CDOIQ Symposium 2024 | John Spens | Thoughtworks

theCUBE.net

Clips
News
More from CDOIQ Symposium 2024

John Spens

Head of Data and AI, North America

Thoughtworks

Revolutionizing decision-making: Embracing data as a product with design thinking and AI

As data continues to be at the epicenter of better decision-making in organizations, viewing data as a product should be part of the compass for enhanced value acceleration.For data to materialize as a product, design thinking is essential because information should be shared in a standardized way using a mesh framework, according to John Spens (pictured), managing director of North America data and AI at Thoughtworks Inc.Thoughtworks’ John Spens takes theCUBE through the data as a product lens.“The principle most quickly embraced by our clients, by individuals I talked to, by myself, was that idea of thinking of data as a product,” Spens said

play_circle_outline Exploring Federated Computational Governance and Data Mesh Principles for Modern Organizations

play_circle_outline Data as a product concept

play_circle_outline Open sourcing data catalog and standards

play_circle_outline Primary barriers are organizational, not technological

Info
Transcript

John Spens | Thoughtworks

John Spens

Head of Data and AI, North America Thoughtworks

The Chief Data Officer Symposium has moved to the Hyatt in Cambridge, Massachusetts after previously being held at MIT. John Spens, the head of data and AI at ThoughtWorks, discusses the industry's current state, including the evolution of the data mesh movement and organizations treating data as a product. Clients are embracing the idea of data product owners combining product and design thinking with technical skills for successful solutions. Governance in data management and challenges in implementing federated computational governance are also important t... Read more

explore Keep Exploring

What is the current status and evolution of the data mesh movement? add

What is one concept from data mesh that has resonated most strongly with clients, and how are they internalizing and applying the idea of data as a product in their organizations? add

What impact does the open sourcing of self lakes data catalog and the industry's adoption of the Iceberg format have on clients' thinking and decision making regarding technology? add

What are the main challenges companies are facing with regards to implementing new technology and data analytics? add

bolt Powered by CUBE AI

John Spens | Thoughtworks

search

>> Hi everybody. Welcome back to Cambridge, Massachusetts. We're here at the Hyatt at the CDOIQ Conference, the 18th annual Chief Data Officer Symposium, used to be at MIT at the Tang Center. This is theCUBE's eighth year, myself, Paul Gillin, Sanjeev Mohan. Paul, eight years doing the CDOIQ.

Paul Gillin

>> I feel old, Dave.

>> We've seen... You look good. We've seen the ascendancy of the chief data officer from this sort of back office, sort wonky data quality position to one that now is front and center with the whole GenAI movement. John Spens is here. He is the head of data and AI for North America at ThoughtWorks, leading thought leader in the industry. John, thanks very much for coming on theCUBE.

>> Thank you for having me.

>> So what's going on at CDOIQ? Are you speaking here, you're just interacting with customers? What's your role here?

>> Well, I am or I have been leading a panel, and we have a couple of our clients also talking as well, but I think my role here, my passion here is getting to interact with all the great practitioners. CDOIQ has so many talented people. It's a great learning opportunity. So first and foremost, on a personal level, I enjoy the opportunity from learning from all these people.

>> Yeah, so you guys... Obviously, we were talking off camera. You were heavily involved in decentralized architectures, participating in the whole data mesh movement that Zhamak Dehghani coined and documented so eloquently. Where are we at now with the data mesh movement. We saw data mesh, and of course, Gartner had to come out with data fabric because they got to invent every term, note to man-

>> Little bit....

>> and women. That caused some confusion, but it seems like things have settled down a little bit. But where are we with data mesh?

>> No, I think that's a great way to describe it. Things are settling down. We're figuring out what works and what doesn't. Zhamak took her experiences, and mapped out a set of principles that were highly effective and really changed how people thought about delivering data solutions. And over the past five years, since we're just celebrated the anniversary of that article, ThoughtWorks has been working with our clients to put those practices in place and learn what changes are necessary to truly make data mesh successful.

Paul Gillin

>> One of the concepts that seems to have resonated most strongly out of data mesh is this idea of data as a product. How are your clients internalizing that idea? How are they applying it in their organizations?

