Reflecting on 10 Years of AI: Abundance, What We Got Wrong, and Innovation at the UN

Curious how AI has actually evolved over the last 10 years? We have the perfect guest.

In this episode of Rendered, Praniti sits down with Rosedel Davies-Adewebi, AI & Digital Transformation Leader at Morgan Stanley and former Head of Innovation at the UN Global Compact, for a wide-ranging conversation about how AI has evolved, what we got wrong, and where it's headed next.

Rosedel had a front-row seat to a lot of the moments that shaped where AI is now: from leading breakthrough innovation programs with Fortune 500 companies at the UN, to driving agentic AI pilots inside one of the world's biggest financial institutions.

In this episode, we get into:

• Whether AI abundance is actually real 

• The 2016 prediction about AI and jobs that turned out to be completely wrong 

• Why blockchain never had its ChatGPT moment 

• The "global compute divide" and why some of the most interesting AI work is happening outside the US 

• What separates an AI pilot that ships from one that dies in the lab 

• Why the next era of UX might not be designed for humans at all

Plus a "Would You Ship?" segment imagining AI compute as a public good.

Transcript

SPEAKER_01 0:00

Welcome to Rundard, where we explore the future of AI and creativity. Today's guest is Rosdell Davies Adawebi, who has spent the last decade building breakthrough innovation programs with Fortune 500 companies at the UN Global Compact, and is now leading agentic AI pilots at Morgan Stanley. We're getting into whether AI abundance is actually real, the 2016 prediction about AI and jobs that turned out to be completely wrong, and why blockchain never had its ChatGPT moment. So let's get into it. Hi Rosdell, welcome to Rendered. Thank you so much for joining. It's so great to be here with you, Pranithi. Over the last 10 years, um, you've had a front row seat to AI becoming real. From from building innovation for good programs with Fortune 500s at the UN to now leading agentic AI pilots inside Morgan Stanley. So congratulations, 10 years is so exciting.

SPEAKER_00 0:55

Thank you so much. I can't believe how fast 10 years has flown by. I'm thinking back to 2016, and that was a really pivotal moment. So there were quite a few foundational models that I saw back then that really set the foundation for what we experience or the capabilities of Gen AI today, right? To step back a little bit and give you a picture of the UN Global Compact. So it's it's the world's largest sustainability initiative. It was launched by um then Secretary General Kofi Anan, who made this call to companies around the world to support the principles of the UN, multilateralism and sustainability, right? Um so the you know, the Global Compact have been on the forefront of really engaging companies in new ways to drive their sustainability efforts and to support UN principles. So in 2016, I remember the day, so just kind of picture it in the office in New York. Um of the senior leaders comes to me and says, Hey, you know, I read this amazing book. It was about abundance. And um we want to start this platform that brings together um UN agencies, our companies, and entrepreneurs within the companies to work on like social impact stuff. And I remember saying to him, maybe hard pass because I had already been working on this social enterprise and impact investing work. And what I had seen was it was kind of window dressing for the company. So I don't, I didn't think I wanted to continue working on that. But he kind of pressed me and I read this book and it gave an overview of the new emerging technologies and what how it could create this world of abundance, right? And I've always been one of these people who truly believed in companies' ability because they're scale to drive positive impact. So I I returned to him and I said, Hey, do you have a concept note for this? And he said, No. And that back in in the UN, the concept note is like the set of slides. It's the document you use to communicate your ideas and influence and and and get things done, right? So I said to him, Don't worry, give me a month, I'll do research, I'll come back with a with a concept note. So I reframed this platform, which ultimately became this breakthrough innovation program, to focus on helping companies understand this new landscape of technology, emerging technology, AI, blockchain, Internet of Things, and help them understand how they can use those technologies to advance their sustainability initiative. So we brought together these 10 companies from around the world, Fortune 500 companies, got some of their most like talented um entrepreneurs, their talented employees, and brought them together in a year-long program to explore the technologies and to design products that address sustainable development goals using these technologies. Back then, I would say there were like four themes that were really prevalent that we were talking to the companies about. One was around big data, right? And how big data really powered the application of AI. And like back then, big data was about, I don't know, like maybe a hundred or one million megabytes, right? One terabyte of data, right?

