The Three Roles of the Chief Data Officer: ADP’s Jack Berkowitz

As chief knowledge officer of payroll and advantages administration firm ADP, Jack Berkowitz has three main duties. One is to supervise the group’s knowledge general, making certain that capabilities like knowledge governance, safety, and analytics, are working nicely. One other is to construct ADP’s knowledge merchandise, resembling folks analytics and benchmark instruments. However the duty that’s of most curiosity to Me, Myself, and AI hosts Sam Ransbotham and Shervin Khodabandeh is Jack’s oversight of the group’s use of synthetic intelligence.

On this episode of the podcast, Jack describes how specializing in the outcomes the group desires to attain results in higher processes and outcomes. He additionally dives into the subject of AI ethics and descriptions how different organizations would possibly think about assembling an AI ethics board.

Learn extra about our present and observe together with the sequence at https://sloanreview.mit.edu/aipodcast.

Subscribe to Me, Myself, and AI on Apple Podcasts, Spotify, or Google Podcasts.

Give your feedback on this two-question survey.

Transcript

Sam Ransbotham: When outcomes don’t encourage synthetic intelligence efforts, how can they achieve success? Learn the way one chief knowledge officer thinks about AI on at present’s episode.

Jack Berkowitz: I’m Jack Berkowitz from ADP, and also you’re listening to Me, Myself, and AI.

Sam Ransbotham: Welcome to Me, Myself, and AI, a podcast on synthetic intelligence in enterprise. Every episode, we introduce you to somebody innovating with AI. I’m Sam Ransbotham, professor of analytics at Boston Faculty. I’m additionally the AI and enterprise technique visitor editor at MIT Sloan Administration Assessment.

Shervin Khodabandeh: And I’m Shervin Khodabandeh, senior companion with BCG, and I colead BCG’s AI apply in North America. Collectively, MIT SMR and BCG have been researching and publishing on AI for six years, interviewing a whole bunch of practitioners and surveying 1000’s of corporations on what it takes to construct and to deploy and scale AI capabilities and actually rework the way in which organizations function.

Sam Ransbotham: At present, Shervin and I are excited to have Jack Berkowitz, chief knowledge officer at ADP. Jack, thanks for becoming a member of us. Welcome.

Jack Berkowitz: Thanks. Glad to be right here.

Sam Ransbotham: Let’s get began. You’re the chief knowledge officer at ADP. Are you able to inform us about what that position means?

Jack Berkowitz: ADP, often known as Computerized Information Processing, is the world’s largest supplier of HR companies, payroll, taxes, issues like that. We function in 140 nations. We now have over 900,000 purchasers. Hundreds of thousands of individuals are getting paid from us day-after-day.

I form of have a two-sided job. On the one hand, I’m liable for all the knowledge that flows via our techniques. We’re a very huge firm. We now have large quantities of information, so [it involves] all of the issues which can be classically related to chief knowledge officers — issues about knowledge governance, knowledge safety, utilization of analytics.

The opposite aspect of my job — and it’s most likely even a much bigger job — is I construct knowledge merchandise, and so my group builds folks analytics, benchmarks, compensation data, all [those] sorts of merchandise that our purchasers are utilizing to take selections concerning the world of labor day-after-day.

Sam Ransbotham: I didn’t hear the phrases synthetic intelligence in there anyplace. How is that concerned?

Jack Berkowitz: I additionally run that for the corporate as nicely, however we use machine studying all through these processes — whether or not we’re cleansing the data, whether or not we’re constructing embedded capabilities in our HR purposes or our payroll purposes, whether or not we’re doing issues like aligning job titles.

Folks would say, “Properly, how exhausting can that be?” , in any given month, we pay about 21 million folks. We now have about 14 million job titles, and we crunch that all the way down to between 6,000 and eight,000 job titles — so [there’s] an terrible lot of very refined pure language processing and machine studying to make that occur.

