Podcast #51: Averting Bias in Future AI

Good day everyone, and welcome to another Momenta Podcast, and today our guest is Mike Flannagan who is currently Senior VP of Intelligent Enterprise Solutions at SAP, but he’s also got a really distinguished career of adding value and creating a lot of value, and being quite a creative thinker in the Connected Industry world. We’re going to expand a bit on some conversations we’ve had before, and hopefully get into some new areas too. Mike it’s a pleasure to have you join us on the podcast. 

Thanks for having me Ed. 

So, just as a quick level-set, could you share a bit of your background, and the context of your background, and what’s really shaped your view of what we call Internet of Things? 

I spent about 16 years with Cisco Systems in a variety of different roles, and a lot of that touched not only IoT, but the underlying data foundations into what needed to be done differently with traditional enterprise data, which is IoT data. I worked with, advised, or served on the board of several start-ups in the connected industry space, and for the last 2½ years I’ve had the privilege of being with SAP, first in the analytics business where obviously analyzing data, and preparing data for analysis was a day-to-day topic, and IoT data certainly brought some complexities to that, we’ve helped customers solve. 

Then for the last few months I’ve really been focused on a broader context than just what we were doing in intelligent technologies like analytics, machine learning, and IoT, but really looking at how do you integrate those intelligent technologies with the core foundational systems which run your business, to become an intelligent enterprise at least in theory, the output of lots of digital transformation initiatives. 

What does the term ‘digital transformation’ mean to you, and from your perspective I’d love to get your thought on the term, because it really has emerged as the new meta term in the industry. 

Well, I think the most important thing is, digital transformation a lot of times ends up in a conversation about the technology, and I don’t think it is about the technology; it is about the fact that technology allows you to solve problems that you’ve been trying to solve in other ways for a long time. My go-to example for that is manufacturing quality; if you’re in manufacturing, you’ve been finding ways to improve the quality of what comes off of your manufacturing line, for as long as you’ve been in business, at least hopefully. The idea of digital transformation is the application in that case of some technology to the age-old problem of how do you improve manufacturing quality, and the focus there has to be on, ‘Can I achieve a better business outcome, by making use of some newer technology?’ 

Certainly, at the first time if you think about IoT and machine learning that a manufacturer has gone through this cycle of, ‘Can I apply something new to help me solve an old problem, and get a better outcome?’ This is just I think an age where the technology is moving so fast, that there is a world of opportunity for things I might apply to that old problem. 

How do you go about assessing and prioritizing potential opportunities for, whether we call it digitalization, or applying advanced technologies; from the perspective of a big software company like SAP, when you look at the customer’s problems there’s almost no limit to where you can focus. What are some of the key factors that will impact your thinking in terms of prioritizing what comes first, and where to focus? 

I mentioned a second ago that in my view it’s not about the technology, it’s about the business result and the business outcome you can get. Those business results and business outcomes are in a lot of cases very specific to the industry dynamics that are in your particular industry, and so some things I think have to be looked at from a very industry-specific lens, or very process-specific lens which then makes it specific to your industry. There are other things though that I think can be said more broadly across industries; every customer I work with is trying fundamentally if you get through all of the buzzwords, the process language, the technology language, and the industry specific language, they’re trying to solve one of three problems… 

• How do I make more money? 
• How do I spend less money? 
• How do I stay out of jail, said differently, how do I reduce my corporate risk? 

Those three things, risk, top-line growth, or bottom-line savings, should be the lens, the filter for every business project whether its digital or not. When you think about digital projects, I think the questions don’t really change from more traditional projects, but sometimes because we’re so enamored with the technology, we’re so enamored with the idea of doing these really interesting things, with really interesting technologies, we lose site of the fact that in the end one of those three outcomes has to be improved by the project. I either have to have better ability to grow my top line, a better cost profile, or a lower over-all risk exposure. So, that’s the filter I would use to look at every project, and then if you think about the combination of those three, or the potential combination of those three, what I’m looking for is really the best return on my investment. 