SPEAKER_01 4:45

Can you imagine? Yeah, which which now is like I need at least three terabytes of data just to store my photos.

SPEAKER_00 4:51

Exactly, right? Like your hard drive is one terabyte of data, right? And now the largest, I think uh large language models have about 1.8 trillion parameters, right? So that's like between one and four terabytes. So like big data seems quaint. You don't like we don't talk about big data. This is a a total way to like carbon date yourself by talking about big data, but that's what we were talking about. So then the second thing that we were talking to companies about, which seemed very abstract at the time, which now is more prevalent and it's really top of mind for so many people, is we talked about the impact on labor, right? So we talked to companies about that there's going to be this duality of labor experience. I remember going to a meeting with the International Labor Organization. So they had this is a uh an agency of the UN, right, that really focuses on labor issues around the world. And the discussion was around AI's impact and disruptive technologies' impact broadly on labor. So we, you know, we talked about these disruptive technologies and how, you know, it's a double-edged sword, you know, the the common narrative that it create it can create jobs and destroy jobs. Yeah, the narrative I think we still hear today, right? We still hear it today, but our thought was that there was low, like the impact would be mostly on low-skilled labor, right? Not what we are starting to see today. Yeah. You know, kind of white-collar professional labor. And um I remember at that meeting we talked about a you know, really interesting topics to mitigate the blow of labor losses from um deployment of these disruptive technologies. You know, everyone talked about universal basic income, but there was one speaker from Brookings who referenced um this book, I think by Robert Kaplan, about this idea of job mortgages. And it was a really unique idea. It never really took off, but the idea then was having companies sponsor employees to get training in skills of the future. And then if and you know, so basically they're like pre-paying for skills, and if the employee was hired, they'd pay back with the cap on like you know the payment amount. But if there was no job, the c the company would absorb the the cost, right? So I think if we look at today, yeah, I think we were wrong on on the prediction of like which groups would be affected most or first, right? Yeah. And we're still grappling with what that impact will be um to mitigate the blow of job losses, because yes, there will be job creation, but sometimes it's not as fast as job losses.

SPEAKER_01 8:19

You mentioned, you know, the technology you were exploring 10 years ago with these companies. It was AI, blockchain, IoT. Within those, when you're you guys were talking about job loss back then and disruption back then, how are you breaking it down by these three different technologies? I know right now most of the discussion is around AI, but was it like that 10 years ago?

SPEAKER_00 8:43

The hype wasn't around AI. AI really separated from the pack over time. So the discussion was around disruptive technologies writ large. One of the things, though, that like I recognized basically 2018, and this is when I um I transitioned to Morgan Stanley in 2019, is that as I'd been working on this breakthrough innovation program, learning more about the technologies broadly, I formed this view that, you know, AI is going to be the most transformative technology out of all of these. And I'd love to understand how this applies within a company because mind you, I'd been working across multiple companies to help them identify use cases in the sustainability field, work on product development, um, you know, identify additional ways they could use these technologies to drive their the performance of their sustainability initiatives. But I really wanted to understand what um what companies could do with AI within their companies outside of sustainability. So that's what really um, you know, precipitated my transition to Morgan Stanley. So I would say in this like time, 2016 through 2018, the discussions were around to start these broad groups of technology. But you know, like 2017, I think that was the time when the paper came out, Attention is all you need, which is what I think started to spur this focus on Gen AI, right? And so this is where I perceived that artificial intelligence started to break away from the pack of this group of disruptive technologies.

SPEAKER_01 10:40

Gotcha. But when you were working with these companies back in 2016 to 2019, were people excited, were people most excited about AI or IoT or blockchain? Um, how was that viewed back then?

