in the sense the company realizes is it's a data-driven business and they are um trying in all sorts of ways to understand their data and develop products and services based on it and so on but sadly I think that's still a minority of companies and in most companies um there doesn't seem to be a data-driven culture um uh a lot of the decision making is still not data driven a lot of the senior Executives still don't understand what's um so important about this and so um in the majority of companies I think even though there may be some data scientists they're still sort of you know kind of on the margins of the of the business um and they're not their work is not given the respect that it deserves I would say um but you know um that was true with traditional analytics as well a relatively small percentage of companies um decided that they wanted to compete on the basis of their analytical capabilities and a relatively small percentage of companies are competing on data science and AI today getting larger and that's um great progress but still a minority I think yeah one thing you know early on that we saw is that data science adoption really shifted by industry as well and changed by industry we saw you know Financial Services organizations try to become data driven and adopt data science much quicker you know data Rich Industries tend to be uh much more skewed towards adoption faster than other Industries can you give us maybe an overview of how data science is being where data science is being most widely adopted today where which Industries you know outside of Technologies are leading the pack and which Industries are maybe lagging behind sure yeah still the case that you know it takes a lot of data to do data science well so if a company doesn't have a lot of data they're they're going to be severely handicapped um um and I still think you know the the biggest industries or financial services where there's a lot of data both Banking and insurance and certainly you know Investments and hedge funds and so on um a lot of data in Telecom so uh Telecom I would say is um it doesn't have the same level of of um kind of cultural centricity as it does in financial services but it's I'm pretty powerful and I just um a few a few weeks ago wrote a piece about a t and how they're um they have a lot of data scientists but I think quite interestingly they have developed a very active sort of Citizen data science program they've done a lot to democratize the activity and I think that's what many organizations need to do if they're serious about data science they can't just rely on a relatively small percentage of you know highly highly trained professionals um other Industries um I recently did a session um with Johnson and Johnson and it's mostly a pharmaceutical company these days and um this was for all the people who are interested in data science is the first one they'd had I've done this a lot speaking at the first one they got 5 000 people to sign up so um that tells you there's a lot of data science going on in the pharmaceutical industry and I see it elsewhere um in that in that industry manufacturing has been relatively slow to adopt data science um it's kind of coming on with um predictive maintenance and digital Twins and so on but it's still behind most organizations some really high-tech manufacturing um has a fair amount of it I've written some things about Seagate for example and they're used of um uh image detection image analysis systems with AI to detect uh problems in electron microscope images so you know that's quite sophisticated but it's relatively rare I think in that industry and um I don't know what else you're starting to see it a bit in Professional Services not just you know Consulting to other organizations but in audit and tax and law even um still and I think it's AI enters law you'll see a lot more of it um I don't know what other Industries are there I can't I can't think of any others at the moment mid-sized companies tend to be less aggressive than either startups or really large companies so that's another distinction to me yeah that's definitely that's definitely the case and we've seen that as well in case of you know data culture adoption as well from our perspective you know keeping here on the theme of data culture and kind of creating a culture of decision making and data-driven decision making as you mentioned that's a obstacle that a lot of organizations have suffered with in the past uh you know decade or so as they increase adoption why do you think that people component and cultural component is still a major obstacle for organizations and what have you seen to be a good pattern or success stories from organizations who've been able to overcome that hump yeah well I think there are two two primary issues that are related um culture you know tends to trickle down from the top of the organization and if you don't have senior Executives who are really committed to data science and analytics and AI as really important resources in the business it's going to be much harder for the rest of the organization to adopt a data-driven culture um and um a related factor is that most organizations you know we've had this feeling there's that old um phrase you can lead a horse to water but you can't make a drink and I don't know if that's true in in Europe it was popular at one point in the United States and um we've we think that because we develop these um uh Information Systems uh and you know analytics and Ai and so on that people will actually use them but that's not always true and we don't really we invest a tiny fraction of um the amount we spend on technology in cultural change and education and initiatives to create a more data-driven culture so you know it's like one percent to 99 so uh and that's despite the fact that um every year I do a survey of large companies typically about a hundred or so large companies mostly in financial services but other Industries as well with new Vantage partners and um every year um the CEO who whom I sort of analyzed this with Randy beans as a question um what's the primary cause of your challenges with um data and analytics and AI is it technology or is it human cultural organizational process factors and generally it's you know between 80 and 90 percent human organizational process culture factors um but you know nobody spends 80 to 90 percent of their their budgets on the um those things so there's a real imbalance between what we spend on technology and and the um attention the technology gets and the attention that the cultural side gives in a lot of ways there's a bit full of shiny toys here the organizations fall into it's very easy to develop a high quality cool robust predictive model that that may not or may or may not get actually deployed into production but a lot of the times enabling folks with local tools and you know data-driven decision making the literacy can get you a long way you know we've seen in a lot of ways organizations shift their uh their priority and create their literacy programs have you seen that succeed in organizations is that only enough to create a data culture or is there an additional mile that needs to be crossed there um well you know it's a good thing to do that while it's a good thing it wasn't nearly enough it should be tailored to um particular parts of a business there should be a human component you know face to face probably where you can discuss these issues not just um uh um watch a few minutes of online video um and so um just go ahead but not enough and you know a friend of mine he um he's recently changing job jobs but he was the head of analytics and AI um data science at Eli Lilly and he said you know um these cultural change programs are really multifaceted you know you have to do data literacy you should do it differently for different levels and different parts of the organization and so on but you should also have one-on-one things with senior Executives you should have um uh communities developer on the organization you should have um uh kind of uh Behavior change programs in meetings you know I always thought the best thing you can do is have somebody in a meeting say excuse me but do you have data to support that hypothesis or or oh by the way you know you're showing a correlation but that doesn't mean there's causality going on that best thing that could happen and maybe the hardest thing to to change you know it's one thing to develop skill sets right but it's another thing to apply in the flow of work and more importantly to develop that healthy data skepticism within the organization where there is an open and honest conversation about the data around how can you action that in a particular business setting now you know let's shift gears and you know discuss also how data science has matured uh from across different dimensions you mentioned earlier in our discussion how the data science role responsibility and skill set has all as well maybe walk us through some of the key shifts that you've seen in the data science role in the past few years how has that differed since you know the nascent days of data science as a profession yeah well one big difference is that you know data scientists were the only job that were supposed to advance data science and you know um we found out that didn't work very well you know these people are not unicorns they can't do it all um they're particularly good at developing models that that fit a set of data and maybe writing some python code to to make it all work but some of the other factors not so good at maybe not so good at interfacing with the business um and building their trust maybe not so good at building um a machine learning infrastructure at scale um maybe not so good at um changing the organization and all the necessary ways to make effective use of their models you know changing the process changing the skills um changing the culture Etc um and maybe not so good at kind of ongoing management of the system once it's been put into deployment assuming you're lucky