Ultimately there’s a financial calculation which comes into play there. Of course, the challenge with traditional industries is that when you go into newer opportunities, you have to calculate a return on a business line that for instance may not be established, or there may not necessarily be a lot of history for a particular company. So, when you think about helping to define a business case for a new initiative, what are some of the key considerations you’ve seen work well for a company looking to digitize processes, or offer new products they haven’t had experience in before? 

You know, I think if you look Ed, what we’re talking about now is digital transformation initiatives, or the kinds of things we’re thinking of when we think about apply IoT, or applying machine learning, or applying advanced analytics in some way; all the buzz around that, and automation as well, things like robotic process automation as an example. Those things right now are kind of en vogue, they’re all the rage, it’s what everybody’s talking about. If I reflect back 10 years, maybe 15, BTO (Business Process Outsourcing) was what everybody was talking about, how do I take advantage of labor arbitrage, lower labor rates in offshore markets, to take well understood processes and move them to somebody who can do them better, faster, cheaper?  

There are a lot of parallels that I believe you can draw between those two things. In both cases I’m trying to figure out how to do better, faster or cheaper, I’m trying to figure out how to get more cost squeezed out of a mature process, I’m trying to figure out how to free-up resources so that I can focus on initiatives that help grow topline in some other way, or I take some of the risk that I have of doing anything in-house that I’m not very good at, and I give it to someone else who’s better at doing it, because it’s all they do all day. Coming back again to grow the top line, reduce the expenses on the bottom line, or reducing my overall risk. 

However, the one lesson I watched a lot of companies learn about business process outsourcing is, if you don’t deeply understand the process internally, you will have a very rocky road trying to outsource that process to someone else. I think the same is true here, if you are thinking about a digital transformation initiative, and you’re thinking about, ‘Where do I apply sensor data to improve this process?’ ‘Where do I apply machine learning to help me predict an element of that process?’ If you don’t deeply understand that process already, you’re going to have a very rocky road applying technology to help that process outcome improve. So, make sure you start with processes that you deeply understand already, because that’s where you’ll have the best ability to understand where to apply these technologies, in order to get a better outcome. 

And the question which comes from that, is that when you understand processes, you’re really looking to bridge a couple of different types of domain knowledge One is understanding business specific context and essentially how processes operate and the dynamics there, but then also being able to translate that into data which is going to be relevant and measurable, and I always like to understand better how one can appropriately determine what data is best to measure, and how you can benchmark for instance goals to measure the success of a project where you have a measure of automation, or digitalization that’s coming out of that? 

Well, I think you’ve touched on the key point which is, you have to understand what it is you’re trying to improve, and the business value of improving that, in order to attack the right place. So, if you think about again the example of manufacturing quality, there are lots of things I can improve in the production of a product, and any one of them I could measure on defect rate. It doesn’t matter what industry you’re in, it doesn’t matter what you manufacture, whether you’re printing books, or you’re making televisions, whatever, I can measure the product that I intended to produce, and the product that I actually produced, and identify defects. However, it is very industry-specific, very company-specific, in many cases very product line-specific, the customer impact of a specific kind of defect. And so, as an example, if I’m in the business of printing books, does it matter to my end-consumer if the text on the book jacket, or the cover of the book is crooked? Yes, that probably matters quite a lot, it looks bad on the shelf, it looks like you worked sloppy in the way you did it, and my assumption is your product is not a very high-quality, and so I probably don’t want to read it and I won’t pick it up off the shelf. 

On the other hand, if I print books and the manufacturing defect is that when the book is packaged, I sometimes put it in the wrong sized box to be delivered, does my customer really care about the size of the box the book comes in? They ordered a book and they wanted it to show up on their door, and it showed up on the door and the box was three inches longer than it should have been, and did anybody even notice? So, if I don’t know anything about your customer, if I don’t know anything about your business, if I don’t know anything about your industry, I simply say I’m going to do a raw count of the number of books that are placed in the wrong sized box, I’m going to implement a digital project which costs me half-a-million dollars, and at the end of it I’m going to reduce the defect rate of putting the book in the wrong sized box, and celebrate that. 