SPEAKER_00 10:58

So I think, you know, it depended on the company, right? So for example, we I worked with a a company, this was a huge like uh energy company in Italy, right? And they were working on building mini-grids. So basically for communities that were off-grid in rural areas. And so for them, what was super interesting was IoT, right? So how do we how do we use IoT to monitor and maintain this equipment remotely? Um, and then for them, the big the then we layer on the big data piece, right, which is okay, what can we do with this data? And then the third piece of it was, okay, now that we have this like this base of data, what can we use AI for? And I would say in in that time, you know, the uses of AI too were were pretty simple, right? Like you had around three things that companies were the were using AI for, which was like recognition, image recognition to start, uh classification and prediction, right? Compare that to today when we're like now talking about AI in terms of content generation, reasoning, taking action, right? Very very different. It's almost like two different worlds. Yeah. As AI started to break away from the pack, one of the things I did see was that blockchain kind of started to fizzle because a lot of people would say, well, it's a solution in search of a problem. Sometimes innovation, or a lot of times, some of the best innovations come from like matching a solution to a problem. And at the UN Global Compact, when I was working on breakthrough innovation, our focus was yeah, understand the technologies, but let's get really into the solution, into the problem. So we would work with these like Fortune 500 companies and and their their like young talent. And I remember one session we were in, we were in Bangalore and we were looking at um like mobile payments and how like that's taken off without like, you know, specific investments and infrastructure. Okay. And so we would like literally we were on the streets talking to vendors and moped drivers to see how they use PayTM so that we understand, like, oh, well, what is the challenge they're trying to solve with these mobile payments? How does it help them? So this was really the type of it's the it's the foundation, at least of the way I practice innovation. It's like, let's really understand the problem and let's understand the people who have the problem, their context, what they're trying to solve for. And that way we can develop innovations that tr truly are innative, innovative, and solve real problems.

SPEAKER_01 14:14

I love that, having a very customer-first lens because I feel like that's sometimes one of the biggest failures of a product. And then, you know, have you come across um an idea or a product that was almost ahead of its time? Um it was maybe a great idea, but maybe the customer, maybe the world wasn't ready for it.

SPEAKER_00 14:35

Wow, that's a good one. I think it was, I think it was blockchain, actually. Right. Remember it was, you know, uh, so there was, I remember we we would talk about blockchain, Ethereum. Yeah, but Ethereum never got the love it should because it was so practical. You could make smart contracts on it, right? There were so many practical uses. Yeah. I think, yeah, I I think what we're seeing now is we're seeing like, you know, uh, well, Bitcoin is being traded. But I I think still the true promise of blockchain has not been has not been tapped, right?

SPEAKER_01 15:17

No, it it never got the publicity or the love that AI did. And I think one of the reasons you kind of put it really well, like AI as a customer, I can, or a user of AI, I can go to Chat GPT and interact with it, right? Like I can under I don't need to understand the deep technology to use it, versus blockchain just never really picked up that traction for you know any customer to go ahead and interact with it and use it like that.

SPEAKER_00 15:42

So it's such a good point. And I'd advise a couple startups maybe in 2018 around that they were trying to do just that, like basically put a user interface on top of like the blockchain, but they never really took off. But I'm sure you definitely appreciate this as a product leader. But yeah, I think one of the biggest challenges was that it seems so esoteric. Like, how do I get um how do I get on the blockchain? What do I actually use it for? Totally. The apps, right? Versus why did Gen AI take off so rapidly for so many reasons, but from a like product perspective, like I open up a a a website and I have a, you know, not even a prompt line, but I have like a box I enter information into or questions into, which is building on behavior I've already been doing, which I think is super powerful when it comes to product. Like I didn't have to learn how to do anything new. I could use a new technology with the same behavior I'd been doing, I'd been, you know, doing for years. Yeah. Which I think is really, really important for product adoption. Definitely. Resistance, resistance was low for your average consumer. Yeah, exactly, exactly. Versus blockchain where it was pretty, it was really hard to you you really had to sit down and research or um yeah, to to get to get access to any of these things.