Shervin Khodabandeh: It looks as if there’s three totally different roles that you just talked about that each one come collectively. And I say this as a result of at many corporations, there are actually three totally different roles for what you talked about — for knowledge governance, for knowledge merchandise, and for AI — which creates perhaps a little bit of siloed-ness and a little bit of perhaps disconnectedness, as a result of all this stuff should work collectively. Remark a bit, please, on the way it took place that it’s one individual main all three. That’s my first query.

After which, my second query is, is the AI involvement solely within the knowledge merchandise, or is it a broader position that you’ve that you just’re additionally supporting AI for the broader enterprise?

Jack Berkowitz: It’s a very good query. The factor to learn about ADP is, sure, we’re a companies firm, within the sense that we offer, for instance, payroll for about 1 in 6, or much more than that, folks within the U.S. However we are also a SaaS [software as a service] product firm, and due to that, there’s a complete bunch of various growth organizations engaged on constructing SaaS merchandise, whether or not it’s for the small companies all the way in which as much as the largest corporations on the planet utilizing our purposes to do HR or recruiting or payroll or taxes, issues like that. And due to that, this position emerged, actually. It began as constructing knowledge merchandise, however to construct knowledge merchandise and issues like reporting, it grew the information platforms. And off the information platforms, it grew increasingly capabilities when it comes to doing machine studying, greatest practices.

We received into the moral use of information and the moral use of machine studying and AI, and that allowed us to be additive when it comes to capabilities. The opposite factor about it, then, is, OK, nicely, the place’s the extent? As a result of we’ve got all of these SaaS purposes, my groups will typically construct the embedded capabilities for different purposes. However we additionally allow these different growth organizations to make use of the frameworks that we construct.

We, for instance, construct a complete bunch of machine studying operations capabilities — issues about bias monitoring and knowledge form monitoring — as a result of that is smart to be completed as soon as in an organization after which permit different folks to make the most of it. We’ve seen an enormous development in folks figuring out themselves as knowledge scientists over the previous 4 years. We’ve been hiring folks and every part else, however they don’t all should learn to do mannequin deployment into manufacturing.

Shervin Khodabandeh: Very fascinating. Is it honest to say that introduction and perhaps scaling of AI extra broadly exterior of the information merchandise that you just do to your clients was form of the information merchandise themselves? The incubation of those knowledge merchandise opened the attention of the group.

Jack Berkowitz: Yeah, precisely. It was precisely that. It was this incubation. After which, off the incubation, we began to see areas of alternative and areas of pleasure. It wasn’t actually a top-down push. It was very a lot a bottom-up, the place groups have been seeing what we have been reaching, after which different groups would come to us and say, “Hey, wait a second. We wish to construct a functionality. Can you’re employed with us?” And so it’s actually turn into an natural development.

Shervin Khodabandeh: I actually love this story. Usually, I get requested to speak to teams or [do] interviews with media round chief knowledge officer roles, and there’s a query round, “What’s the appropriate chief knowledge officer position?” And I’ve all the time been saying that position needs to be actually, actually linked to the use of information, not simply to the governance of information and to constructing issues with knowledge. And I believe you’re an excellent instance of the appropriate setup of that position and success with that position.

Jack Berkowitz: It’s fascinating, as a result of my profession’s all been about product growth or outcomes. It’s been about ensuring that you’ve enterprise outcomes. Are you constructing one thing, are folks shopping for it, or are you constructing one thing and also you’re popping out with a greater functionality? We carry that product mindset to even our knowledge governance. Sure, we have to do governance, nevertheless it’s not for regulatory compliance, essentially. It’s actually about ensuring that we perceive the data such that anyone can construct an excellent knowledge product on high or an excellent machine studying functionality on high. In any other case, why are we doing all this?

I graduated proper on the time of a recession. Sound acquainted? I spent about 10 years in engineering consulting, principally for DARPA, which is the Protection Superior Analysis Tasks Company, which will get you concerned in fascinating issues. From there, I made a decision to start out an organization with a couple of buddies and received into the startup world, after which I had an excellent alternative to hitch Oracle, perhaps 11 years in the past now, and actually loved my time there. After which I used to be capable of carry it to ADP 4 years in the past. ADP has actually been the head of my profession. I couldn’t have requested for a greater state of affairs when it comes to combining all these learnings of the way you watch a person doing one thing, the way you begin a enterprise in these little startup corporations that have been VC-backed, to the information and the expertise.