In the end your customer doesn’t care, they don’t pay you any less if you put it in the wrong sized box, assuming the boxes are of equal cost there’s no benefit to the bottom line from doing it, and I didn’t really have any risk that I introduced by putting it in a slightly larger box. And so, I’ve got this digital project that I’ve done, I’ve celebrated it, and we’ve had a team get-together where we had a cake, and everybody got excited about the accomplishment, and in the end, you have actually moved an inch at all for the business. So, start with things you understand, start with things you can measure, start with things that matter in terms of the overall customer experience, and the outcomes you’re trying to drive; whether that’s better topline growth, better bottom line results, or lower risk, and make sure you are doing things you understand not only how to improve them, but that it improvement matters, and it can be translated to something which is tangible for your organization, whether that’s a net promoter score from a customer, or top-line results/bottom-line results. 

I’d love to get your perspective, since you’ve approached the industry from the angle of a technology vendor, how do the technology vendor effectively work with the customer to help them see how to change business process? And, potentially overcome objections that say, ‘I’m running a manufacturing company’, and saying, ‘Well, you guys just do technology and we love your back-office software’, or, ‘We need your networking technology to manage our internet traffic’, but how do you understand what we need as a business? I’d love to get your perspective of how tech companies have been able to lead transition toward digitalization, and true connected industry, but as an outsider what are some of the advantages, and maybe some of the challenges in helping customers in different industries to take that journey? 

Well, I think as a customer-consumer of technology, one of the things that is most challenging which you really need to pay attention to is, not all companies who are put in the technology company umbrella are created equally. There are companies that are entirely focused on the technology itself, and the application of that technology is really what they understand, they understand the domain, let’s just say its machine-learning – they understand the domain of building the algorithms, and they understand the processing required to do deep learning, those sorts of things. Those companies that are technology-focused, technology-oriented, are different from companies which are business process-oriented, or they’re oriented towards a business focus and use technology as a means to the end of delivering the outcome of the process.  

Not to turn this into a commercial for SAP, but I would submit that whilst SAP is certainly lumped into the high-tech sector, that what SAP’s real focus is, is business process. Enterprise resource planning is ultimately very much what the business was built on, it’s the heart and soul, its where the company has grown up, and there’s lots of technology that SAP now brings to customers, but it’s all done in the process-orientation, and it’s all done in the orientation of your industry, your business process, and how do we then apply those technologies to the process? But rather than starting with technology it starts with an understanding, a deep understanding of the industry and the business processes within that industry. 

So, as a consumer, if I were CIO and I were thinking about digital transformation, or any other C-Suite thinking about digital transformation projects, the first question I would want to ask is, ‘Who really understands my business?’ ‘Who really understands my industry?’ ‘Who really understands my processes?’ Then we can figure out how those people, either on their own, or together with partners were technology-oriented can apply technology to help with that process, but if all you understand is the technology and you do not understand the business, it can be very tough to apply the technology correctly. 

Yes, and of course, technology is just one part of a business transformation, right? Because you need to be able to map potential business models, and the changes which may be required to adopt for instance a new go-to market model, or even realign an organization to a different business process, do you have any thoughts how organizations have been able to effectively adopt digitalized business processes, where there may be a pretty significant reallocation of where labor, where human involvement is needed in the organization. In other words, what I’m referring to is when you start to automate processes and people’s roles change, what are some best practices you may have seen to help guide the types of transitions that may incorporate some pushback or disruption in an organization? 

Well I think you framed the topic very well and danced all around the word that I’m dying to say, without actually saying it. Culture is a significant barrier to the success of transformative projects of any kind, and that includes digital transformation. Attitudes of the people who have to trust the technology, and trust the post-transformed way of working, and if you fail to take into account the serious importance of transforming the culture, transforming the way employees approach things, the way they think about things, the way you are rewarding and recognizing people, then the technology cannot get you there.  