SPEAKER_01 17:18

Yeah, and I will admit, I have watched multiple blockchain videos. I've read about the technology, and it's still don't ask me to describe it.

SPEAKER_00 17:27

You know what I mean? Well, you know what was interesting though, and that's why I I feel that, you know, it's really worth it to study the innovations that happen, like particularly outside of the US, because we talk about constraints. Some of these constraints create such opportunity. So I remember reading about like um Sierra Leone, and I I I have an affinity for Sierra Leone because that's where my mother emigrated from like 55 years ago. Wow. And um they had used this in I think the 2019 uh election, national election. They had used blockchain to basically verify votes. Right? So they were looking at voting IDs on the blockchain and using the blockchain to verify. There was another company that the name escapes me, but they were using blockchain to help refugees establish their identity when they had to basically leave and you know go into refugee camps and await like asylum hearings. So, so basically they had built this app and then family members around the world or other people who knew them could verify pieces of their or give pieces of information to verify their identity. Because oftentimes when people are leaving in situations where they have to become refugees, they are not able to bring all of their documents with them. So this company was helping to identify that, and basically all of those pieces of verified information was stored on a blockchain, which then the app would access, and this person could use these various pieces to basically show or prove their identity. So, you know, these were all examples of innovations built under real constraints. So I think it's really worth looking at constraints and how they can drive creativity sometimes.

SPEAKER_01 19:36

I love that you just said that because my mom is an artist, and one thing she always loves to remind me, you know, when I'm complaining about this or this, I can't do this because I have this constraint. She always tells me that creativity thrives under constraint. So I really like that you said that, and I think those examples speak to that.

SPEAKER_00 19:56

My mother would always tell me necessity is the mother of invention. Yes, that's another good one.

SPEAKER_01 20:06

Yeah, because I think when people think about creativity, right, when they think about AI, it's so broad, right? You can go anywhere with it. And I think a few things you said about focusing on the actual problem, focusing on necessity, focusing on the constraint is what actually then can deliver a really impactful product at scales. Okay, so now we've reflected a little bit on what you've seen in the AI space over the past 10 years and where we're at now. So as you think forward, right, let's say to the next 10 years, what technology and what shifts are you expecting to happen?

SPEAKER_00 20:40

I always laugh when I get this question, like no matter how it's phrased, right? Like, where do you see yourself in 10 years? Or, you know, first of all, just to share why I chuckle, because 10 years, at least from a business perspective, is two business cycles, right? Yeah. And so, and just look at the rapid amount of change we've had in the last five. Yeah. So it's really hard for me to predict accurately, or not even accurately. I mean, nobody can do this, what the next 10 years will look like. But I could say here's what I'm seeing over the next two to three years, right? So we talked about constraints, and at least from the perspective of like kind of corporate AI investments, this is where constraints are gonna come into sharp relief, right? Where companies spent a lot of time in the last like two, three years bringing in AI capabilities, investing in them, doing proofs of concepts. Now there's gonna be a lot of pressure internally and externally for companies to say, yeah, these investments created these capabilities that solve these real problems and create commercial values. Yeah. One of the biggest things that I'm seeing is just the advancement of agentic AI. Really powered by these incredible reasoning models. So so many things now that can be done. Like 18 months ago, it was really hard to automate a process end-to-end using a gentic AI. Now we're looking at, you know, agents that are true that are truly autonomous, like according to the anthropic definition that they're, you know, they dynamically choose tools, you know, make decisions with minimal human intervention or almost none. We're starting to see those come online, right? And again, this is where constraints will come in in terms of like how far or how how autonomous can these agents be? How autonomous will the companies let them be? And how autonomous will anyone who's building agents actually be comfortable letting them be?

