Latest / CXOTalk / Mastering AI: Designing and Implementing Effective Data Strategies
Transcript
- 0:00Today on episode #793 of CXO Talk, we're speaking about Data
- 0:07and a I. Our guests are Inderpal Bandari,
- 0:12the Global Chief Data Officer of IBM, and Anthony Scriffiniano,
- 0:18the former chief of Data scientists at Dunn and
- 0:22Bradstreet Inderpal, welcome to CXO Talk.
- 0:26It's great to see you. And please.
- 0:27Tell us about your work at IBM. I'm actually a four time Chief
- 0:31Data Officer. When I first became Chief Data
- 0:34Officer in 2006, there were just four of us globally.
- 0:38I was the first in healthcare and then the profession and the
- 0:41related professions like Chief Analytics officer,
- 0:44transformation officers, the century that took off and I
- 0:48happened to be fortunate enough to ride with it and I've done
- 0:51this job four times, IBM being the 4th and perhaps the most
- 0:55complicated. At IBM, my strategy, data,
- 1:00strategy has been to make IBM itself into an AI enterprise and
- 1:04then use that as a showcase for our clients and customers,
- 1:07because our clients look very much like us.
- 1:09So that's what I've been doing for the last seven years, 7 1/2
- 1:12years or so. And Anthony Scriffiniano,
- 1:15welcome back to CXO Talk. It's your good friend.
- 1:20It's great to see you and tell us about your work these days.
- 1:24Thank you very much, Michael. It's, it's great to see both of
- 1:27you. So as you mentioned, I was with
- 1:31Dunn and Bradstreet for quite a long time, over 20 years.
- 1:34Right now I'm doing a number of things and probably front foot
- 1:39is as a distinguished fellow with the Stimson Center, which
- 1:42is a think tank, think tank. I'll put in quotes because
- 1:46there's a lot of what I would call applied research or action
- 1:49research where they actually get involved in doing things, not
- 1:52just writing about them I've been involved with.
- 1:54Things that are called a I. The term's been around probably
- 1:58since the 50s but I've been involved with it in as it's
- 2:01become computational from its from its birth and and I know
- 2:06Interpol has as well lots of things going on in the world
- 2:09right now in terms of regulatory focus on a I as well as new
- 2:16types of a I becoming sort of the greatest new shiny object
- 2:20and everyone pays attention to them.
- 2:23And I I stand for the, the, the size behind it.
- 2:26What do you have to believe? What has to be true in order for
- 2:29you to do that thing that you think is so cool and and why is
- 2:32it better than what you're doing today and and what is the cost
- 2:36of it? So I try to ask those emperors
- 2:39new close kind of questions and that's the role I'm playing
- 2:42right now. So we're talking about data and
- 2:47AI. And I think where we need to
- 2:49start is when we talk about an AI data strategy, what actually
- 2:56is that Interpol, you wanna maybe take a crack at that to
- 3:00start? AI is only as good as the data
- 3:05that is used to train that AI, because AI has a training
- 3:09sequence and then an inference sequence.
- 3:12The training sequence has to do with seeing.
- 3:15All kinds of related data, so that it can actually then train
- 3:21itself to figure out what the right output is when it's shown
- 3:25an input that it may not have seen before.
- 3:27So if the data to begin with is flawed or low quality, the AI
- 3:33will not work effectively. It's the garbage in, garbage
- 3:36out, that kind of phenomenon that you would have.
- 3:40So they go hand in hand. And very often you think of
- 3:43people talking about AI, and if they haven't really looked at
- 3:48the data, but they embark on a data on an AI strategy that is
- 3:52going to be very high risk, it'll most likely fail because
- 3:55they'll have to go back and straighten out the data strategy
- 3:58first, just so that it's fit for purpose.
- 4:01Now when you say fit for purpose, what that means is.
- 4:05If you know what the business objective is that you're trying
- 4:08to serve, so it could be something quite narrow like a
- 4:11specific objective. It could be something like I
- 4:14want to understand what segments of my business.
- 4:18Should I try to try to expand to increase my top line then, in
- 4:24which case, if it's segments of business, you know, data about
- 4:27your clients, about your products, etc, those things
- 4:30become very important. You'd want to make sure that
- 4:32they are, that that data is of very high quality.
- 4:36And on the other hand, if it's something at a strategy level,
- 4:40which is what kind of what happened when I joined IBM, I
- 4:43mean, you know, IBM wanted to be a cloud and AI company.
- 4:46And to be a cloud and AI company, eventually we landed at
- 4:49the point that, well let's transform ourselves internally
- 4:52before we actually show this off to our clients and customers.
- 4:57That became a strategy that was enterprise wide and we realized
- 5:01that now while for instance not only do we have to make sure
- 5:05that our structured data is in order, but also our unstructured
- 5:09data because we are going to go after this.
- 5:11And transform ourselves into an AI company.
- 5:14So there are two, two aspects there that are relevant.