>> Oh, you're absolutely right. The principle most quickly embraced by our clients, by individuals I talked to, by myself, was that idea of thinking of data as a product, how you're delivering it to the people who are consuming it, making sure they're successful, crossing that last mile with those consumers. And our clients, as I said, embrace that concept. The interesting experience for them and for most practitioners is bringing people in who are now data product managers, so bringing together those skills of product thinking, design thinking, with those hard skills of understanding how to use these solutions, the technologies that are involved, the roles of data governance. So we are creating another unicorn in our industry, this individual called the data product owner. So helping those organizations we work with develop those talents and build that cadre of product owners is one of the biggest challenges they face.

Paul Gillin

>> Well, you mentioned design thinking. How does design thinking fit into data product management?

>> One of our experiences in the technology industry has been we have to stop thinking about what our clients might be asking us for directly and really exploring what they need, so taking a step back and thinking about how that data will be consumed, the old Ford adage, "If I gave people what they wanted, it would've been faster horses." There are times where... And especially now with the introduction of AI, we have the opportunity to circumvent old processes, and produce more useful solutions if we understand what the need is, what our customers, our data consumers are trying to produce.

Paul Gillin

>> So thinking about those principles, the data ownership, which I always found was, okay, that's like a starting point. You got to get everybody together and agree on who owns what data as product. I think generally, people are comfortable with data as product. Self-service, I think with cloud, everybody kind of has self-service on the mind, but the hardest of the four principles was that sort of federated governance. I think you called it federated computational governance, which I heard was the high degree of automation, so we didn't have to have hall monitors worrying about this. So is that what you find with clients, that sort of the starting point is one, and if you can get that right, things start to flow, but then you run into number four, which is the big gap in the business today?

>> Yes, that is absolutely the path that most individuals go down in sort of understanding data mesh. And the experience too is when you start to think of data as a product in delivering these products... Well, when Zhamak put forward the principles of DATSIS or the fair principles for data that we're familiar with, describing how data has to be findable and needs to be interoperable, all of those characteristics, you start to understand how governance plays such a critical role in ensuring that data is readily consumable. So I think that most organizations, most practitioners that we work with follow that journey of understanding, all right, if this data is going to be consumable by these different organizations, now I'm starting to get the idea. And the federated model is simply saying, as we move to scale, we understand that we need to engage the whole organization, like that saying quality is job one, quality is everybody's job, data quality is everyone's job. So now how do we distribute? How do we federate that responsibility? Because like a federal system, you still need that strong centralized policy creation group, the compliance and risk individuals who are driving this is why it's really important to keep this data secure, protected and at high quality, but then enabling the organization to act on that, and of course, using automation as a way to ensure that happens.

Paul Gillin

>> We've seen some progress this year with the open sourcing of self lakes data catalog with the industry really being to come together around the Iceberg format and the standard data platform for lakes and lake houses. Does this change the equation somehow? Are your clients thinking differently because they have access to busy standards that are emerging?

>> So being from ThoughtWorks where we have long been espousers passionate adopters of open source software, we're celebrating, in a way, this move that was seen. And I've been in the data industry for over 30 years, and so that move from very singular large proprietary solutions to an open source format, more of that sharing we're seeing, yes, our clients are able to make more granular decisions about the technologies they leverage, and they're getting the opportunity to evolve their technology strategy more rapidly because they have the opportunity to swap those in. There's the greater interoperability.

Paul Gillin

>> Is this a technology first phenomenon? Do they have to get their governance house in order before they think about what technologies they use?

>> I've got to say no because nobody's ever done that. Nobody's gotten their governance house in order first. It's always been that cleanup afterwards, and they realize that's the real challenge. I think it has been one of the struggles in our industry, is that organizations do think of it as a technology first problem. And data, in my experience, my opinion is that greatest representation of the intersection of technology in the business as how we're describing how the business operates, how the business is operating right now with bits and bytes. And so understanding all of those definitions, how things work, what it means when we say this is our performance or this is the cost of building that tractor, we need the definitions from the business. So if we don't start with an organizational business first mindset, we're going to just build new technology.

>> And that's frankly the brilliance of ThoughtWorks business model. You've always... In our industry, you remember NetWear consultants?

>> Oh my god, yes.