enough to get it into deployment not not very many systems were that was part of the problem so um we've had this evolution of of um disaggregation of jobs so now you have data scientists and you have machine learning engineers and you have data product managers which I think are the single in a way most important job because they're the ones who integrate all this stuff you have ml Ops engineers in some organizations you have translators I think many of those translation functions can be right data product managers but um you really have seen a proliferation um a a data Engineers more broadly to take some of that data wrangling off the data scientists um you've seen a proliferation of these jobs and um I think it's a great thing but you know they have to be coordinated um in the context of a particular project and that's where the data product management role often comes into play so in a lot of ways the story of the past few years have been you know I've seen a debate raging on in the data science space between you know is the optimal path for a data science profession specialization or generalization do you think that you know as data science has matured the path for succeeding in data science is specialization you know becoming a specialist in a particular area of the data science pipeline for example well you know I I think it depends to some degree on what your skills are there certainly are some data scientists who like dealing with with managers who like overseeing a project um who you know understand uh how to create a model But realize that's only a small fraction of a job but they're pretty rare to be honest and most of the time people got into data science because they like modeling you know they like coding um those were the initial skills that were considered most important and most valuable that's why you needed you know phds and these quantitative disciplines and in many cases um you know and I often say you know um uh Librarians sometimes like books more than people um system developers like computers More Than People data scientists sometimes like models more than people so um I think you have to go with the um kind of skills and inclinations that you have in most cases I think you're going to be better off with that kind of specialization and the data scientists who really like that broader range of activities can you know move into Data product management data oversight of the data science function and so on where they have more managerial activities and less you know kind of day-to-day model and stuff you know and what's interesting speaking of skills here is not just how the skill sets and roles and responsibilities of data scientists have evolved is the also the educational landscape within data science and how would that how you know education within the space has evolved to unlock that specialization you know can you shed some light maybe on the changes of the educational landscape within the data science Industry and what that looks like today yeah well you know you guys are in that business so um you probably know more about it than than I do but from my perspective being a professor um you know we at Babson College have a uh uh a Masters of Science in um business analytics uh there are literally hundreds of these programs in the United States alone five years ago I wrote about this and there were over 200 just in Business Schools and they're also programs in data science and analytics and AI in engineering schools computer science schools Etc so they're really all over the place and I in in many cases I think there are two problems one is it's hard for somebody who's applying to one of these programs to know okay well what am I getting mostly am I getting um something that lets me do hardcore modeling if so I better have some pretty good statistics skills before I go in because these are mostly one-year programs um or am I getting something that's more oversight of the entire process or um you know Babson my school is mostly known for entrepreneurship and we try to take an entrepreneurial span on business analytics which is a little unusual but valuable if you want to go into that that combination of of emphasis so students don't know what they're getting and in many cases I don't think that one year is enough to create you know a great State data scientists people I think it somebody was telling me the other day they think what it really does is create excellent citizen data scientists and that's probably true and then you end up focusing on some part of the business domain supply chain or marketing or whatever in terms of getting a lot of expertise in that and knowing how you can use data science to advance that that particular aspect of the business and you mentioned that especially it's great like you know a lot of programs are great creating citizen data scientists you know we see that on the employer side as well you know a lot of times employers you know have a lot of ease in getting you know Junior data scientists data analysts in the door when it comes to hiring right and recruiting but struggle in getting that extremely proficient Advanced Data science sound within the organization that has a lot of experience in machine learning modeling deploying these models Etc what do you think needs to change in the education industry today or education in general to be able to create a healthier pipeline from that Junior talent to Advanced Talent well you know it's interesting there are a ton of Master's programs not very many PHD programs in the world in data science so that would be one one thing I think um you know data science is almost by definition a multi-disciplinary activity and universities don't tend to do well at that you know they they're good at creating statisticians or physicists or whatever but if you're combining a bunch of different skills which may come from different parts of the of the University not so good generally um uh I think you know um somebody who has a master's degree could be encouraged to go back and get more um training maybe to get a PhD or more specialized training in certain aspects of data science you know I I think if you are highly motivated to acquire new skills there are tons of places where you can get them your organization you know there are tons of these online courses by uh Coursera and udemy and so on um if you're highly motivated you can get the education that you want even I mean YouTube but um most people don't really know enough about what they want and so I think it means that organizations need to develop um sort of you know categorizations and certification programs within their companies to say okay here are the skills that you need to really be you know a level one data science just I'm here some places you can get them here if you want to be a data product manager here's what you need if you want to be a data engineer here's what you need a few companies are doing that but they're not nearly enough um out there yeah and this connects to another question that I wanted to ask is that you know a lot of organizations well they've made strides in building data teams I think if you want to compare data science maybe to a natural counterpart which is software engineering software engineering tends to be a lot more mature when it comes to leveling right you get a junior software engineer level one there's a level two level three you become a staff engineer after that distinguished engineer Etc and so on and so forth or you become a people manager data science doesn't necessarily have that it's not codified across the industry you see organizations trying to do that for their specific use cases what is the solution to the lettering problem and data science because I do see that in the long term this can create problems where data scientists feel stagnant depending on the type of organization they they work in yeah well you know it's interesting in systems um on the system side you tended to have some dominant vendors and a lot of Microsoft certified systems Engineers for example you don't really have that on the data science side so it would be helpful if some vendors would step up um and um there are some organizations out there there's one that I've done a little bit of work with called iadss I forget um what the letters stand for that's created a certification approach there's an analytics certification approach that um is fairly popular um it's called a cap program certified analytics professional it doesn't really deal with all aspects of data science so so um we need more people developing certification approaches and I think if if you know that some vendor could inject a lot of money and and um resources into it that would probably help even though it'd be great if sort of Open Source it but um that hasn't happened okay that's great so you know let's talk about we talked about data science how is it matured a lot but you know your latest book on all in on AI we used to talk about that as well given especially the the many changes and movement that we're seeing in the AI space so you know as data science is becoming more institutionalized maybe AI is this new frontier for organizations you mentioned in the book that less than one percent of large organizations view themselves as AI driven uh maybe the first set definitions trait what does it mean to be an AI fueled organization there's a lot of sort of breadth and depth the breadth is lots and lots of use cases um you know um with Deloitte participated in some surveys um where you know maybe 25 of large organizations have seven or more use cases implemented and are getting some positive outcomes out of them so that's still a small number um there are there are companies with literally thousands of use cases that have been implemented so you need a lot you need a variety of different types of Technology AI technology in your organization conventional machine learning deep learning even rules automation oriented Technologies um conversational AI which is increasingly based on deep learning but wasn't always