- 5:17One is at a strategy level when you're aligning, aligning to the
- 5:20business strategy or it could be more narrow to a specific
- 5:23business objective. Anthony, the challenge of
- 5:28aligning the data strategy to the business objectives is
- 5:34something that many organizations struggle with.
- 5:38What thoughts or advice do you have on making that work?
- 5:42You've seen so many different scenarios.
- 5:45You would really have to unpack what Interpol just said quite a
- 5:48bit to really get at the essence of it, and I did when I was
- 5:52listening to him, but he was using some terminology very
- 5:55carefully there. A lot of times organizations
- 5:59don't have one strategy, I mean make more money, you know grow,
- 6:02grow, you know fill in the blank, right.
- 6:05The the things that we learn in Business School, you can serve
- 6:08your shareholders, you can serve your customers, you can serve
- 6:11your employees. Kind of hard to do all of those
- 6:13things at the same time because very often optimizing for one is
- 6:18is less optimal for one of the others.
- 6:21So the strategy of which we speak.
- 6:25And we start to talk about AI has some very serious
- 6:30implications, these methods that we're talking about.
- 6:33And I should say that these days it's rare that only one method
- 6:38gets applied very often. There are many methods being
- 6:40applied simultaneously. There are some commonalities.
- 6:44So one of the commonalities is that the quality of the data has
- 6:49many dimensions, so truth. If if your a I is going to
- 6:52ingest data, it's going to probably presume it's all true.
- 6:55Well, all data is not necessarily simultaneously true.
- 6:59It may have been true at the time that it was created, but
- 7:03maybe not so much anymore at the time that it's curated.
- 7:07So how old is the data? How is it still true?
- 7:10How would you know that it's still true it before you consume
- 7:14it into an algorithm or an approach that presumes that?
- 7:17I love to say that. When we go to court, we swear to
- 7:19tell the truth, the whole truth, and nothing but the truth.
- 7:22That's because those are three different things, and those are
- 7:25three different ways to manipulate veracity or
- 7:27understanding. So when you get back to this
- 7:30concept of strategy, well, whose strategy?
- 7:33What part of the organization? Specifically, what objective?
- 7:37How would we know when we were successful?
- 7:40And asking those questions is very often a source of intention
- 7:44because the people in the room that all think they want the
- 7:46same thing. Realize they don't.
- 7:49And when you unpack it a little further, they realize that to
- 7:51get what the guy on the left wants, you have to get less of
- 7:54what the person on the right wants.
- 7:56And it's not pretty. So it's not really a technical
- 7:59problem as much as it is an alignment problem.
- 8:02And, you know, sort of getting everybody to agree on what they
- 8:06want so that we would know that this strategy of which we speak
- 8:09is actually what this a I of which we speak, is delivering
- 8:13really difficult. These roles like the Chief Data
- 8:17Officer, the Chief Transformation Officer, the
- 8:19reason these are CXO roles is because of what Anthony just
- 8:23said. You have to be part of that
- 8:25discussion. It's not so much like there is
- 8:28the concrete business objective. Sometimes you get into those
- 8:31situations where it's very clear cut, but more often than not
- 8:36it's a strategic discussion in terms of a understanding,
- 8:39clarifying, perhaps even adding to the business strategy.
- 8:43And then relating it back to what you're trying to do with
- 8:45data and AI, and unless you're in a position to have that kind
- 8:49of conversation and you also have the wherewithal to pull
- 8:52that off, you know that you won't be able, you won't really
- 8:55be successful. So that's why these are CXO
- 8:57roles, because it's really part of the negotiation that goes on
- 9:02to align the business strategy to the data or the AI strategy.
- 9:07Maybe I can just add a little bit to that, that the whole
- 9:10concept of being in the room. It's so important back in the
- 9:14day that the the goals and the objectives would come down from
- 9:18on high and that the folks with the the keyboards will just make
- 9:22it so you know, it doesn't work that way anymore and it can't
- 9:26really work that way anymore. And so it's so critical that the
- 9:30folks in the roles that Inderpal and I have had have a seat at
- 9:34the table understand what went into the ask and not just the
- 9:38ask very often what. Folks want and what they need
- 9:44are two very different things. And so without being arrogant
- 9:48about it, you you asked a lot of questions and you get at what
- 9:51they really needed in the 1st place, which is probably not
- 9:54what they started out asking for.
- 9:55You're talking about organizational alignment with
- 10:00business strategy, and at a high enough level, this is true for
- 10:05every business decision that needs to be made.
- 10:07And yet, when you hear people talking about a I.
- 10:11And data strategy, the conversation turns and very
- 10:15quickly to what kind of data do we need?
- 10:18Where do we get that data? What's the technology that we're
- 10:20going to use to aggregate and to manipulate that data?
- 10:23What kind of models are we using?
- 10:26And so now I'm confused because you're talking about one thing
- 10:30and I hear the entire world talking about something
- 10:32different. The world tends to focus on the
- 10:35hammers and the nails, and it tends to focus on.