I think the difference between now and 10 years ago when you look at something like machine learning, the mathematics behind what’s done on machine learning are not new, the statistical algorithms that are used for advanced analytics are not new; most of them have been around longer than you and I. What is new is the processing power that’s available to apply those things to ridiculous amounts of data, at an extremely fast pace, so that we can apply those things to enough data in the right amount of time, to get a result that matters, in a time that matters.  

The technology is there to do amazing things, but if your job for 20 years Ed has been figuring out when to push this button, and I show up tomorrow and say, ‘You push this button when the light goes on’, and that’s it, that’s all the change management I’ve done in the organization, what are the chances that you’re going to believe the light coming on, over your gut instinct which has been developed over 20 years of experience? And if you don’t figure out how to overcome that cultural barrier to adoption, you can have the best run, best delivered, most impactful digital transformation initiative in the world, and it will fail to deliver the business value, because it won’t be adopted, it won’t be trusted, and people will look for excuses to find fault with it. 

In terms of best practices obviously I’m showing my hand a bit here, that having a parallel cultural initiative, change management initiative within the organization to the technology initiatives that are going on digital transformation, is very important. The other thing I touched on briefly is rewarding and recognizing people; do I reward and recognize you for making a diving catch based on your gut instinct, today? Well, if I do, I’m reinforcing that, I’m incentivizing you to look for things at the very last minute, dive and catch, be the hero, and I’m giving you bonuses, and performance awards based on that behavior. If I continue to reward and recognize you on that, wow I’m doing a digital transformation initiative that’s meant to introduce the idea of predictive intervention… I’m not really doing what I need to do to change your behavior, to incent you to accept the earlier warning signs and go intervene earlier in the process, to prevent the crisis, because I’m still giving you reward and recognition to recover from the crisis. 

So, those sorts of things are just as important as the technology. In the end, there is no generalized artificial intelligence today which can replace a human being in the workplace, it may be a reality someday, but we’re a long way away from it; which means that every digital initiative we do for the foreseeable future, is going to rely on the human-machine interaction, and that means the human has to be considered in the project, as well as the machines. 

That’s a great insight, and again you’ve hit on the word that I was dancing around, ‘Culture’. We find that moving these biological machines, I guess meat machines is what we call them, they are the most difficult machines of all to manipulate! 

I’d love to back up a little bit and ask you a bit about a high-level view of the market. You’ve been focused on connected industry and digitalization of industrial sectors for a while, I’d love to get your sense of where are in the market. I know a few years ago there was a lot of hype, the excitement cooled off a bit, there was a period of a little bit of disappointment; it seems now there’s a much greater interest particularly amongst some sectors that are lagging, for instance construction and real estate being forward thinking in adopting technologies. I’d like to get your sense of where are we in the broader market evolution, and whether there are any notable used cases, companies or industries which stand out to you? 

Well, the maturity I’ve seen over the past year to 18-months as I’ve worked with customers and met people at industry conferences and heard their stories, has been that we’ve gone from this view of digital initiatives are separate and distinct from the core business, to realizing the digital initiatives are core and essential to the business. By that I mean, I think there’s a broader realization that there’s not going to be a world in the future where I have ecommerce over there, and [0:26:30.6] commerce over here. Those two things are not separate, they’re not distinct, they for retailers are coming closer and closer together, the lines are getting more and more blurred.  

If you look at Amazon five years ago and what people were saying, it was the death of retail, all the stores were going to be closed, retail is going out of business, and if you look today at what Walmart is doing as an example, and by the way, this is not based on any working I’ve done with Walmart; just to be clear, this is my outsiders view from watching publicly available information about what Walmart is doing, if you look at what Walmart is doing, they have figured out how to leverage their brick and mortar infrastructure for things like in-store pick-ups which provide them with a different way of operating than Amazon, where everything’s delivered to your door.  