We now have to run these techniques 24-7. Firms rely on these techniques to pay their workers, which is, one would argue, some of the necessary issues that exists in an organization, significantly at present, in at present’s surroundings.

Shervin Khodabandeh: Are you able to share with us some makes use of of AI within the merchandise that you just construct to your clients, in addition to perhaps these which can be broadly used for the enterprise or for perhaps core processes or extra internally targeted?

Jack Berkowitz: We’re within the HR house, so [that] runs a variety of capabilities. One of many issues that we’re doing proper now [that] we’re actually enthusiastic about is, we’ve used that job title data together with plenty of different pure language processing to provide you with a expertise graph — a 100% data-driven expertise graph. A whole lot of different distributors do it with hand-cranked ontologies. Downstream from that, that expertise graph reveals up in a wide range of locations, whether or not it’s in recruiting purposes, the worker profile inside an organization so that folks can discover new roles, issues like that. We now have a gaggle that’s doing suggestions for folks in retirement packages. We now have a giant retirement program for folks in small companies. A whole lot of occasions [with] small enterprise, folks aren’t supplied well being care or retirement [benefits]. ADP makes these companies obtainable to small-business folks to supply to their workers. And there’s a functionality that recommends to folks to say, “Hey, folks like you’ll really make investments kind of of their retirement program.” And in order that’s a machine learning-based functionality.

However then we additionally, identical to every other firm, use machine studying throughout the board in different areas. We do plenty of issues in our gross sales and advertising channels. However, extra importantly, we do plenty of issues in our service [channels]. So we’re doing an terrible lot proper now to create a self-service surroundings for our purchasers. Our skill to create a greater service surroundings for them creates a greater expertise, proper? They get extra correct pay, or they get a greater expertise for his or her workers. In flip, that’s higher enterprise for us, so we’re spending as a lot cash and as a lot time on making that service expertise nice as we’re [on] making the core product nice.

Sam Ransbotham: I do know you’ve advocated for the concept of an AI ethics board. I’ll take the counter: Why is that necessary? What’s the good thing about that? Why hassle organising AI ethics boards?

Jack Berkowitz: We began it as a result of we simply felt that the tempo of expertise and the tempo of information most likely weren’t representing the values that we wished to characterize with our purchasers. We began it initially as a result of we thought it was the appropriate factor to do.

The place it’s gone from there has actually been fascinating. We’re studying a heck of quite a bit, each when it comes to our personal product growth but additionally externally about tips on how to educate not simply our purchasers however even our ADP associates [and] when it comes to how we consider the place we wish to do enterprise, like biometrics or voice recognition, and even what knowledge entry rights imply. There’s additionally now, three years later, a giant regulatory push, each within the EU, the FTC of america, the EEOC [U.S. Equal Employment Opportunity Commission], and so we’re not reactive to that. We all know what to consider. We’re in an excellent place to cope with it by considering a little bit bit forward.

Shervin Khodabandeh: From a setup and accountability perspective, do you see the subject of ethics and accountable AI ruled by a board or ruled by an individual suggested by a board?

Jack Berkowitz: We’re way more within the latter, and we wish to do this for a motive. We carry exterior specialists onto our board. We now have folks from the HR area. We’ll have folks from the machine studying world. We’ll carry [in] ethicists.

We would like that board to have freedom of thought. We now have very structured product launch processes, whether or not it’s for the merchandise that we launch to purchasers or whether or not it’s the merchandise that we use internally. And we’ve got governance all through there, and safety; when you can think about, knowledge safety is on the high of our checklist in the meanwhile. Additionally on the board is our chief privateness officer. So we would like the board to have freedom of thought. And it’s an adviser. Undertaking groups should current to the board as a part of going to marketplace for areas.