- 10:38The the tools that are going to be used for the the, the
- 10:43purpose. If I come to your house and I
- 10:46say and if then you want to put an addition on your house and
- 10:50I've met with the architect and I understand your objectives,
- 10:53let's talk and someone else comes to your house and says I'm
- 10:56going to build you a beautiful addition and I'm going to use
- 10:59the hammer, right. You don't really care about the
- 11:02hammer, so of course the hammer is important.
- 11:04It's very important that we have the right data, the right tools,
- 11:07the right technology, the right people.
- 11:09So it's people, process, tools and mindset.
- 11:12All of those have to be aligned in order to get this right, but
- 11:15it starts with making sure you're focused on the right
- 11:17mission. Yes, that number changes that
- 11:20piece of, you know, aligning back to the business it's.
- 11:23So the way I would put it is, yes, no matter how promising the
- 11:27technology, no matter how dramatic the advance, etcetera,
- 11:31it doesn't let the organization off the hook for coming up with
- 11:36a sound business strategy and then of aligning these elements
- 11:40to that business strategy. That's still going to be very
- 11:43much needed, in fact maybe even more so than before as you try
- 11:47to go after these new approaches and methods.
- 11:50Be sure to subscribe to our YouTube channel and hit the
- 11:55subscribe button at the bottom of our web page and you can
- 11:59subscribe to our newsletter and we'll tell you and notify you
- 12:03about our excellent upcoming shows and guests.
- 12:05We have lots of them. So would you say that the
- 12:09business side is more difficult or harder to achieve than the
- 12:15data and technology foundations in your experience?
- 12:19I would say that if you take something new, so you probably
- 12:23want to draw the distinction between a mature technology and
- 12:26a technology that's more recent or Macent or emerging.
- 12:30And if you take something new like that like the latter then
- 12:34it's you know there. There is a tremendous amount of
- 12:37complexity on the technology side as well.
- 12:41So early on when getting into this game, you know when you we
- 12:46were working on on. AI for instance, at IBM, it
- 12:52became very clear as we went forward that there were four
- 12:55elements that had to move in lockstep, data, technology,
- 12:59workflow and culture. And those four kind of had to
- 13:03move at the same time. Otherwise the adoption was not
- 13:07going to be not going to be effective and.
- 13:11The technology piece for an emerging technology, so at that
- 13:15time, you know the cloud was emerging, there was a lot of AI
- 13:20techniques that were emerging, the deep learning stuff with,
- 13:24you know, GPUs and things like that.
- 13:26You have to make all that stuff work together.
- 13:28So there is a significant complexity in the technology
- 13:31piece, but there is also a significant complexity in the
- 13:34data piece and the workflow workflow piece and then
- 13:37eventually in the culture piece of the organization.
- 13:40The the stuff that we were talking about in terms of the
- 13:43negotiation, working with the C-Suite, you know, there's a lot
- 13:47of the cultural aspect that goes into, there are many
- 13:50organizations one could go into and you would essentially not,
- 13:54Anthony said. They would still want to give
- 13:57you a set of objectives and say here, go off implement this.
- 14:00We really don't want to hear from you about anything else.
- 14:03These are your marching orders. Go off and implement this.
- 14:06Well, that's the wrong approach when you're trying to bring in
- 14:09an emerging technology and use it to impact the business.
- 14:12Anthony, we have a question exactly on this topic from
- 14:16Twitter from Arsalan Khan and maybe you can share your
- 14:19thoughts on this. He says we when we talk about
- 14:22alignment, there's business strategy, enterprise business
- 14:26architecture, change management culture and now data strategy.
- 14:33All right, Anthony, so what's your prescription than to make
- 14:38all these layers work together and align sounds almost
- 14:42impossible? Almost impossible is a synonym
- 14:45for possible. So if you said it was
- 14:48impossible, you know, now we have to, we have to talk, right?
- 14:53I think that first of all, thank you for the question.
- 14:56From someone who knows that I ask a good question, I would say
- 15:00it's really important. That you start with the question
- 15:04with the objective. Everybody wants to jump to the
- 15:06technology. They want to jump to the the
- 15:08date of the deal. The the thing that we're going
- 15:11to the the, the, you know that there's there's two factions in
- 15:16the room. The one faction is focused on
- 15:18the the revenue, the growth, you know what's going to happen to
- 15:21the organization and the other faction is focused on, all
- 15:25right, let's get going, Let's start, you know, doing stuff,
- 15:28let's start cooking in the kitchen.
- 15:31I'm usually the 1:00 somewhere in the middle of those two
- 15:33saying let's make sure we're answering the right question
- 15:35here. And I, you know, I'm not slowing
- 15:38you down. I'm actually making sure we get
- 15:40done in a way that we don't fall over the finish line.
- 15:44So it is very difficult to get all those factions in the same
- 15:48place. Probably the most important
- 15:50thing you have to do is be able to listen to each other and not
- 15:53start immediately talking about hammers and nails or immediately
- 15:56talking start talking about what color we're going to paint the
- 15:59finished product, right. But you know, somewhere in the
- 16:02middle is, you know, why are we doing this?