But if you look at their results, if you look at the financial results, what they’re doing seems to be working, it doesn’t suggest the death of retail, it doesn’t suggest Walmart stores are going to start closing, because everybody wants everything shipped in a brown box to their house. I also don’t think that means what Amazon is doing is not working, but they are different business models, and figuring out how to leverage your existing assets, your existing strengths, what you already do well as a company, and add to that augment it, surround it with new things which are well connected to your strengths; that’s a little different to what we heard a few years ago where it was, whatever you’re doing in the old world, ‘Oh that’s just old stuff which we’re going to throw away, here’s this new world of digital’.  

I don’t see as much of that sort of, everything is green field thinking which I saw a few years ago, what I see now from customers is, realization that they have to leverage the strengths of their existing core business, and tie digital initiatives very closely to that in order to achieve not only a digital transformation, but a digital transformation which feels real and genuine to their customers, and leverages the strengths they have in the competitive advantages, that they’ve been working to build for years and years. 

That’s a pretty significant transition, I think it really reflects a maturing of the perception of technology from the standpoint of business people, and the way they are looking at their industry is not the zero-sum game I think that was the initial perception, particularly retail is a great example. What about from the technology standpoint? Are there some technology dynamics, trends or enablers you can point to, which you feel are having an outside impact on the conversations? 

Well, these days the thing every customer that I talk with is excited about, everybody I meet at a trade show or a conference, every analyst I talk with at Gartner, it’s all about AI, and it’s all about how to apply AI to various parts of their business, their business processes. And of course, the things that are happening in AI are also pretty exciting in their own right, from deep learning, hyper learning models, I think there are things that are also happening to mature that space; as an example, a few years ago if I built a machine learning algorithm that decided whether or not to admit you to a college, if three years later you sued me and said, ‘Hey, that practice was discriminatory. You used information in that model which you shouldn’t have used’, most organizations couldn’t have gone back and produced the model they were using three years ago, because they refined the model, they continue to provide training data, and they refined the model and couldn’t go back and provide you with a clear traceability to how that model, that algorithm made the decision that it made. 

The same is true in any other application, I just happened to choose college admissions as an easy example. That lack of traceability, that lack of transparency is problematic, because when we rely more and more on artificial intelligence to make decisions, you know, we make mistakes as humans all the time, and we shouldn’t expect the machine to be perfect, there are going to be mistakes that are made by artificial intelligence, there are going to be decisions that are taken which are not the right ones, and the important thing is for us to be able to understand what happened, and then learn from it, and not to expect more from the AI than what is reasonably possible, but the maturity of being able to version models, the maturity of being able to go back and find the model that made the decision three years ago, and understand why it made the decision that it did, so we can understand whether it was correct, or incorrect, is an important part of the maturity of artificial intelligence.  

In the same way that if you have someone in your organization making hiring and firing decisions, and they fire an employee for a reason that they shouldn’t have been able to fire an employee for, you go back, and you hold that person accountable. There’s a conversation, there’s training, there’s coaching that happens about how to make that decision better next time, and we need to be able to do that with our artificial intelligence and the models we’re using. I’m excited to see that kind of maturity coming, and I think what that means now, that technology because it is becoming more mature, and getting some of what we would call probably enterprise features, there’s an opportunity for us to do more with it at scale and enterprise, knowing that it doesn’t leave us exposed because of a lack of enterprise capability or functionality. 

Are there certain skillsets you find are in short supply when it comes to AI and machine learning, which you see a critical to ensuring that adoption will be effective, and appropriate? We hear a lot about the shortage of data scientists, and people with really advanced skills around machine learning. At this point it seems to be a bit of a constraint, and you get the baby with the hammer syndrome where people want to try out technologies almost everywhere, they can find them. But how do you ensure if you’re a company looking to incorporate some of these technologies, and as a software vendor, what can a company like SAP provide to ensure that there’s the best match between the technology, and the skills and capabilities and goals of the organization using it? 

Well I think obviously a business understanding, there are a lot of people being churned out of university who have an understanding of Hadoop, Spark, Python and all these great software and technology elements that can help do incredible things. But to the discussion we had earlier, if they’re not able to understand the business context, how am I going to take somebody who deeply understands machine learning, but knows nothing about retail, and ask them to apply machine learning correctly to help me do a product assortment? If you don’t even know what a product assortment is, we’ve got a problem regardless of what technology you’re able to bring.  