SPEAKER_01 22:56

Yeah, absolutely. From a design and from a product standpoint and an engineering standpoint, you're no longer just building product and experiences for humans, right? Your customers, you're actually also going to have to support agents as well. As we start to see more computer use agents in place, basically doing what humans are doing and, you know, booking travel for us and doing stuff like that, companies are gonna have to make sure that their websites are able to support these agents as well.

SPEAKER_00 23:25

Yeah, you know, and on that, like I think I I had seen uh in one of these like newsletters that I get, AX. So I was thinking, what is AX? Autonomous experience? And I started to look that up, and I saw that there's a whole conference on agentic experience. So how do you design your digital platforms for a good, not user experience, but agentic experience? Yeah. Like, I mean, it's fascinating how the UX practice will actually have to shift to accommodate the fact that they're the users of platforms are not necessarily humans, but agents, right? Yeah. So that's something like a future trend, right? Maybe if we come back two or three years, yeah, and have this conversation again around AX, how that seemed in two or three years, it might be might have sounded so abstract. Yeah, in two or three years, it's part of our um, you know, our our design toolkit, right? So yeah. So like just like big data now sounds quaint, yeah, where it was quite abstract in 2016 for the companies we were engaged with, AX might sound abstract now, but in two or three years it would be very quaint, right?

SPEAKER_01 24:51

Yeah, we might have more designers actually building.

SPEAKER_00 24:54

We might have more agents, I guess, we might be building for AX as designers, right? So totally, yeah.

SPEAKER_01 25:02

I love it. In your experience, what is the strongest predictor that an AI proof of concept will actually become a real impactful product that scales?

SPEAKER_00 25:12

Uh well, it's a tough one, right? Because I think there are three. One, it's like, does this solve a real problem that is painful for the business? So I think like a lot of proof of concepts are great and they teach you a lot about what the capabilities of the technology are. But ultimately, like if after come like it doesn't come out of the lab if it can't prove that it solves a problem that is really pressing for the business. Two, one of the ways I've seen POCs fail is a lack of proper success criteria. Like, how do we know if we're actually solving the problem? Right. And because things are experimental, people don't want to limit the POC with like metrics. But if I can't measure if this is solving the problem, solving it better than before, I'm not likely to invest in it, especially as a a business. So I actually would say, you know, out of those two, the success criteria is a really important one to come to the table with understanding how will I judge this POC as a success. Yeah. It may not be exact, you may change it over time, especially when it the solution's in production. But to come to the table with like, let's just find out what we learned makes it more likely that the POC will stay in POC stage than actually move into production.

SPEAKER_01 26:41

Yeah. So that's almost like innovation for the sake of innovation versus innovation for actual impact.

SPEAKER_00 26:47

Yeah. And so I mean, like, both have their place, right?

SPEAKER_01 26:52

Yeah.

SPEAKER_00 26:52

And so I think, you know, sometimes you do the innovation for innovation's sake to learn about capabilities, options, get new ideas. So I would I would definitely never knock that. But from my perspective, I think, you know, and this is a lot of the work I did when I was in leading our design thinking practice within Morgan Stanley, it was let's let's define the problem really well. That's 50% of the the challenge. And then let's start to build decide or determine solutions that meet the problem. Yeah. Am I solving a real problem that's causing pain? Right. Am I cognizant of the the constraints that will be placed in the real world, right? When this is out of the labs, will my proof of concept stand up to those constraints, right? Yeah. Yeah. And you know, now that we're on this conversation on constraints, I've been thinking about that a lot because, you know, I think sometimes constraints can be used to weaponize or shut down innovation, right? Oh, well, we can't do that. We, you know, we have this regulation, we cannot do this, we cannot do that. And I think the the framing and timing in which you introduce the constraints, the, you know, is really critical in innovation.