so um a lot of a lot of different Technologies a lot of support from senior Executives a lot of investments in the area a lot of talent um and some sort of ethical framework to make sure you your AI work doesn't go off the rails um you need a governance structure at least a center of excellence um to kind of manage these activities and select tools and manage their careers of people and and so on so you know those are the kinds of things and yeah it takes takes a lot of money to make that work but um the companies that we found that are doing it seem to be quite successful definitely takes a lot of money takes a lot of resources so maybe let's unpack a lot of that and start at the beginning you know you discuss and the book on how to become AI driven that there are three strategic archetypes usually the AI organizations can adopt right uh can you explain maybe these archetypes if an organization is trying to get started wants to look at a road map a framework for developing their AI strategy what should that look like and you could do more than one but I think um it's good to sort of um have the Strategic discussions around the organization about well you know what what are we going to do with AI so there's the doing something new idea new strategy new business model um um important new products and services that you're going to offer um and that to me is the most interesting one although um you know it can be can be hard to do requires a lot of conviction on the part of of seeing your Executives and it requires a lot of people to understand the business strategy and AI so they can kind of make that connection um the second one is operational transformation in a not just improving things on the margin but doing dramatic change in how you do key aspects of the operation so one of our examples of that is shell where they're um using things like AI image recognition to inspect all of their refineries with you know drones and um image recognition to identify potential problems in piping and valves and so on so they can send a human out to look at what what appears to be a problematic area just a dramatic change in how you do maintenance and they're also doing some really interesting things with subsurface stuff and some probably not enough yet for the planet but some things also in the space of um of uh non-carbon-based energy energy forms um you know where to put charging stations on that kind of machine learning analysis and then um uh the third category is really trying to change customer Behavior where you are um uh trying to do something with AI to make life better for your customers and that tends to be insurance companies who used to you know they'd pay you if you died or if you got sick or if you crash your car but they sort of realized gee wouldn't it be better if we helped our customers not die not crash their cars you know not get sick um so it's early days of that but using kind of AI and Behavioral Science nudges to create healthy behaviors better driving behaviors Etc so those are the three things operational transformation is the most common um changing customer Behavior probably the least common so maybe let's talk about operational transformation because you know mentioning that shell case that sounds like a very ambitious project right what would be your you know advice here for an organization starting off when trying to prioritize use cases right because you know conventional wisdom always says you need to start small and trade from there right versus going for a big project that has a lot of complexity requires a lot of talent a lot of data a lot of overhead what would be your advice when organizations and leaders are trying to make that trade off I think there's you know the um what kind of opportunity is there for this and um uh you know you can assess how much money are we spending in that particular area now how much could we save if we were doing this a different way you know for Shell it was um takes years to inspect a refinery in the whole refinery in the traditional approach and it takes days to inspect it in the AI enabled approach so obviously that's a big big payoff and you know the maintenance and inspection is an important part of you know keeping things going and successful so um uh and then there's the you know how viable is it and you have to sort of think about since some of these things take a while you have to think well how rapidly is the technology maturing we talked about a company in the book not well known as a kind of medium-sized company about approaching a billion in revenues called CCC intelligence Solutions and what they do is enable this capability to for insurance companies to let their customers take a photo of their um car collision damage and get an immediate estimate for what it's going to cost to fix it and they started on that well before you know deep learning based image recognition was at the capability it needed before camera resolution was at the smartphone camera resolution was the at the level that it needed to be and and so on um so it requires some ability to sort of look ahead into the future and see how these Technologies are going to be developing okay that's really great and we'll definitely talk about what you think about generative AI here but maybe before uh we talk about and look forward to the future in the book you also talk about the human side of becoming an AI driven organization this connects back to our earlier conversation on their literacy and data culture uh you know succeeding in AI you mentioned is the most important actually developing The Human Side is the most important aspect of succeeding in AI not the technology side something we wholeheartedly believe in a data Camp as well can you explain maybe why this is so important in the context of AI what an AI driven culture looks like and how can leaders cultivate such a culture well I think you know traditionally an AI driven culture required a lot of you know professional data scientists um but increasingly you know in part because of generative Ai and Automated machine learning and so on I think we're going to be seeing an environment where it has to be a democratic pursuit in the sense uh we we need tons of people doing this even at Shell you know I think they've now um certified 5 000 of their Engineers as knowledgeable in AI you know you know they had some online programs that they recommended and and so on so Airbus did a similar thing in um uh getting their engine a lot of their Engineers up to speed so um I think if if you're starting from scratch now and you don't have that strong um human culture for AI I think you're much better off thinking about how do I enable a large number of people in my organization many of whom are already employees to acquire these skills and you know the skills themselves are changing um there aren't too many programs out there on you know what do you need to know in order to be a good citizen data scientist how do you use um Automated machine learning effectively even more I think they're going to be machine learning programs that have uh some companies have already announced them I'm not sure they actually exist yet but but the front end is generative AI so how do you tell a generative AI system what you want out of a machine learning algorithm and um you know how do you refine your prompts how do you interpret the output Etc so we're going to need some new programs of that type as the as these Technologies mature okay I couldn't agree more and you know you mentioned here generative AI let's dive deep into it you know of course Johnson the way I have garnered quite a lot of attention recently maybe discussed I haven't heard I haven't heard of any maybe discuss what the generative AI impact or landscape of impact looks like for large organizations Enterprises today how should they think about leveraging these Technologies well you know it's um potentially quite earth-shattering it's quite early days but um very wide range of potential use cases um now I think we're seeing the relatively less important ones you know writing blog posts or writing copy for ads or marketing materials um really easy to do that and um you know it's what's created is not going to be world class but it's going to be certainly a good input to an expert copywriter and will very much improve their productivity um I think the whole world of coding is going to change dramatically and I was I'm talking about this yesterday with some of my friends at Deloitte we were together doing a presentation for a big private Equity Firm and Deloitte is saying they expect improvements in coding productivity of like 50 percent um uh which means that we need you know far fewer programmers than we thought we needed and I think more and more the programmers are going to be like everybody else are going to become editors of um generative AI code rather than you know people who write it in the first place um any writer is going to become an editor more than a first draft writer so um that's um something that we have to deal with I think there's going to be a lot of applications in r d both kind of product design in terms of image related Technologies but also um pharmaceutical companies are starting to figure out that um just as uh generative AI can model text sequences it can also model protein sequences or molecule sequences or something like that and see which ones are most likely to be be effective so big changes coming they are a huge amount of change I think in Professional Services law is I'm going to change dramatically so um uh you name it I think they're going to be a lot of impacts and now what should you be doing well I think a lot of experimentation um May you probably want some production applications um but I wrote a piece last week about Morgan Stanley um the big wealth management company is using it they they fine-tuned trained on a hundred thousand of their documents and they're using it to provide their financial advisors with high quality answers to questions they have on you know what are they