So, that industry understanding and taking people who have it, and teaching them the technology, I think that needs to be given as much consideration as hiring the best and brightest who are coming out of MIT. because the best and brightest who are coming out of MIT deeply understand technology, statistics, and mathematics, but they may lack that business context, and I think as we look at artificial intelligence in particular one of the concerns is the robot taking over the world, aside, the robot taking my job which I think in a lot of cases is a real concern. We have tasks which have historically been done by humans that now are going to be done by machines, and where does that lead the human?  

What a machine doesn’t have is the depth of understanding of the contextual things that are outside of the dataset that are used to train a specific model. Artificial intelligence is very domain specific, and so the idea of having the people who may be otherwise displaced by a certain set of things, that will be done by artificial intelligence instead of by a human, the opportunity there is to leverage the knowledge that they have of all the contextual things around the edges, to work together with the algorithms, and make overall better decisions. So, I think that’s one, and that really speaks to me of a need for us to focus more on workforce retraining, and reskilling of an existing workforce, so that we don’t run into this problem where the machine is suddenly doing what the human used to do, and the human now doesn’t have a seat at the table. But rather we change the look and feel of that seat at the table, so that the human and the machine work together to do things that deliver the best results. 

So, that’s one. And two, this one for me is really key because of the fact that I spend most of my time in front of customers and partners, if you understand data, if you understand whether it’s how do you use sensors to generate a bunch of data in a connected context, or its artificial intelligence, or its analytics, or master data management, and you can’t tell a business relevant story… you are far less valuable, because in the end somebody has to go in and talk with the CEO and the board about how we’re going to make more money, how we’re going to save a bunch of money, or how we’re going to reduce our risk. That’s a business conversation, it’s not a technology conversation, so that storytelling skill is one that is so important, and I think often overlooked in technology-oriented degree programs. 

That’s a real problem because if you churn out people from masters programs, or PT programs, who really understand the data, the mathematics, the science, and the statistics, but they don’t understand how to tell a business-relevant story, they may be applying that technology in a great way, but they’re never going to get funding for their project. 

Yes, you’ve hit on a topic of a really interesting book, Daniel Pink’s Whole New Mind, where the sub-text is that right-brainers will rule the future, because you can automate so many engineering-type tasks or wrote tasks, but that ability to create a narrative and connect with people on the human side of course, that at least for the time being you can, we’re not at the point where virtual assistants are going to displace true connection with people! 

I’d love to get your thoughts looking forward a bit, were there some opportunities and used cases, and potential of some of the emerging technologies that could really change the way that we think of different types of roles? I know in our prior conversation you were talking about natural language processing, but as you look forward, what gets you excited and really optimistic about the creativity that could be unleashed by the potential of new technologies? 

That’s a broad topic, and I could answer that in a lot of different ways, but the thing that immediately came to mind is, in the end businesses have shareholders, and their shareholders are thinking about their investment and the return on their investment. Every project that involved sexy-flashy technology has to be able to relate back in the end to that objective, because that’s what the board is in place to make sure happens; the board hires the CEO, the CEO is there to make sure that that happens. If you don’t think about every technology, every technology project, every new facet of technology in the context of how does it actually help me do something better… and I haven’t given any consumer examples, of course there are a bunch of those as well, but if you think about it in the enterprise context, shareholders want return on their investment, and if that comes from the application of better technology, great, but it can’t be technology for technology sake when you think about how to leverage technology in the future. 

So, the thought behind that for me when you asked the question was, I’m not sure that the technology is what’s really opening up the future possibilities, as much as it is the intersection that’s happening of technology and business. Historically things like machine-learning was in the realm of mathematicians and statisticians, and I’m talking not the last couple of years but further back than that, largely academic applications and things that were not necessarily business relevant. What’s really exciting in the trend that I see which I think is going to be the most impactful over the next five or ten years, this is our C-Suite and board level conversation about what are we doing with IoT, what are we doing with machine learning?  