- 16:04What are we not doing while we're doing this?
- 16:07Do we know there's a big difference between can we do it
- 16:09and should we do it. So what are we giving up while
- 16:11we do it? What about compliance, What
- 16:13about regulatory, what about making?
- 16:16How do we know that the data that we have is the right data
- 16:18to make the decision you want Just because you believe it and
- 16:22you have your confirmation bias and you found one or two pieces
- 16:25of data that support your hypothesis doesn't make you
- 16:28right. So we have to ask these
- 16:29difficult questions and there's a very fine line between being
- 16:33right and being dead. So you have to be able to ask
- 16:35them in a way that doesn't annoy, it can annoy them a
- 16:38little bit, but you have to annoy them just to the point
- 16:41where they don't kick you out of the room and and keep asking
- 16:44those, you know, help me understand kind of questions
- 16:46until we get to a shared understanding of what it is
- 16:50we're trying to achieve and the opportunity cost of all the
- 16:53other things that we're not doing.
- 16:55Interpol. But you're a technologist?
- 16:58So if this is strictly then a business issue of organizational
- 17:05alignment, why do technologists play such an important role in
- 17:11this discussion of such a foundational fundamental?
- 17:14Role. I think the best way to think of
- 17:17my role of people in similar situations is that of a
- 17:20changing. The catalyst for the change is
- 17:24the technology. The change has to be affected in
- 17:28the organization and in the business.
- 17:31So you have to be able to bridge those two to be able to do this
- 17:36successfully. So you know it's a
- 17:39transformation and the transformation typically has
- 17:42those elements that I talked about for what we do, what I do
- 17:46data, technology, workflow and culture and I'll give you one
- 17:50other thing, I mean it's. There is there is a lot to be
- 17:55done in terms of changing the culture of an organization when
- 17:58you try to bring bring about this change.
- 18:01What we what we saw at IBM when we pushed forward with our data
- 18:06and AI strategy was that the adoption of the platform was
- 18:11triggered far more by the bottom up measures that we put in
- 18:16place. So we actually had a team that
- 18:18was empowered to engage with other teams.
- 18:21That were working in the business, you know doing
- 18:25workflows, go to cash procurement, you know things
- 18:31like that supply chain. And so we have an empowered team
- 18:34on the technology side which was didn't really need to come back
- 18:38for direction or instruction but if they found a like minded team
- 18:41they could go ahead and and move forward with the transformation.
- 18:45We found that 85% of the adoption actually came from that
- 18:49part. As opposed to the top down path
- 18:51and so forth. So it really is all about how
- 18:54you effect the change. But obviously if the catalyst is
- 18:57the technology then you've got to be able to walk that walk as
- 19:01well. So but you you can't discount
- 19:04the other side of it, you have to really be the bridge.
- 19:07Like I smiled when you called into call a technologist and he
- 19:11very diplomatically didn't didn't respond.
- 19:14I think you can tell by that answer that you have to be much
- 19:18more than. Just an expert in the technology
- 19:20to get what he just said right in large organizations there,
- 19:25what's happening right now is a massive federation of data and a
- 19:31I capability. It's not like you go to the room
- 19:34where the people that know how to do that live and ask them to
- 19:37do it for you. Almost anybody can get these
- 19:41capabilities on their desktop. It doesn't mean that's the right
- 19:44place to do it, but they can start doing it there and
- 19:46everyone feels like they're an expert.
- 19:49Just like when we all first got, you know, I'm trying not to name
- 19:54a product, but I think I can say Harvard graphics or you know,
- 19:56like in the days even before PowerPoint where all of a sudden
- 19:59we could all, you know, lay things out on the screen, we all
- 20:02thought we were experts in design and layout and font
- 20:05selection and and all of that. And of course we weren't.
- 20:07And and there's an old joke where the punchline is death by
- 20:10PowerPoint. We we all know versions of that
- 20:13joke. And I'm not picking up
- 20:15PowerPoint. Federating A capability like
- 20:18that across an organization or across the world comes with some
- 20:23risk that those who really honor practitioners who know.
- 20:27What? The differences between what you
- 20:29can do and what you should do, who understand the implications
- 20:34of going down a certain path and the difficulty of changing
- 20:37course once you get too far down the path.
- 20:40They have to be able to hear what's going on so for.
- 20:42When Interpol is describing to work, well 85% of the time does
- 20:46require an organization that actually talks to each other or
- 20:50or at least talks up to people who talk to each other up and
- 20:53down. But you know that's not always
- 20:55the case. So you can't just throw
- 20:57everything out in the middle of the floor and say here you go,
- 20:59everybody play with this and do whatever you want.
- 21:01That will not work. That will end in tears.