invent what are the firm's analysts saying about particular Investments um from a recommendation Point how best to do certain tasks like how do I establish an irrevocable trials for my grandchildren um uh what regulatory requirements have to be met in in certain domains so the hundred thousand documents is not that many they have to be well curated to make sure that they're accurate and everything but I think it's going to revolutionize you know what we used to call Knowledge Management which was an early enthusiasm of mine um making it much much easier to capture high quality organizational knowledge this is this is a very fascinating space you know you mentioned here really uh the breadth and depth of use cases that generative AI can unlock in a lot of ways we're really still early if you want to compare this maybe the last technology that was that could have this potential of change is maybe the invention of the iPhone for example you know where you had a massive amount of software ecosystems that could have unlocked new type of application software maybe you know we're still really early days right now in the generative AI space but maybe walk us through how you think the economy will be impacted in the next few years you know I'm gonna look at it from two different perspectives here uh you know a lot of people tend to feel worry and anxiety when these types of tools come in right um as they can be displaced jobs disrupt jobs and leaders as well are having to you know address create a vision for generative AI within the organization that incorporates The Human Side as well what would be here your advice for you know professionals today that uh feel this worrying anxiety and what would be your advice for leaders who are trying as well to communicate a vision that both incorporates humans and generative AIS part of the organizational strategy and future well you know I've been a pretty um persistent advocate of augmentation over automation um and I've written two books on the subject one with 29 case studies of people whose work is augmented um day to day called working with AI another a little bit more theoretical but with some examples back in 20 2016 um and I I I didn't really suspect that large-scale automation could be coming anytime soon I'm less sure of that now with generative AI um uh you know in content creation Fields it is just so um amazingly productive and effective already you say in the early days that I do have some concerns um but I think um for now you're nuts if you take the output of generative Ai and release it directly into the marketplace without lots of you know editing and trying not multiple prompts and so on so there's still a human component at the beginning and the end of generative AI process I don't know that that will always be the case as it improves in capability but right now that's true so the most important thing you can do is start to use these tools as an individual and be more productive with them so that uh you know people can say oh yeah we're getting some value out of this we didn't have to hire an additional copywriter because our existing copywriters were so productive in what they did um sort of a recessionary climate and some economies now um we're already seeing a fair number of layoff you you want to make yourself as valuable an employee as possible so um you know by all means tell your boss what you're doing um where we're doing some I'm doing some work with another guy on the whole idea of Citizen development and some people have used these citizen development tools even before generative AI they've used them to hold down multiple remote jobs at once as they could be so much more productive that I do not recommend I think it would long run be not a good strategy but um and so managers senior managers should be saying this is a very powerful tool um let us know what you're doing but we think it's going to be really critical to our future and we'd like you to experiment with it and I think they should performing communities of practice within the organization to say you know what have we learned about this um what works what doesn't work which tools work best for us um are there legal issues uh clearly you know you want to get your general counsel and your legal staff involved for things that you know go out into the into the world in terms of production there's some maybe some uh security issues ETC some ethical issues I um go to a fair amount about unilever's approach to ethics AI ethics in my all in an AI book and I got a message um from uh the woman who focuses on ethics primarily their yesterday saying that they're quite focused on generative Ai and what does that mean for their ethics approach so um you you I think they're going to be a lot of potential ethical issues around misinformation and deep fakes and the usual issues of transparency and bias and so on yeah I couldn't agree more this is definitely an interesting exciting uh but also wary space in a lot of ways that I'm looking at maybe Tom took cap off our discussion uh I know it's always it's always annoying to put someone on the spot to give predictions uh given how fast the generative AI space is is moving right what would be your prediction for this space in 12 months from now well I think there will be um generative AI front ends to almost every software product that we use you know we've already seen some announcements of these for at blow and HubSpot and um uh um Salesforce CRM and so on um uh more and more companies are going to say um command line interfaces or even point and click interfaces are going to be replaced by conversational interfaces we're going to see finally I think chat Bots that are really quite useful um it's not going to happen overnight because they're going to have to be fine-tuned trained and uh going to be have to be on high quality customer service content which many companies haven't you know bothered to curate in the past um I think we'll see almost every online activity that we performed search or on social media or um uh just you know internet e-commerce will be mediated by a conversational front end um I think we'll still have um prompts but I suspect within 12 months we'll have some alternative to prompts where a generative ai's system would sort of lead us into creating a prompt with a sort of intelligent front end as as a prompt Creator um I think there will be many many many different customized models fine-tuned models Millions probably in 12 months um so I mean just think about in the legal profession there's already there are already a couple of uh you know large language models trained for the legal field but then you think about okay um there are differences between U.S law European law UK law so they're going to be versions for different countries or Regions they're going to be different versions for types of real estate law within those regions like real estate law and even individual law firms will say Okay Alan and Avery a UK firm we're going to have the Allen and ovary UK real estate law model that nobody else has it's not exactly like ours so they're going to be lots and lots of these things and no matter what we do you want to take a vacation in Peru there will be a model custom designed for where do you go where do you stay how do you spend your time in in Peru that is very very interesting and very fascinating it's going to be a very interesting future in next 12 months and Beyond uh finally Thomas do you have any final call to action or final words before we wrap up today's conversation well you know I think the time to move on these things is now um I mean the the Morgan Stanley people have been working with open AI now for 18 months and um Deloitte has been working with open AI for 18 months on code generation so um there are some companies that are already ahead of yours um in this regard so you don't want to just sit back and wait for things to develop it it may be a hard thing to be a fast follower on so um start experimenting aggressively get your content in shape this is a degenerative AI space it's a AI space in general I think um you can start small because AI 10 tends to be kind of small use cases small level of functionality individual tasks so far but think big about what it's going to do for your organization and how it's going to transform it so think big but you start small if you if you need to okay thank you so much Thomas for coming on the podcast my pleasure thanks for having me foreignwell you know I think the time to move on these things is now um I mean the the Morgan Stanley people have been working with open AI now for 18 months and um Deloitte has been working with Oakland AI for 18 months on code generation so you don't want to just sit back and wait for things to develop it it may be um a hard thing to be a fast follower on so um start experimenting aggressively think big about what it's going to do for your organization and how it's going to transform it so think big but you start small if you if you need to all right Thomas Davenport it's great to have you on the show thanks for having me you know 10 years ago you go you co-authored the article with DJ Patel in the Harvard Business Review titled data science is the sexiest job of the 21st century you know there's a lot to unpack and how data science has evolved since the article but maybe to set the stage for today's conversation is data science still the sexiest job of the 21st century and if so why yes and if not why not well first I would say um you know I think it was a memorable title but there were many non-sexy aspects of data science um at the time and I remember thinking as I was interviewing the people for the article um I said you know there's a whole lot of data Plumbing going on here that's not sexy at all but um it was obviously sexy in terms of the desirability of you know hiring these people and the labor market and so on um I think that's still the case um it's changed a bit in that um there are many many other sources of data science expertise than sort of the traditional sort