The fact that those conversations are now board-relevant conversations, means there’s an opportunity for that technology to intersect the business at the level where it can make a huge impact on shareholder value. That’s exciting I think probably more than anything else, that conversation is now happening at the right level, to have real significant impact on global corporations. 

And as you’ve mentioned, I think it’s not an evolution of technology, but it is this intersection of awareness and a bit of a change in C-level culture as well. I think that’s an important point that you’ve made. 

What about concerns, obstacles, and risks ahead? What are the things that keep you and your clients up at night, tossing and turning, sleeping like a baby and waking up every two hours crying!? 

Well, there are a couple of things, I don’t know how they would have otherwise come up in this discussion today, but now that this conversation’s turned a lot to artificial intelligence in particular, I just can’t let the conversation go without making a couple of points; one is that algorithms are trained using data, data comes from the past, things that have happened in the past that you’re using to try to predict the future to some extent. In the past there’s been in companies a lack of diversity, a lack of inclusion, a lack of people who don’t look like you or me if we’re the ones doing the hiring – getting the job, there’s been a lot of bad behavior in the past, and one of the things that keeps me up at night is that if I take historical data about salaries, and I apply that data as training data to an algorithm without the awareness that historically many companies were guilty of underpaying women, minorities, and just let the algorithm make the decision about how much people should be paid, I will take the bad behavior of the past and turn it into systemic future decision-making, in a way that will get me a very-very bad result in terms of how my workplaces are viewed by people, who have historically been underpaid, or not paid fairly, not paid evenly with their other colleagues. 

We have to apply that human understanding of the failures of the past, to how we train our algorithms to help us make decisions for the future. We cannot allow that to happen, we have to make sure that algorithms that are written, are written specifically to remove bias from the hiring process. They’re written to remove lack of equity in compensation. Those things are very important, and one of the things that concerns me, one of the things that keeps me up at night is for all of the conversation that we have as a society these days, about diversity, inclusion, equality, and the importance of those things. If you look at the board of director’s makeup of the Fortune 500, it is still largely a very non-diverse group, and I’m not sure that this topic that I just touched on gets as much passionate dialogue at the right level that it should.  

There’s a lot of thought about how to improve the bottom line, but in improving the bottom line we also have to make sure that we’re behaving in a sociably responsible way. We know that there have been inequities in the past that cannot exist in the future, and we need boards of directors and C-Suite executives to institutionalize equality and fairness, and diversity and inclusion in a way that the historical data won’t do without some intervention. And so, I think that has to be a topic of conversation, it has to be something we’re not afraid to talk about, and it has to be something that we are all committed as executives and boards of directors to address. 

I think that’s a great point you’ve made. Again, if you look at what’s transpired historically and try to build a predictive model based on data that has really just emerged through practices that may have inherent of undiscovered flaws, you kind of miss out on the opportunity. I think you’ve raised an interesting point about data; the data about investing in companies that are led by female founders for instance, there have been several studies that show proportionately companies that are led by female founders are under-invested in the venture world. However, they deliver outsize returns, statistically they tend to perform better as a whole proportionate to the investment. Do you think there are ways to use the new technologies, to identify potential say disconnects in identification of value, maybe it’s almost like an application of sabermetric or money ball to corporate governance? Do you think there might be a way to address this, harnessing existing technologies? 

I think so, absolutely. If I look at some of the things that are done with artificial intelligence to remove bias in the process of selecting the right résumé from a pile, it is absolutely human nature, and I cannot fault anyone for this because its down there deep somewhere in the stem of our brain, that if I can’t pronounce your name, I am less likely to pull your résumé from the pile and call you for an interview. Because it’s embarrassing for me that I can’t pronounce your name, its uncomfortable for me and so I’m less likely to do it. That means that people who come from a different background, either speak a different native language, or live in a different part of the world than where I grew up, who have a very different name and its one that’s uncomfortable and makes me not want to pick up the phone and call you, are less likely to get a phone call. They’re less likely to get the job, obviously 100 percent of people who don’t get an interview don’t get the job! 