- 21:04So you have governance, you have focus on these foundational
- 21:09pieces. What about the interface between
- 21:14the technology and what you're describing the whole world and
- 21:18and organizations by and large tend to focus on that technology
- 21:23piece. And so can you now maybe talk a
- 21:26little about technology management as it relates to what
- 21:31you're just describing and also selecting the right kinds of
- 21:35technologies and especially? Selecting the right kinds of
- 21:39data to match with the problems that you're trying to ultimately
- 21:43address. And I would add time that it's
- 21:45still relevant. I think I have a good example
- 21:47for you. When when the pandemic broke
- 21:50out, I don't think anybody was really expecting that.
- 21:55All of a sudden organizations shifted to almost exclusively
- 21:59working from home. There's laws about what data you
- 22:03can access from home and what data you can access at your
- 22:05desk. You have a different firewall
- 22:08when you're working in the office than you do when you're
- 22:09working at home. You've got developers that used
- 22:13to be Co located that are not Co located anymore.
- 22:17Organizations had to absorb all of that change while still
- 22:21trying to serve their customers, and in some cases failure to do
- 22:25so could have been life and death.
- 22:27So you know there's an urgency about this as well.
- 22:31You can't take forever to do it and you have to have good
- 22:34discipline in place. So that when the unexpected
- 22:38happens in the middle of the other unexpected that was
- 22:40already happening, you have the resiliency survive that and come
- 22:46out of that stronger. I'm not going to suggest,
- 22:50although I could that I BM is one of those organizations, but
- 22:54you know mature organizations that get it and do that.
- 22:58We saw a lot of organizations that weren't so mature not
- 23:01getting it in the middle of all that disruption.
- 23:04So it's it's just a very big question you're asking.
- 23:06The example of the pandemic actually was particularly
- 23:09instructed, I think it goes to your data AI questions of the
- 23:14earlier in the in the segment as well.
- 23:16So when the pandemic hit in terms of being able to run your
- 23:21business, for instance, make financial forecasts, make
- 23:25forecasts about your supply chain, about your procurement
- 23:28abilities etcetera, all the models that were in play were
- 23:32essentially useless. Because we have now embarked on
- 23:36a situation that was completely new.
- 23:40And so no matter what technology we had in there from an AI
- 23:44standpoint or a model standpoint, it had been trained
- 23:47in a completely different world. And that's Anthony, was
- 23:50Anthony's point, right? I mean, it may not be true now.
- 23:52In fact, it wasn't true. What was true though was if we
- 23:56were able to get the data, accurate data, pristine data.
- 24:02Into the hands of the people who were running those different
- 24:05departments along with an overlay of what was actually
- 24:09happening in the pandemic, you know, where COVID-19 was
- 24:12breaking, what were the incident reports in different areas.
- 24:15So if you could like geographically, then overlay
- 24:18that on what these guys were working on, whether it be
- 24:20financial forecasts or sales with, you know, they expected to
- 24:24close or procurement sites that were endangered, things like
- 24:28that. They could make something out of
- 24:30it and move forward with it. And then so that that's I think
- 24:33also an instructive example of the relationship between data
- 24:38and AI and how that plays out as it as things really unfold that
- 24:42you know that are truly unexpected.
- 24:44Let me draw first blood on saying something super nerdy.
- 24:48There's a concept. I call it decision elasticity.
- 24:51I kind of stole it from economics.
- 24:53But how wrong can you be and still make the same decision
- 24:56effectively? So you don't have to be perfect
- 24:58to make a decision. And Interpol is talking about
- 25:01training. There's an implication there
- 25:03that you have longitudinal data data from the past that you can
- 25:07project into a near term future that looks reasonably similar
- 25:12and you can measure the elasticity of your decisions.
- 25:15How wrong are they? And then if they start getting
- 25:17wronger and wronger to coin a term, then you can stop and and
- 25:21reexamine those methods. The problem is when you have
- 25:24something completely disruptive, there is no data and the most
- 25:27dangerous situation you can find yourself in.
- 25:30Is when the world is changing faster than the data that
- 25:32describes it. That's exactly where we were at
- 25:35that moment. You can't just throw your hands
- 25:37up and say, well wait, when you have five years worth of data,
- 25:39come back and I'll, I'll retrain everything and and we'll be good
- 25:42to go. You have to have methods in
- 25:44place that are effective in a situation where and that's what
- 25:49this environment taught us, that you can't just rely on one type
- 25:53of learning, one type of projection into the future.
- 25:58At that time, I was very involved with watching bad guys
- 26:00do bad things well. When there's disruption, the
- 26:03best bad guys, especially if they think they're being
- 26:06watched, they change what they're doing.
- 26:08If you model based on what they were doing, you're modeling how
- 26:10the best ones are no longer behaving out of a dangerous
- 26:13thing to do, right? But we know this.
- 26:15And so the the flip side of that coin is if you know that the
- 26:20environment changed such that the bad guys are going to
- 26:22probably try to take advantage of it.
- 26:25That many of them are probably going to do that unarthfully.
- 26:28And so you may be more easily able to see them as they run.
- 26:32You know, you turn on the light and the the, you know, the
- 26:35little creatures run away, you can see that.