of PhD in physics types who were really the only thing around at the time um and um the job has fragmented a fair amount there are these different roles and then um we can do more with citizen data science than we could um at the time certainly um um so you know we realize now these are not unicorns they're pretty good at creating models but now we realize that's not all there is to data science um to successful data science anyway so I don't think you know there's a lot of clarity in organizations about who is a data scientist and who isn't and we haven't really succeeded very much in creating certification for them and skill testing and so on despite you know a few attempts but in general you know it's still a desirable job and um if you have the skills it's great to be a data scientist yeah that's great and you recently co-wrote another article with DJ Patel on is data science still sexy uh and you know data science in a lot of ways you mentioned in the article has become a lot more institutionalized you know we've seen over the past few years a significant increase in investment in data science by organizations across various Industries how would you characterize the current landscape of investment for data science that organizations and leaders are doing and how do you think that shift and that increase has impacted the state of data science adoption today in comparison to when you first wrote the article it's interesting I you know I think data science is much more institutionalized um in some companies it has become you know it's kind of the um preeminent job in the sense the company realizes is it's a data-driven business and they are um trying in all sorts of ways to understand their data and develop products and services based on it and so on but sadly I think that's still a minority of companies and in most companies um there doesn't seem to be a data-driven culture um uh a lot of the decision making is still not data driven a lot of the senior Executives still don't understand what's um so important about this and so um in the majority of companies I think even though there may be some data scientists they're still sort of you know kind of on the margins of the of the business um and they're not their work is not given the respect that it deserves I would say um but you know um that was true with traditional analytics as well a relatively small percentage of companies um decided that they wanted to compete on the basis of their analytical capabilities and a relatively small percentage of companies are competing on data science and AI today getting larger and that's um great progress but still a minority I think yeah one thing you know early on that we saw is that data science adoption really shifted by industry as well and changed by industry we saw you know Financial Services organizations try to become data driven and adopt data science much quicker you know data Rich Industries tend to be uh much more skewed towards adoption faster than other Industries can you give us maybe an overview of how data science is being where data science is being most widely adopted today where which Industries you know outside of Technologies are leading the pack and which Industries are maybe lagging behind sure yeah still the case that you know it takes a lot of data to do data science well so if a company doesn't have a lot of data they're they're going to be severely handicapped um um and I still think you know the the biggest industries or financial services where there's a lot of data both Banking and insurance and certainly you know Investments and hedge funds and so on um a lot of data in Telecom so uh Telecom I would say is um it doesn't have the same level of of um kind of cultural centricity as it does in financial services but it's I'm pretty powerful and I just um a few a few weeks ago wrote a piece about a t and how they're um they have a lot of data scientists but I think quite interestingly they have developed a very active sort of Citizen data science program they've done a lot to democratize the activity and I think that's what many organizations need to do if they're serious about data science they can't just rely on a relatively small percentage of you know highly highly trained professionals um other Industries um I recently did a session um with Johnson and Johnson and it's mostly a pharmaceutical company these days and um this was for all the people who are interested in data science is the first one they'd had I've done this a lot speaking at the first one they got 5 000 people to sign up so um that tells you there's a lot of data science going on in the pharmaceutical industry and I see it elsewhere um in that in that industry manufacturing has been relatively slow to adopt data science um it's kind of coming on with um predictive maintenance and digital Twins and so on but it's still behind most organizations some really high-tech manufacturing um has a fair amount of it I've written some things about Seagate for example and they're used of um uh image detection image analysis systems with AI to detect uh problems in electron microscope images so you know that's quite sophisticated but it's relatively rare I think in that industry and um I don't know what else you're starting to see it a bit in Professional Services not just you know Consulting to other organizations but in audit and tax and law even um still and I think it's AI enters law you'll see a lot more of it um I don't know what other Industries are there I can't I can't think of any others at the moment mid-sized companies tend to be less aggressive than either startups or really large companies so that's another distinction to me yeah that's definitely that's definitely the case and we've seen that as well in case of you know data culture adoption as well from our perspective you know keeping here on the theme of data culture and kind of creating a culture of decision making and data-driven decision making as you mentioned that's a obstacle that a lot of organizations have suffered with in the past uh you know decade or so as they increase adoption why do you think that people component and cultural component is still a major obstacle for organizations and what have you seen to be a good pattern or success stories from organizations who've been able to overcome that hump yeah well I think there are two two primary issues that are related um culture you know tends to trickle down from the top of the organization and if you don't have senior Executives who are really committed to data science and analytics and AI as really important resources in the business it's going to be much harder for the rest of the organization to adopt a data-driven culture um and um a related factor is that most organizations you know we've had this feeling there's that old um phrase you can lead a horse to water but you can't make a drink and I don't know if that's true in in Europe it was popular at one point in the United States and um we've we think that because we develop these um uh Information Systems uh and you know analytics and Ai and so on that people will actually use them but that's not always true and we don't really we invest a tiny fraction of um the amount we spend on technology in cultural change and education and initiatives to create a more data-driven culture so you know it's like one percent to 99 so uh and that's despite the fact that um every year I do a survey of large companies typically about a hundred or so large companies mostly in financial services but other Industries as well with new Vantage partners and um every year um the CEO who whom I sort of analyzed this with Randy beans as a question um what's the primary cause of your challenges with um data and analytics and AI is it technology or is it human cultural organizational process factors and generally it's you know between 80 and 90 percent human organizational process culture factors um but you know nobody spends 80 to 90 percent of their their budgets on the um those things so there's a real imbalance between what we spend on technology and and the um attention the technology gets and the attention that the cultural side gives in a lot of ways there's a bit full of shiny toys here the organizations fall into it's very easy to develop a high quality cool robust predictive model that that may not or may or may not get actually deployed into production but a lot of the times enabling folks with local tools and you know data-driven decision making the literacy can get you a long way you know we've seen in a lot of ways organizations shift their uh their priority and create their literacy programs have you seen that succeed in organizations is that only enough to create a data culture or is there an additional mile that needs to be crossed there um well you know it's a good thing to do that while it's a good thing it wasn't nearly enough it should be tailored to um particular parts of a business there should be a human component you know face to face probably where you can discuss these issues not just um uh um watch a few minutes of online video um and so um just go ahead but not enough and you know a friend of mine he um he's recently changing job jobs but he was the head of analytics and AI um data science at Eli Lilly and he said you know um these cultural change programs are really multifaceted you know you have to do data literacy you should do it differently for different levels and different parts of the organization and so on but you should also have one-on-one things with senior Executives you should have um uh communities developer on the organization you should have um uh kind of uh Behavior change programs in meetings you know I always thought the best thing you can do is have somebody in a meeting say excuse me but do you have data to support that hypothesis or or oh by the way you know you're showing a correlation but that doesn't mean there's causality going on that best thing that could happen and maybe the hardest thing to to change you know it's one thing to develop skill sets