That bias can be removed by allowing an algorithm to look through a résumé, and decide who comes in for interview based solely on criteria like their skillset, which is really what we should be looking at, who’s going to be more effective at doing this job, not, ‘Do I feel comfortable pronouncing your name?’. There are things that can be done like that all throughout an organization, undertaking a look at the processing power, and the algorithms exist now to look at compensation from a hundred different views. Compensation in most large companies is made up of a mix of your base salary, bonus, stock incentives, those sorts of things, all of those can be sliced and diced in a million different ways by machine learning algorithms, to point out where equities exist, based on… maybe not originally based on, but along the lines of bias against a certain group. 

It is a pretty well-known fact that in most organizations' women make less for doing the same job than men make, for doing that same job. Most organizations in their C-Suite and their board would say, if they were asked, ‘That’s not fair, that’s not right, we should fix that’. But I think in a lot of cases the question is, ‘Okay, what are you actually doing to fix that?’ and there is technology application that can help provide guidance there to managers, so that managers make decisions that are fair and equitable, and not based either overtly, or accidentally on whether you are a man or a woman, a minority or not, etc. Those things are things we can do to not only help the company to run better, but get better results, you mentioned women founders tend to have better return for investors, and that means something. Companies that have diversity in their C-Suite and their board of directors, and they have better financial results, that’s a tangible benefit to every shareholder, what company wouldn’t want that? 

Absolutely. I think it’s terrific that you’re highlighting this, and it’s a conversation that we need to have more actively, and not just rely on California for instance to pass laws insisting on representation. It needs to be representation of women on public company boards, and we need this to be much more of an organic conversation overall. 

I certainly agree with that, and I think the fact that California did that and led the way in it is great, but the reality is, if you’re a shareholder of a company, as I talked about earlier, your main concern is the return on the investment you’ve made in that company. If statistics prove that companies are more diverse, deliver better financial results, as a shareholder you should be advocating for that, it shouldn’t have to be down to the state to have to pass a law. I certainly applaud California for what they’ve done, but as shareholders of companies, and public companies, we should all be advocates for them delivering the best financial results and doing it in a socially responsible way. 

Absolutely. Well again, I really appreciate you spending some time on this because its topical, and really critical. I think we’re just about running up to the end of our time here, and I always like to ask a question about any recommendations that you might have to share with our listeners, for a book or other resource? 

I have two sitting here on my desk. One is written by Nick Poulson and James Scott, it’s called ‘AIQ, How People and Machines are Smarter Together’. Doctor Scott is a Professor here in Austin where I live, at the University of Texas, at Austin; a brilliant guy, a PhD in Statistics from Duke, and also studied at Cambridge. Had the opportunity to listen to him speak the other day, and I have a copy of his book and I’ve been devouring it. Really interesting read, I would definitely-definitely recommend that one. 

I’d be remiss not to put a small plug in for my friend Tom Davenport. If you are a fan of analytics or artificial intelligence, you’ll know Tom, he’s written a bunch of books, they’re all really good. But in this one I think Tom touches on some things that are important to me, and which we’ve talked about today, which is the business value of artificial intelligence is solid, not sexy or splashy, at least that’s what the inside of the book-jacket says. I’ve read about half of this book and I think he makes the point very well with some depth, that it’s not about the sexiness or the splashy appeal of technology, it’s really about the business results. Tom’s a fantastic writer, a great guy, so, I’d recommend that one, it’s called ‘The AI Advantage’. 

Awesome. I think a lot of people would point to competing on analytics as a pivotal work that really made the case for the advent of Big Data and analytics, about a little over a decade ago. So, that’s a great recommendation. 

That’s been as always, a fascinating conversation and we’ve touched on a lot of super-interesting topics. This is Ed Maguire, the Insights Partner at Momenta Partners, and thank you Mike Flannagan once again for sharing your time, it was extremely fascinating and a lot of fun. Really enjoyed it, thank you. 

My pleasure, thanks for having me. 

[End]