- 26:38And so there might be an opportunity there along with
- 26:41that risk. So it's it's sometimes these
- 26:44these situations are, I would say almost never are they all
- 26:48bad or all good. There's always something in it
- 26:51that can teach you. There's always something in it
- 26:53that can make what you're doing better if you have enough time
- 26:57to breathe and observe what's going on and use the energy in
- 27:01the best possible way. It doesn't mean the bad thing
- 27:03will stop happening, but it may mean that you emerge from it in
- 27:07a better way because you you took that time to be more
- 27:10thoughtful about it. So we have a question from
- 27:13Twitter. Elizabeth Shaw says the issues
- 27:16you're describing are true of any business or technology
- 27:21transformation. Are there particular points
- 27:24issues that are more problematic for AI enabled initiatives?
- 27:30Can you kind of drill down into that?
- 27:33If you look at the advent of AI, the progression of AI, it's
- 27:37moved very, very quickly in the consumer space, but not so fast
- 27:42on the business space. And that's because in the
- 27:45business context, people don't trust AI.
- 27:49And they don't trust AI for multiple reasons.
- 27:52I mean, there's the, we talked about some of the issues about
- 27:55the data. So the robustness of the data,
- 27:58the quality of the data, the currency of the data.
- 28:01Then you also get into issues that have to do with the
- 28:04fairness of the algorithms. You know that the results they
- 28:08produce are going to treat people fairly if they pertain
- 28:11to, if the, you know, the data pertains to people, you have the
- 28:15issue of privacy being invaded. In terms of the algorithms
- 28:22discovering something new, you know there's this famous example
- 28:26of or infamous example of retail retailer, large retailer
- 28:32actually looking at the shopping shopping data, shopping patterns
- 28:36and shopping data and then inferring that.
- 28:40That this person is is pregnant and actually mailing their home
- 28:44and it turns out to be a young lady and you know, it was, it
- 28:49was it was really a complete invasion of her privacy.
- 28:51So those aspects come in. Then there are the issues around
- 28:54the job displacement and things of that nature.
- 28:57You know, if you're applying AI in the enterprise, there are two
- 29:01flavors of it. There's the automation flavor
- 29:04which has to do with when things are kind of straightforward.
- 29:08And you go from one step to the other and you know what those
- 29:12steps are and you can automate all that.
- 29:14So there's job displacement associated with that.
- 29:17But even on the decision making side, where the AI is actually
- 29:20helping make a decision, there's a decision maker in play and
- 29:23they have to trust it. They have to say, well, this is.
- 29:26Going to, you know, this won't displace me.
- 29:28So and extending that further, the executives as you put AI, we
- 29:33kind of know by now that AI has to be infused into the major
- 29:37workflows of the business, things like procurement, supply
- 29:40chain, etc. That's the kind of IP that
- 29:43doesn't get published in papers or patented or anything.
- 29:46Those are the trade secrets of a company.
- 29:49So they have to be able to trust whoever the vendor is of this
- 29:52software that this is not something that's going to
- 29:54disintermediate. Furthermore, the decision maker
- 29:57that's working with the system has to understand it.
- 30:00You know, years ago we I did this computer program called
- 30:03Advanced Scout that ended up being used by every coach in the
- 30:08NBA. And I remember the first time it
- 30:10had a counterintuitive finding. It basically asked the coach to
- 30:14play 2 backup players in a playoff game that they were on
- 30:18the verge of elimination. And he, you know, he was very
- 30:21concerned about that because he felt if I make, if I do this and
- 30:25I lose, I'm gonna lose my job and reputation as well in
- 30:28addition to the series. And we kind of solved that
- 30:31problem by letting him see the video clips of when those two
- 30:34players were on court. But that's the explanation
- 30:37piece, right? So if you tell a doctor,
- 30:40amputate the left leg. They're going to have all kinds
- 30:43of questions, OK, why amputate? What other options were
- 30:46considered? Why is amputation the right one
- 30:48for this patient? Etc.
- 30:50So explanation is another big part of it, and the AI systems
- 30:53today don't do a good job of all that.
- 30:55So those are the special aspects of AI and trust that come into
- 31:01play. I think that was a fantastic
- 31:03list and I won't vain to add to it, but I will suggest another
- 31:08dimension to it. So great question.
- 31:11Like how how are the A, I issues?
- 31:13What's special about the A I issues?
- 31:15I would say another one is that you have the opportunity to fail
- 31:19faster and at larger scale. There's a tendency once these
- 31:24sorts of systems are implemented, someone says, well,
- 31:26it's 99% accurate, it's 92% accurate, it's 87% accurate and
- 31:32you assume that means that 87% of the time the prediction will
- 31:35be right. Well, no, that's based on the
- 31:37past. In the future, right?
- 31:39Very rarely do we measure fast enough to stop every conceivable
- 31:47bad thing from happening. Interpol hinted at something
- 31:51which is an observer effect that people, when told what to do by
- 31:59a quote UN quote machine, will sometimes.
- 32:02Think they know better or not want to be told what to do by a
- 32:06machine and do something different just because a machine
- 32:10told them to do it? To prove that they can do
- 32:12something and not necessarily thinking it out loud.