>> The Cisco CCIE, or you could have a consultancy to do that type of stuff. And then you get sort of locked into that domain, and then when that domain gets disrupted, you get disrupted. But so ThoughtWorks, it seems to me anyway, publicly has always tried to avoid picking winners of technologies with the understanding that the customer's going to have multitude of tech. And I think that's the case here with these catalogs. You mentioned Snowflake with Polaris. They're kind of keeping horizon to themselves. You've got Unity catalog, which Matei on stage made open source. You've got Microsoft doing Purview, which is going to be a catalog of catalogs. You've got AWS with Glue and data zones, and they got different meta. So customers have it all.

>> They do.

>> So you're never going to be able to solve it with all the wood behind one arrow, to use a Scott McNeely term. So having said that, how do you see this... Sort of picking up on Paul's question if I can't get my governance act in order beforehand, your response to that was, "Yeah, you kind of figure it out after." But it can't, so-called, can't be a bolt on either.

>> Absolutely.

>> So help square that circle for me, John. How do I not bolt-on governance as an afterthought, and yet put in a governance structure? I think the answer is going to be from a... It's going to be a business answer, but I'd like to hear it. How do I do that given all these choices that I have? And I know my business is going to... some the data scientists are going to choose Unity, my data warehouse guys are going to choose Horizon, and my AWS guys, my cloud architects are going to choose AWS. How do you make sure that the governance can actually be federated and automated?

>> Yeah, it's a fantastic question. And our approach with our clients in trying to tear this apart, is to start small, is to work in a singular or set of simple use cases with the eye to the future. We're not just going to build one thing because then we'll just create yet another silo of a different technology and a different challenge, but saying we are going to implement not just the technology, but we're going to put the entire process in place for a selected set of use cases and subsequently a selected set of data. So we're going to do the full steel thread through the entire process rather than trying to change the whole organization, and build momentum for it. So you're giving people what they want. They want, one, they want to see something come out of their investments, they want some sort of tangible return, even if it's minuscule. They want to see a path forward to how their organizations are going to change. So we use a model. We call it show, shift, and scale. Demonstrate the value, shift from a proof of value to a way of working, and then scale it across the enterprise. So it's a journey. It's not a silver bullet. It takes time. And the panelists that we were talking to today are all of our clients who have been on this journey, some where it's between six months and three years. And as one of them said, this is not an endpoint. This is just a step in our evolution. And so taking that evolutionary approach is near and dear to our hearts and has been the greatest way to make our clients successful.

Paul Gillin

>> I know was one of your panelists was from Pfizer, and 90,000 people and everybody very opinionated about how they do things there. How does an organization like that that's has very siloed, and really siloed by design, create any kind of unified governance?

>> Right. And to be clear too, that panelist, as he said in his session is part of an acquisition. And so he's just gone from CGEN, a small oncology focused firm, to Pfizer. And so he's grappling with those very challenges himself. You do need to... The experience that we've had that has never been successful has been to try and say, "Listen, we're all going to sing from the same hymnal. We're all going to jump in this together, and we're going to change the whole organization. We have to build the momentum." So you're right, a large organization like Pfizer working with... We work in federal space, very large organizations there. Driving that change has got to have first support those champions, but drive it incrementally across the organization. And it's more of a viral approach to bring the change.

>> Paul, one of the first case studies I ever did was, I'm looking at it now, July 2021, how JP Morgan is... I'm probably thinking data mesh.

>> Data mesh.

>> And so that was a really good example, John, if somebody starting... Don't know if you guys were involved, but they were starting small, right? But I remember saying, okay, do they have a roadmap and a journey? And we talked to them, and they recognize the importance of that, but they're like, "Look, we just wanted to dive in and learn." But there was glue and there was catalogs, and there was the Hadoop legacy stuff. I did another one with HelloFresh, as you might recall, and they had a really completely different approach, although by the way, Hadoop was still part of that. So it'd be interesting to go back, and you'd be really good at this, at seeing how these guys have evolved, whether it was a cul-de-sac in a kind of one-way door, as some people say, or whether they actually could evolve it in that way. Are there examples? Now that the data mesh has been out for... You just say five years is anniversary.

>> Yeah.

>> Are there examples of folks that started small and have been able to successfully, your client's been able to successfully extend that start small and are on that path?