right but it's another thing to apply in the flow of work and more importantly to develop that healthy data skepticism within the organization where there is an open and honest conversation about the data around how can you action that in a particular business setting now you know let's shift gears and you know discuss also how data science has matured uh from across different dimensions you mentioned earlier in our discussion how the data science role responsibility and skill set has all as well maybe walk us through some of the key shifts that you've seen in the data science role in the past few years how has that differed since you know the nascent days of data science as a profession yeah well one big difference is that you know data scientists were the only job that were supposed to advance data science and you know um we found out that didn't work very well you know these people are not unicorns they can't do it all um they're particularly good at developing models that that fit a set of data and maybe writing some python code to to make it all work but some of the other factors not so good at maybe not so good at interfacing with the business um and building their trust maybe not so good at building um a machine learning infrastructure at scale um maybe not so good at um changing the organization and all the necessary ways to make effective use of their models you know changing the process changing the skills um changing the culture Etc um and maybe not so good at kind of ongoing management of the system once it's been put into deployment assuming you're lucky enough to get it into deployment not not very many systems were that was part of the problem so um we've had this evolution of of um disaggregation of jobs so now you have data scientists and you have machine learning engineers and you have data product managers which I think are the single in a way most important job because they're the ones who integrate all this stuff you have ml Ops engineers in some organizations you have translators I think many of those translation functions can be right data product managers but um you really have seen a proliferation um a a data Engineers more broadly to take some of that data wrangling off the data scientists um you've seen a proliferation of these jobs and um I think it's a great thing but you know they have to be coordinated um in the context of a particular project and that's where the data product management role often comes into play so in a lot of ways the story of the past few years have been you know I've seen a debate raging on in the data science space between you know is the optimal path for a data science profession specialization or generalization do you think that you know as data science has matured the path for succeeding in data science is specialization you know becoming a specialist in a particular area of the data science pipeline for example well you know I I think it depends to some degree on what your skills are there certainly are some data scientists who like dealing with with managers who like overseeing a project um who you know understand uh how to create a model But realize that's only a small fraction of a job but they're pretty rare to be honest and most of the time people got into data science because they like modeling you know they like coding um those were the initial skills that were considered most important and most valuable that's why you needed you know phds and these quantitative disciplines and in many cases um you know and I often say you know um uh Librarians sometimes like books more than people um system developers like computers More Than People data scientists sometimes like models more than people so um I think you have to go with the um kind of skills and inclinations that you have in most cases I think you're going to be better off with that kind of specialization and the data scientists who really like that broader range of activities can you know move into Data product management data oversight of the data science function and so on where they have more managerial activities and less you know kind of day-to-day model and stuff you know and what's interesting speaking of skills here is not just how the skill sets and roles and responsibilities of data scientists have evolved is the also the educational landscape within data science and how would that how you know education within the space has evolved to unlock that specialization you know can you shed some light maybe on the changes of the educational landscape within the data science Industry and what that looks like today yeah well you know you guys are in that business so um you probably know more about it than than I do but from my perspective being a professor um you know we at Babson College have a uh uh a Masters of Science in um business analytics uh there are literally hundreds of these programs in the United States alone five years ago I wrote about this and there were over 200 just in Business Schools and they're also programs in data science and analytics and AI in engineering schools computer science schools Etc so they're really all over the place and I in in many cases I think there are two problems one is it's hard for somebody who's applying to one of these programs to know okay well what am I getting mostly am I getting um something that lets me do hardcore modeling if so I better have some pretty good statistics skills before I go in because these are mostly one-year programs um or am I getting something that's more oversight of the entire process or um you know Babson my school is mostly known for entrepreneurship and we try to take an entrepreneurial span on business analytics which is a little unusual but valuable if you want to go into that that combination of of emphasis so students don't know what they're getting and in many cases I don't think that one year is enough to create you know a great State data scientists people I think it somebody was telling me the other day they think what it really does is create excellent citizen data scientists and that's probably true and then you end up focusing on some part of the business domain supply chain or marketing or whatever in terms of getting a lot of expertise in that and knowing how you can use data science to advance that that particular aspect of the business and you mentioned that especially it's great like you know a lot of programs are great creating citizen data scientists you know we see that on the employer side as well you know a lot of times employers you know have a lot of ease in getting you know Junior data scientists data analysts in the door when it comes to hiring right and recruiting but struggle in getting that extremely proficient Advanced Data science sound within the organization that has a lot of experience in machine learning modeling deploying these models Etc what do you think needs to change in the education industry today or education in general to be able to create a healthier pipeline from that Junior talent to Advanced Talent well you know it's interesting there are a ton of Master's programs not very many PHD programs in the world in data science so that would be one one thing I think um you know data science is almost by definition a multi-disciplinary activity and universities don't tend to do well at that you know they they're good at creating statisticians or physicists or whatever but if you're combining a bunch of different skills which may come from different parts of the of the University not so good generally um uh I think you know um somebody who has a master's degree could be encouraged to go back and get more um training maybe to get a PhD or more specialized training in certain aspects of data science you know I I think if you are highly motivated to acquire new skills there are tons of places where you can get them your organization you know there are tons of these online courses by uh Coursera and udemy and so on um if you're highly motivated you can get the education that you want even I mean YouTube but um most people don't really know enough about what they want and so I think it means that organizations need to develop um sort of you know categorizations and certification programs within their companies to say okay here are the skills that you need to really be you know a level one data science just I'm here some places you can get them here if you want to be a data product manager here's what you need if you want to be a data engineer here's what you need a few companies are doing that but they're not nearly enough um out there yeah and this connects to another question that I wanted to ask is that you know a lot of organizations well they've made strides in building data teams I think if you want to compare data science maybe to a natural counterpart which is software engineering software engineering tends to be a lot more mature when it comes to leveling right you get a junior software engineer level one there's a level two level three you become a staff engineer after that distinguished engineer Etc and so on and so forth or you become a people manager data science doesn't necessarily have that it's not codified across the industry you see organizations trying to do that for their specific use cases what is the solution to the lettering problem and data science because I do see that in the long term this can create problems where data scientists feel stagnant depending on the type of organization they they work in yeah well you know it's interesting in systems um on the system side you tended to have some dominant vendors and a lot of Microsoft certified systems Engineers for example you don't really have that on the data science side so it would be helpful if some vendors would step up um and um there are some organizations out there there's one that I've done a little bit of work with called iadss I forget um what the letters stand for that's created a certification approach there's an analytics certification approach that um is fairly popular um it's called a cap program certified analytics professional