- 32:15Like that Question I get asked a lot is, you know, what about
- 32:19someday when will people be reporting to robots or robotic
- 32:24bosses of some sort And you say, oh, of course not.
- 32:26I would never do that. And then the GPS tells you to
- 32:28turn left or right and you do. And Outlook tells you to go to a
- 32:32meeting and you go. We're already taking a lot of
- 32:34direction from automation. I won't call it a I necessarily,
- 32:39but from automation and the the the human fact of what we do as
- 32:44human beings to accept or reject that device is essential to get
- 32:48at trustworthy a I to get at making sure that we don't
- 32:52marginalize others that are already marginalized more
- 32:55because they don't have access to these technologies.
- 32:58This concept of good and and not good is is a very, it depends on
- 33:02where you're sitting sometimes whether it's good or not good.
- 33:05There are certainly lots of volumes, books, committees
- 33:11focused on trustworthy, A, I and explain ability.
- 33:14There's legislation as we speak being considered that will hold
- 33:19the feet to the fire of anyone who is implementing anything
- 33:22called a I. So you know to say that it's not
- 33:26being adopted by business. The adoption is lower, I think
- 33:32in some ways because of some of these human factors.
- 33:36It's not a lack of technology, it's a reticence to just push
- 33:42that button so quickly. And you know, technology will
- 33:45always outpace regulation. So you have to be careful where
- 33:48you could find yourself in a in a world of hurt where now
- 33:51they're coming after you because you use that technology that
- 33:53made a better decision. Good luck trying to prove that
- 33:56sometimes. This is a question from who Wong
- 34:00and he says we can sometimes measure the cost of implementing
- 34:06data solutions. But how can we measure the
- 34:11operational costs when a business decides not to
- 34:14implement certain solutions, such as governance or data
- 34:19quality? The opportunity cost, the cost
- 34:22of not doing something. And thank you way for that
- 34:26question that would. That's that's a big one and I
- 34:29think it's an important one. If we're going to decide not to
- 34:32do something, we should decide not to do it on purpose, not
- 34:37just because we got tired of arguing about it or because we
- 34:40didn't want to take the effort to get all the data that will be
- 34:43necessary to make that decision. So one of those annoying
- 34:46questions that I usually bring into the conversation.
- 34:50Is if we're going to decide not to do this because there's some
- 34:53other thing that we want to do and that other thing has been
- 34:56deemed more important, great let then let's make that decision.
- 34:59But let's understand the opportunity cost, the cost of
- 35:02not doing in. Many cases it does become clear
- 35:05cut because you might have regulations that then levy huge
- 35:09fines, for instance in the European Union GDPR for
- 35:13instance, if you don't have the right set up for governance and
- 35:16privacy and so forth. You'll be hit by a major fine.
- 35:20In other cases though, when you know they're making these
- 35:24decisions, they might choose not to do the governance of the
- 35:28data, but it'll end up reflecting in the actual output
- 35:32that's being produced, and then somebody has to go back and fix
- 35:35it. So keeping a keeping tabs of
- 35:38that, you know, I'm assuming here that you've lost the
- 35:40argument and they've gone ahead with it without actually, you
- 35:44know, then keeping tabs on that and raising that every time it
- 35:46happens. I think very quickly you'll be
- 35:49able to make a difference in the in the way people are viewing it
- 35:53because nobody wants, you know wants a disaster.
- 35:56And if the if the if what's if they've skipped that step which
- 35:59is which has major magnitude, you know, as sometimes that'll
- 36:03happen and the collaboration is the name of the game.
- 36:06So you just want to then keep an eye on it, warn people that this
- 36:10is going to happen. And every time it happens or
- 36:12even before it happens, you raise your hand and say look, I
- 36:15told you about this, now let's do it.
- 36:17This is from Jav Boshinov, who is a professor at the Harvard
- 36:22Business School, and he's also been a guest on CXO Talk, and he
- 36:27says this and Interpol. Maybe I'll ask you first, is
- 36:33there anything different between generative A I and more
- 36:38traditional A I? And how should organizations
- 36:42approach? This I think the best way to
- 36:45think about generative A I. The promise is that you can do
- 36:49things conversation. So just as you and I can have a
- 36:52conversation and we can discuss something and try to get get to
- 36:55some resolution. That's the hope.
- 36:57So now if you apply that in a large organization and say I've
- 37:01got some intelligence that can now conversationally help me do
- 37:04client support, employee support, my IT operations, etc.
- 37:08That's you know, hugely, hugely promising.
- 37:12On the on the other hand, I mean the way these systems work
- 37:15today. You know, the best way to
- 37:17understand generative AI that I've been able to get my mind
- 37:21around it is in a sense, each word is predicted, and then the
- 37:27word, essentially that word is fed back into the input and then
- 37:30the next word is predicted. It's almost like when you and I
- 37:33are talking, I'll sometimes do this.