>> Yes. Yes, we do have some clients that's... In talking about bringing out names in front of cameras, I'm-

>> You can't do it, but you mentioned this applies because they were public in the panel.

>> Right.

>> They were in a panel.

>> And in-

>> Yeah, I get it, but even generically.

Paul Gillin

>> Sure. And the interesting thing is in the life science of space, we have a number of clients who we've helped on this journey and have gone on for multiple years on the journey and expand across their organizations. And so we have seen that success, and especially working in the life sciences space, moving either starting in the clinical trial space or in the discovery space, but then moving through into the commercial process, moving through the entire value chain as they start to see that value have the impact. So we have seen multiple organizations, a number of organizations, I should say, that have gone on that journey, particularly life sciences. We've had some in manufacturing. So it is there. The thing that differentiates it... It was interesting, a conversation I had down at one of the tables. People were saying, "Well, people aren't patient, and they can't wait for those results." And I said, "That's right. You have to find those pioneers, those people who see the big investment, the long tail benefit and are willing to show that patients because they know there will be some return." Now eventually as technology evolves, we'll solve the pioneers' problems, and we can bring that back to the people who move quicker.

>> And in those examples, what has been the business impact? It sounds like it still may be somewhat fuzzy. And what are the primary barriers to achieving the vision? Are they technological, or are they organization, maybe a little bit of both? But I wonder if you could

>> Got it. So the benefits and in life sciences space and interesting learning for me was drugs are going to come to market. They are expensive and take a long time. Our biggest opportunity is actually finding out the ones that fail as quick as possible. Move the chips to the best bet, as it was explained to me. So we have helped organizations accelerate the process of killing off those non viable treatments. And so that's been a successful return for us and for our clients. So we've seen success in that. We've seen success in a number of industries where complex, one particular, complex manufacturing chains where the cost of scrap the loss of productivity has been dramatically reduced in multiple digits. So those are the types of benefits because we enable what data has always tried to do, cross organizational boundaries, cross domains. Now, the challenges, overwhelmingly they are organizational. The technology challenges with the rapid pace of innovation right now, if there's a true technical challenge, well, there's somebody at Google, at AWS, at Databricks, wherever solving that problem right now. But the organizational challenge, the fact that these source systems are owned by people who are focused on one responsibility and these data consumers have a very different responsibility, that's where the friction lies. And I've experienced that since the 90s.

>> I love that business benefit example. I always bring things back to horse racing. Kentucky Derby, there's 22 horses. You got to throw somebody out. And you're helping pick better, bet on the sure thing.

>> To be clear, we haven't worked with any racing because-

>> No. Great business.

Paul Gillin

>> But the two examples we cited were both cost production, clinical trials, and saving on scrap. Is that the best place to start?

>> I think it's the most tangible. When you talk about... For example, when you talk about accelerating the drug discovery process, that's a very open-ended question. There's a lot of risk in there. But if I talk about reducing cost, reducing pain and friction in the process, that's a great place to start. But the interesting point that comes from that was working with actually one of the record producing companies that, don't call them that anymore, but they had a very arduous process of coming up with quarterly results. And so we accelerated that and we made it a push button event. Now the thing that happened was, yes, we reduced the amount of effort required to get that report out, that analysis out. But what we also did was gave them the ability, the insight to know how our performance is not at the end of the quarter or not two months later, but right now. So being able to take those snapshots and start looking at how the performance is, now, it moves from cost reduction, the effort of the people to produce those reports to we have greater insight. And that translates into things like variable pricing, demand pricing. So that ability, the insights starts often with a cost reduction, but then now you know something and you can do something.

>> Was that a virtual close you're describing in a quarter, within a quarter to three weeks, 10 weeks in, you could do a virtual close? Is that-

>> Yeah, absolutely. Yeah, we were able to get that down to a period of about... It would run through and reconciliation. You always have to do that. So we could complete that in about a course of a week, which had taken the full quarter before.

>> Right. Amazing. All right. Hey, we got to run. John, thank you very much. I really appreciate your time and your insights, and congratulations.

>> Thank you so much. It's been a real pleasure.

>> All right, keep it right there, Paul Gillin and I will be back. Sanjeev Mohan is also in the house. You're watching CDOIQ from Cambridge, Massachusetts on theCUBE.