it doesn't really deal with all aspects of data science so so um we need more people developing certification approaches and I think if if you know that some vendor could inject a lot of money and and um resources into it that would probably help even though it'd be great if sort of Open Source it but um that hasn't happened okay that's great so you know let's talk about we talked about data science how is it matured a lot but you know your latest book on all in on AI we used to talk about that as well given especially the the many changes and movement that we're seeing in the AI space so you know as data science is becoming more institutionalized maybe AI is this new frontier for organizations you mentioned in the book that less than one percent of large organizations view themselves as AI driven uh maybe the first set definitions trait what does it mean to be an AI fueled organization there's a lot of sort of breadth and depth the breadth is lots and lots of use cases um you know um with Deloitte participated in some surveys um where you know maybe 25 of large organizations have seven or more use cases implemented and are getting some positive outcomes out of them so that's still a small number um there are there are companies with literally thousands of use cases that have been implemented so you need a lot you need a variety of different types of Technology AI technology in your organization conventional machine learning deep learning even rules automation oriented Technologies um conversational AI which is increasingly based on deep learning but wasn't always so um a lot of a lot of different Technologies a lot of support from senior Executives a lot of investments in the area a lot of talent um and some sort of ethical framework to make sure you your AI work doesn't go off the rails um you need a governance structure at least a center of excellence um to kind of manage these activities and select tools and manage their careers of people and and so on so you know those are the kinds of things and yeah it takes takes a lot of money to make that work but um the companies that we found that are doing it seem to be quite successful definitely takes a lot of money takes a lot of resources so maybe let's unpack a lot of that and start at the beginning you know you discuss and the book on how to become AI driven that there are three strategic archetypes usually the AI organizations can adopt right uh can you explain maybe these archetypes if an organization is trying to get started wants to look at a road map a framework for developing their AI strategy what should that look like and you could do more than one but I think um it's good to sort of um have the Strategic discussions around the organization about well you know what what are we going to do with AI so there's the doing something new idea new strategy new business model um um important new products and services that you're going to offer um and that to me is the most interesting one although um you know it can be can be hard to do requires a lot of conviction on the part of of seeing your Executives and it requires a lot of people to understand the business strategy and AI so they can kind of make that connection um the second one is operational transformation in a not just improving things on the margin but doing dramatic change in how you do key aspects of the operation so one of our examples of that is shell where they're um using things like AI image recognition to inspect all of their refineries with you know drones and um image recognition to identify potential problems in piping and valves and so on so they can send a human out to look at what what appears to be a problematic area just a dramatic change in how you do maintenance and they're also doing some really interesting things with subsurface stuff and some probably not enough yet for the planet but some things also in the space of um of uh non-carbon-based energy energy forms um you know where to put charging stations on that kind of machine learning analysis and then um uh the third category is really trying to change customer Behavior where you are um uh trying to do something with AI to make life better for your customers and that tends to be insurance companies who used to you know they'd pay you if you died or if you got sick or if you crash your car but they sort of realized gee wouldn't it be better if we helped our customers not die not crash their cars you know not get sick um so it's early days of that but using kind of AI and Behavioral Science nudges to create healthy behaviors better driving behaviors Etc so those are the three things operational transformation is the most common um changing customer Behavior probably the least common so maybe let's talk about operational transformation because you know mentioning that shell case that sounds like a very ambitious project right what would be your you know advice here for an organization starting off when trying to prioritize use cases right because you know conventional wisdom always says you need to start small and trade from there right versus going for a big project that has a lot of complexity requires a lot of talent a lot of data a lot of overhead what would be your advice when organizations and leaders are trying to make that trade off I think there's you know the um what kind of opportunity is there for this and um uh you know you can assess how much money are we spending in that particular area now how much could we save if we were doing this a different way you know for Shell it was um takes years to inspect a refinery in the whole refinery in the traditional approach and it takes days to inspect it in the AI enabled approach so obviously that's a big big payoff and you know the maintenance and inspection is an important part of you know keeping things going and successful so um uh and then there's the you know how viable is it and you have to sort of think about since some of these things take a while you have to think well how rapidly is the technology maturing we talked about a company in the book not well known as a kind of medium-sized company about approaching a billion in revenues called CCC intelligence Solutions and what they do is enable this capability to for insurance companies to let their customers take a photo of their um car collision damage and get an immediate estimate for what it's going to cost to fix it and they started on that well before you know deep learning based image recognition was at the capability it needed before camera resolution was at the smartphone camera resolution was the at the level that it needed to be and and so on um so it requires some ability to sort of look ahead into the future and see how these Technologies are going to be developing okay that's really great and we'll definitely talk about what you think about generative AI here but maybe before uh we talk about and look forward to the future in the book you also talk about the human side of becoming an AI driven organization this connects back to our earlier conversation on their literacy and data culture uh you know succeeding in AI you mentioned is the most important actually developing The Human Side is the most important aspect of succeeding in AI not the technology side something we wholeheartedly believe in a data Camp as well can you explain maybe why this is so important in the context of AI what an AI driven culture looks like and how can leaders cultivate such a culture well I think you know traditionally an AI driven culture required a lot of you know professional data scientists um but increasingly you know in part because of generative Ai and Automated machine learning and so on I think we're going to be seeing an environment where it has to be a democratic pursuit in the sense uh we we need tons of people doing this even at Shell you know I think they've now um certified 5 000 of their Engineers as knowledgeable in AI you know you know they had some online programs that they recommended and and so on so Airbus did a similar thing in um uh getting their engine a lot of their Engineers up to speed so um I think if if you're starting from scratch now and you don't have that strong um human culture for AI I think you're much better off thinking about how do I enable a large number of people in my organization many of whom are already employees to acquire these skills and you know the skills themselves are changing um there aren't too many programs out there on you know what do you need to know in order to be a good citizen data scientist how do you use um Automated machine learning effectively even more I think they're going to be machine learning programs that have uh some companies have already announced them I'm not sure they actually exist yet but but the front end is generative AI so how do you tell a generative AI system what you want out of a machine learning algorithm and um you know how do you refine your prompts how do you interpret the output Etc so we're going to need some new programs of that type as the as these Technologies mature okay I couldn't agree more and you know you mentioned here generative AI let's dive deep into it you know of course Johnson the way I have garnered quite a lot of attention recently maybe discussed I haven't heard I haven't heard of any maybe discuss what the generative AI impact or landscape of impact looks like for large organizations Enterprises today how should they think about leveraging these Technologies well you know it's um potentially quite earth-shattering it's quite early days but um very wide range of potential use cases um now I think we're seeing the relatively less important ones you know writing blog posts or writing copy for ads or marketing materials um really easy to do that and um you know it's what's created is not going to be world class but it's going to be certainly a good input to an expert copywriter and will very much improve their productivity um I think the whole world of coding is going to change dramatically and I was I'm talking about this yesterday with some of