Sam Ransbotham: It’s fascinating, Shervin. , Jack, I don’t wish to low cost how necessary and delicate your knowledge is, as a result of clearly it’s very delicate knowledge. But it surely’s fascinating. Shervin and I speak with folks; we’ll speak with folks in medical and well being, and everybody appears to have this second of, like, “Oh gosh; oh, our knowledge is admittedly delicate and necessary.” And I believe perhaps that’s ubiquitous now, that each one knowledge appears to be like that. I assume lots of people can be taught from the board arrange like this.

Jack Berkowitz: It’s not a nasty factor to be defending folks’s curiosity of their data.

Sam Ransbotham: It’s actually fairly fascinating, as a result of there are many sources of details about wage, and folks self-report in a number of areas, however you’ve received floor truths on plenty of details about what really hits their financial institution accounts. It provides nice perception into actually what’s occurring within the economic system.

Jack Berkowitz: It creates a singular functionality, each to have the ability to present that data to our purchasers or to their workers or associates, but additionally to deal with it correctly, proper? We now have an excellent alternative to deal with it correctly. And so all the degrees of information safety, all the degrees of all the good issues CDOs care about — you realize, knowledge governance, windfall, lineage — we’ve got a beautiful alternative to apply the sector.

Sam Ransbotham: One space that I believe you most likely are considering mentioning is this concept of evaluating DEI [diversity, equity, and inclusion] metrics. That’s an excellent place that you just’ve been capable of present benchmarking and provides perception into what’s actually occurring versus what folks would love you to assume is occurring.

Jack Berkowitz: Yeah. It’s an excellent level. The corporate really began that in 2017, [when it] printed its first pay fairness explorer, which allowed corporations to try how they have been doing when it comes to pay fairness gaps for deprived teams. Now we’ve got the benchmarking functionality that permits an organization to see, for his or her location, for his or her business, for his or her firm dimension, how are they doing when it comes to creating a various surroundings, after which additionally, how are they doing not simply bringing folks in, however really advancing them of their careers?

By bringing all that collectively, through the use of our benchmarking functionality, by fixing an issue, by taking a look at an final result, we’ve had nice success.

We are able to run multiregression evaluation all the way down to the person — not simply inside their firm however in opposition to the various inhabitants in that native geography or business. However then we are able to say, “OK, listed below are 4 or 5 finances situations.” As a result of it’s one factor to say, “Hey, you might have pay fairness points.” However, you realize, perhaps the corporate has budgetary issues, to allow them to make some decisions about budgetary situations, and it tells them, “OK, if you wish to shut this finances, these are the folks that you just’re capable of cowl.” And they also mainly can change that. They arrive out after which, increase, they will make these modifications straight into folks’s paychecks. And it’s a significant affect.

Shervin Khodabandeh: That is fantastic. The query I’ve, which might be very fascinating for our viewers, is, the place do you get began? As a result of we see … I imply, I see this in my work, [I] see this with Sam after we interview and analysis the subject of AI deployment, there’s a query round, how a lot do you construct capabilities earlier than you start to monetize or commercialize or construct knowledge merchandise or use instances, versus how a lot value- and use case-driven you’re. And I’m actually considering your perspective, each for ADP and any recommendation you might have for others within the early levels of their journey.

Jack Berkowitz: The best way I’ve all the time checked out it’s, when you’re constructing a product, whether or not you’re a startup firm or every other, is, construct the thread from one nook of the piece of paper as much as the opposite nook of the piece of paper. And use that thread — in different phrases, a use case or two — that will help you outline what you want in your state of affairs together with your firm at the moment.

You could possibly say these are prototypes, however in my thoughts, a prototype is ineffective until you really attempt to have an effect with it, since you don’t study tips on how to measure final result. You gained’t learn to measure what you really want.

The factor that’s been misplaced in machine studying and all the thrill over the previous six, seven years is that each one machine studying and all AI is focused to an final result. To me, it’s actually about fielding some functionality. Off of fielding that functionality, you’ll be taught what ranges of machine studying operations you want, you’ll be taught what ranges of information you want. And I do know that there’s 1,400 distributors. Matt Turck has the good FirstMark Capital matrix of it, and I bear in mind when that was solely 30 distributors, by the way in which.