- 37:35I'll go out on a limb, I'll start saying something and the
- 37:38thought hasn't fully formed. Usually I'll manage to come out
- 37:41of it. And but many times, you know,
- 37:44I'll end up with my foot and my mouth.
- 37:45So the generative AI techniques are essentially going out on a
- 37:48limb every time because it's, which is also why they're not
- 37:52always consistent with the response.
- 37:54You know, you might have the same problem to give you a
- 37:56different response because it's actually working off a
- 37:58probability distribution. So I think there's a tremendous
- 38:01amount of promise, but also a tremendous amount of work that
- 38:04needs to be done to address some of the issues that we've raised
- 38:08earlier. Anthony, Differences between
- 38:11Generative AI. And traditional AI and
- 38:14implications for the enterprise and pretty quickly please
- 38:17generative. AI is is making stuff that
- 38:20didn't exist before based on stuff that it observed.
- 38:22And that stuff can be taxed, it can be images, it can be
- 38:27anything that we as humans consume.
- 38:29So the the the challenge to it is that you look at all the
- 38:32stuff in the past and you you kind of compute on it and do a
- 38:35lot of math and then you generate something that looks
- 38:37like a human said it and a human didn't say it.
- 38:41And so when the world changes and the corpus of data that it's
- 38:45looking at didn't change fast enough, that nuance gets lost
- 38:48and we lose the ability to understand something nuanced.
- 38:53So if the if the purpose is to provide customer support based
- 38:57on frequently asked questions, or if the purpose is to
- 38:59summarize a whole bunch of things that you should have read
- 39:02and didn't have time, it's a fantastic idea.
- 39:05If the purpose is to to write some new thought leadership on
- 39:09something. Maybe it's a starting point, but
- 39:11it would be very careful when we consider that to be an ending
- 39:14point. Share final thoughts on advice
- 39:18that you would give to business and technology leaders who want
- 39:24to be more effective using data using AI.
- 39:29Interpol. You want to jump in with that
- 39:31one first. I've been doing this for the
- 39:33last 2025 years, starting from the days when I did that program
- 39:38for the NBA to now. And I've always felt that
- 39:41whenever I was doing it, I thought, oh, it can't get better
- 39:44than this, but it always seems to get better than that.
- 39:47And I think we're now in one of those moments where there is the
- 39:50potential and the opportunity to have a tremendous impact not
- 39:54just on business but also on society.
- 39:57And I think because of that implication that there are these
- 40:01major societal considerations as well, we absolutely have to get
- 40:05involved. And that would be my biggest.
- 40:08You know, advice to people either on the business side or
- 40:11on the technology side, you need to really get involved with
- 40:14what's happening here. And there's just tremendous,
- 40:16tremendous potential and it's there's never been a better time
- 40:19to be involved in data in there. Anthony, it looks like you're
- 40:22going to get the last word here. Number one, I would say ask why
- 40:26a lot? Why are we doing this?
- 40:28What do we have to believe? Why this data?
- 40:31Make sure that you understand before you jump into.
- 40:34That method with that data. Make sure that method and that
- 40:38data are in some way justifiable, like not only
- 40:41against what you intend to do, but against what you're not
- 40:43doing by by doing that instead. And then the second thing is
- 40:47make sure that you pay very close attention that how the
- 40:50environment is changing so that you don't get caught by the
- 40:54change that makes what made sense no longer sensible.
- 40:58And then the last thing is something I always advise, which
- 41:01is to be humble. It is extremely rare.
- 41:04When you know everything you need to know and have all the
- 41:07information you need without widening that circle and
- 41:10bringing in others that have some sort of expertise or some
- 41:13sort of perspective that you don't have.
- 41:15So inviting that expertise and that perspective is not a sign
- 41:18of weakness, It's a sign of great strength.
- 41:20With that, unfortunately, we are out of time.
- 41:23I just want to say a huge thank you to Anthony Scriffiniano and
- 41:28into Paul Bandari. Anthony, thank you.
- 41:31It's. Wonderful that you're you've
- 41:34been here again, and I hope you'll come back another time.
- 41:36Absolutely. Thank you so much, Michael.
- 41:38And Interpol, so honored that you joined us.
- 41:42And again, I hope you as well will come back and be a guest on
- 41:45CXO Talk again and on another date.
- 41:48Delighted to do that, Michael. Thank you for having me.
- 41:50And for those with unanswered questions, please, you know
- 41:53Lincoln and we can continue the conversation.
- 41:56Everybody, thank you for watching, especially those folks
- 41:59who just ask such great questions.
- 42:01You are such a smart and bright audience and we love your
- 42:07questions and keep watching. CXO talk, go to cxotalk.com, be
- 42:12sure to subscribe to our YouTube channel and hit the subscribe
- 42:17button at the bottom of our web page and you can subscribe to
- 42:22our newsletter and we'll tell you and notify you.
- 42:24About our excellent upcoming shows and guests, we have lots
- 42:27of them. Everybody, thank you so much.
- 42:29Hope you have a great day and we'll see you again next time.
- 42:31Bye, bye.