my friends at Deloitte we were together doing a presentation for a big private Equity Firm and Deloitte is saying they expect improvements in coding productivity of like 50 percent um uh which means that we need you know far fewer programmers than we thought we needed and I think more and more the programmers are going to be like everybody else are going to become editors of um generative AI code rather than you know people who write it in the first place um any writer is going to become an editor more than a first draft writer so um that's um something that we have to deal with I think there's going to be a lot of applications in r d both kind of product design in terms of image related Technologies but also um pharmaceutical companies are starting to figure out that um just as uh generative AI can model text sequences it can also model protein sequences or molecule sequences or something like that and see which ones are most likely to be be effective so big changes coming they are a huge amount of change I think in Professional Services law is I'm going to change dramatically so um uh you name it I think they're going to be a lot of impacts and now what should you be doing well I think a lot of experimentation um May you probably want some production applications um but I wrote a piece last week about Morgan Stanley um the big wealth management company is using it they they fine-tuned trained on a hundred thousand of their documents and they're using it to provide their financial advisors with high quality answers to questions they have on you know what are they invent what are the firm's analysts saying about particular Investments um from a recommendation Point how best to do certain tasks like how do I establish an irrevocable trials for my grandchildren um uh what regulatory requirements have to be met in in certain domains so the hundred thousand documents is not that many they have to be well curated to make sure that they're accurate and everything but I think it's going to revolutionize you know what we used to call Knowledge Management which was an early enthusiasm of mine um making it much much easier to capture high quality organizational knowledge this is this is a very fascinating space you know you mentioned here really uh the breadth and depth of use cases that generative AI can unlock in a lot of ways we're really still early if you want to compare this maybe the last technology that was that could have this potential of change is maybe the invention of the iPhone for example you know where you had a massive amount of software ecosystems that could have unlocked new type of application software maybe you know we're still really early days right now in the generative AI space but maybe walk us through how you think the economy will be impacted in the next few years you know I'm gonna look at it from two different perspectives here uh you know a lot of people tend to feel worry and anxiety when these types of tools come in right um as they can be displaced jobs disrupt jobs and leaders as well are having to you know address create a vision for generative AI within the organization that incorporates The Human Side as well what would be here your advice for you know professionals today that uh feel this worrying anxiety and what would be your advice for leaders who are trying as well to communicate a vision that both incorporates humans and generative AIS part of the organizational strategy and future well you know I've been a pretty um persistent advocate of augmentation over automation um and I've written two books on the subject one with 29 case studies of people whose work is augmented um day to day called working with AI another a little bit more theoretical but with some examples back in 20 2016 um and I I I didn't really suspect that large-scale automation could be coming anytime soon I'm less sure of that now with generative AI um uh you know in content creation Fields it is just so um amazingly productive and effective already you say in the early days that I do have some concerns um but I think um for now you're nuts if you take the output of generative Ai and release it directly into the marketplace without lots of you know editing and trying not multiple prompts and so on so there's still a human component at the beginning and the end of generative AI process I don't know that that will always be the case as it improves in capability but right now that's true so the most important thing you can do is start to use these tools as an individual and be more productive with them so that uh you know people can say oh yeah we're getting some value out of this we didn't have to hire an additional copywriter because our existing copywriters were so productive in what they did um sort of a recessionary climate and some economies now um we're already seeing a fair number of layoff you you want to make yourself as valuable an employee as possible so um you know by all means tell your boss what you're doing um where we're doing some I'm doing some work with another guy on the whole idea of Citizen development and some people have used these citizen development tools even before generative AI they've used them to hold down multiple remote jobs at once as they could be so much more productive that I do not recommend I think it would long run be not a good strategy but um and so managers senior managers should be saying this is a very powerful tool um let us know what you're doing but we think it's going to be really critical to our future and we'd like you to experiment with it and I think they should performing communities of practice within the organization to say you know what have we learned about this um what works what doesn't work which tools work best for us um are there legal issues uh clearly you know you want to get your general counsel and your legal staff involved for things that you know go out into the into the world in terms of production there's some maybe some uh security issues ETC some ethical issues I um go to a fair amount about unilever's approach to ethics AI ethics in my all in an AI book and I got a message um from uh the woman who focuses on ethics primarily their yesterday saying that they're quite focused on generative Ai and what does that mean for their ethics approach so um you you I think they're going to be a lot of potential ethical issues around misinformation and deep fakes and the usual issues of transparency and bias and so on yeah I couldn't agree more this is definitely an interesting exciting uh but also wary space in a lot of ways that I'm looking at maybe Tom took cap off our discussion uh I know it's always it's always annoying to put someone on the spot to give predictions uh given how fast the generative AI space is is moving right what would be your prediction for this space in 12 months from now well I think there will be um generative AI front ends to almost every software product that we use you know we've already seen some announcements of these for at blow and HubSpot and um uh um Salesforce CRM and so on um uh more and more companies are going to say um command line interfaces or even point and click interfaces are going to be replaced by conversational interfaces we're going to see finally I think chat Bots that are really quite useful um it's not going to happen overnight because they're going to have to be fine-tuned trained and uh going to be have to be on high quality customer service content which many companies haven't you know bothered to curate in the past um I think we'll see almost every online activity that we performed search or on social media or um uh just you know internet e-commerce will be mediated by a conversational front end um I think we'll still have um prompts but I suspect within 12 months we'll have some alternative to prompts where a generative ai's system would sort of lead us into creating a prompt with a sort of intelligent front end as as a prompt Creator um I think there will be many many many different customized models fine-tuned models Millions probably in 12 months um so I mean just think about in the legal profession there's already there are already a couple of uh you know large language models trained for the legal field but then you think about okay um there are differences between U.S law European law UK law so they're going to be versions for different countries or Regions they're going to be different versions for types of real estate law within those regions like real estate law and even individual law firms will say Okay Alan and Avery a UK firm we're going to have the Allen and ovary UK real estate law model that nobody else has it's not exactly like ours so they're going to be lots and lots of these things and no matter what we do you want to take a vacation in Peru there will be a model custom designed for where do you go where do you stay how do you spend your time in in Peru that is very very interesting and very fascinating it's going to be a very interesting future in next 12 months and Beyond uh finally Thomas do you have any final call to action or final words before we wrap up today's conversation well you know I think the time to move on these things is now um I mean the the Morgan Stanley people have been working with open AI now for 18 months and um Deloitte has been working with open AI for 18 months on code generation so um there are some companies that are already ahead of yours um in this regard so you don't want to just sit back and wait for things to develop it it may be a hard thing to be a fast follower on so um start experimenting aggressively get your content in shape this is a degenerative AI space it's a AI space in general I think um you can start small because AI 10 tends to be kind of small use cases small level of functionality individual tasks so far but think big about what it's going to do for your organization and how it's going to transform it so think big but you start small if you if you need to okay thank you so much Thomas for coming on the podcast my pleasure thanks for having me foreign\n"