I do know all 1,400 distributors will let you know [that] it’s worthwhile to purchase their stuff straight away, and that’s simply not true. That’s simply not true. You’ve received to purchase a few of it, although, with a view to get that preliminary factor fielded.

Shervin Khodabandeh: Yeah, I’m so blissful you say that, as a result of that’s, actually, been the unlock we’ve seen each in our analysis, Sam, the place we see that corporations that simply take expertise first perhaps get a little bit little bit of worth however there’s a giant piece they don’t get. But additionally, in our work at BCG, that’s been the main unlock, to be value-driven and outcome-focused. And I like the way you speak concerning the thread, since you can not simply construct this stuff in silos, nevertheless it doesn’t should imply that you just construct … the total stadium earlier than the baseball [game] begins. You could possibly begin taking part in.

Jack Berkowitz: Precisely. At an organization earlier than ADP, we used to name it a “Subject of Goals marketing strategy.” It’s like, “No person ever invented baseball; why are we constructing a baseball discipline now?”

The entire concept is, get that thread working. And, you realize, perhaps it’s not all the way in which related. Possibly you continue to have anyone standing up with a floppy disk and working to the opposite pc to make all of it work, however not less than you might have an concept. After which you may broaden out that thread over time. That’s all.

Sam Ransbotham: For our listeners, floppy disks have been issues that folks needed to put into computer systems to retailer data.

Jack Berkowitz: Yeah, thanks, Sam. Thanks, Sam. You may see it — you’ll really see it on icons on previous Macs and previous PCs, so, yeah.

Shervin Khodabandeh: Jack, thanks a lot. This has been so fantastic and insightful. This brings us to the following part of our present, the place we ask you 5 questions.

Jack Berkowitz: Nice.

Shervin Khodabandeh: And we count on some fast reactions to those questions. So I’ll begin. What’s your proudest AI second?

Jack Berkowitz: My proudest AI second is when my algorithm went into manufacturing.

Shervin Khodabandeh: Excellent.

Sam Ransbotham: That ties in nicely.

Shervin Khodabandeh: What worries you about AI?

Jack Berkowitz: I believe the following reoccurrence of the AI winter. Having been via it the primary time and seeing that come quick. Let’s not overpromise.

Shervin Khodabandeh: Mm-hmm. Your favourite exercise that includes no expertise.

Jack Berkowitz: Kayaking on the Chattahoochee River.

Sam Ransbotham: Ah, been there, completed that! I’ve kayaked Chattahoochee many a time. I’m really from Atlanta, from Smyrna.

Shervin Khodabandeh: The primary profession you wished: What did you wish to be if you grew up?

Jack Berkowitz: I wished to be an astronaut, like each different child born within the ’60s.

Shervin Khodabandeh: Your biggest want for AI sooner or later?

Jack Berkowitz: My biggest want is that we assist folks stay higher lives.

Shervin Khodabandeh: Thanks. Very insightful.

Sam Ransbotham: Jack, nice assembly you. I believe there’s quite a bit right here that folks can be taught, significantly among the particulars about the way you’re organized. I believe that’s one thing lots of people can be taught from. We actually respect your taking the time.

Jack Berkowitz: Thanks, Sam. I actually respect the dialog.

Sam Ransbotham: Thanks for becoming a member of us. Subsequent time, Shervin and I speak with Ameen Kazerouni, chief knowledge and analytics officer at Orangetheory Health.

Allison Ryder: Thanks for listening to Me, Myself, and AI. We consider, such as you, that the dialog about AI implementation doesn’t begin and cease with this podcast. That’s why we’ve created a gaggle on LinkedIn particularly for listeners such as you. It’s known as AI for Leaders, and when you be a part of us, you may chat with present creators and hosts, ask your personal questions, share your insights, and acquire entry to useful assets about AI implementation from MIT SMR and BCG. You may entry it by visiting mitsmr.com/AIforLeaders. We’ll put that hyperlink within the present notes, and we hope to see you there.