SPEAKER_01 28:15

We're going to close with this quick segment that we do on Rendered called Would You Ship? So, Rosedell, I'm gonna pitch you a futuristic product idea in your space. Okay, and you tell me if you would ship it. Okay, so here is the idea for you. What if we treated AI compute like a public good? Sort of how we treat libraries. So imagine a global library organized by the UN Sustainable Development Goals. So there are 17 SDG goals today. So imagine a library with 17 floors, right? So floor one would be the no poverty floor, floor two is the zero hunger floor, all the way up to 17. But instead of borrowing books, organizations can borrow these impact kits, which would have these ready-to-deploy AI tools, the safety guardrails, the deployment playbook, plus this time box compute loan to actually run these programs. And the key here is that they don't need to pay in dollars. Instead, they could pay with proof of impact. So if their pilot measurably improves outcomes, you know, like lower admissions, higher school enrollment, better health outcomes, et cetera, you earn more compute and access to better kits. And the funding could be blended. We could have cloud providers donating discounted or off-peak compute. Foundations could fund the governance. Fortune 500s could join as members to sponsor specific SDG floors of the library that they're excited about. So I can kind of just walk through a really quick example, right? Uh, let's say a company or a region wants to focus on SDG too, zero hunger. So they could go to the, you know, the the second floor. They could go to the look at all the kits. And so one could be, you know, helping farmers spot crop disease early via photos, or having a kit that's focused on school meals forecasting so they can predict demand so kids get fed with less waste, or a third kit, which could be like a food rescue routing kit that can match surplus from grocers and restaurants to shelters in real time. So essentially like a public library for AI good that's designed to ship solutions safely. Yeah. So, Rosdell, what do you think? Would you ship this?

SPEAKER_00 30:35

Yes, I would. I love it. And I really love the idea of time box compute. Yes. Yeah. A couple this was in December. I'd gone to a round table on AI and diplomacy, and I ended up talking about how the comp the global compute divide is the new digital divide. Okay. So I love how this addresses that issue of compute divide because it's a real challenge, especially for emerging and developing countries, to contribute their own solutions to key challenges, right? Using using AI, particularly generative AI. So the the fact that you have like this compute built-in is great.

SPEAKER_01 31:25

Yeah, because conduit's almost the new form of currency, right? And you know, I think when people hear about AI, there is this whole narrative about AI abundance. But we talked about this actually a little bit at that lunch, right? In order for that to happen, right, AI would fully need to be democratized, and with that also compute. And so I think right now that seems like, you know, as we talk about constraints, that seems like one of the biggest constraints.

SPEAKER_00 31:51

Yeah, biggest constraints, like the compute, the the energy required, the water, the data. But again, going back to compute as a driver of or constraints as a driver of creativity, you know, in a lot of countries, and particularly I'm seeing this trend like in West and East Africa, so in Ghana, in in Kenya, where they're building small language models. Um yeah, which are very efficient and using things like mixture of experts kind of architecture to still be highly performant. So one of the things I talked about in this um in this round table was this trend towards quantization. I'm sure your audience is very technical, but in essence, like almost compressing the models so that they still perform highly. So like compressing the pixels in your like photo so it can fit in your phone, but it still looks great. So thinking about the design choices and the architecture of language models to fit the constraint of min of like minimal access to compute and data and energy. Yes, another way to start using those constraints as creativity to drive broader access.

SPEAKER_01 33:17

I I love that. I completely agree with that. You know, I know people who go to ChatGPT to ask what the weather is, right? So I think that's definitely the direction we're heading. And with that, um, focusing on, you know, optimizations so we don't just have to continue. So maybe one solution is we put, you know, data centers off of Earth, and you know, we can do all that. But in the meantime, there are things I think that we could explore to make these models and these processes a little bit more efficient. Yeah, for sure. Great. Okay, so let's go ahead. Let's ship this product.

SPEAKER_00 33:50

All right. I love it. I love it.

SPEAKER_01 33:53

Awesome. Well, that is the end of the episode. Rosedell, thank you so much for joining.

SPEAKER_00 33:59

Thank you so much, Pranitia. I really appreciated the opportunity. Great talking to you and the audience. Absolutely.