AI Agents, digital identity, wallets and personal data
An interview with Jamie Smith (May 8, 2025)
The following is a transcript of our December 16th 2024 interview with Jamie Smith on Masters of Privacy. The original recording can be found on the podcast’s website, as well as on your favorite podcasting feed.
Jamie Smith is the CEO and Founder of Customer Futures Ltd, a company focused on digital identity and customer-controlled personal data. He has been working at the forefront of digital transformation for nearly 15 years, helping deliver innovative solutions for some of the world’s largest organizations. Jamie has previously worked at Evernym, Ctrl-Shift, BT and Deloitte, before embarking on various recent projects, always in the same space.
INTERVIEW
Sergio Maldonado
Jamie, welcome. We’ve spoken, we met when I was in London. I really enjoyed it. And then you’ve kept on building on top of these ideas, agency and control, things that I’m very passionate about. Where are we now? What’s the most exciting thing for you now in this space?
Jamie Smith
Thanks for having me. There’s a funny thing with innovation where you can’t know everything. So either you know what will happen with certainty, but you don’t know what it means. Or you don’t know. So a regulation comes in or there’s going to be a technical breakthrough, but you don’t know what the impact will be. Or you know exactly that something will happen on a specific date.
And you’ll say, actually, how do we understand what’s going to happen after that? We don’t know what’s going to happen. So I think one of the reasons I’ve been so optimistic about this shift is I don’t know exactly when it’s going to happen, but I think directionally, we can be really clear that individuals are going to be more empowered with their data. If you look at every data protection regulation, the technology is getting better and better at the edge. It wasn’t possible for individuals to control their data. And when you talked about digital trust, it was about an organization making the privacy policy more readable, right, or getting the checkbox better, but it was still happening on the business side. And the reason I’m optimistic is things are going, now, technology is getting cheaper and faster and better on the customer side. And that means directionally, we can be sure that it’s only going to become more useful for individuals to have their data.
And so then the question is, okay, what does that mean? In terms of what am I most excited about? I was actually just posting recently, you people talk about how payments is the kind of killer app for identity, you know? And, you know, what do you tell? Because having an identity is kind of, nobody really cares about that thing. Like, I have a document, I cross a border every so often. Most companies don’t need to know who I am. They care what I want, my preferences, my payments and so on. So payments is the killer app for identity. And I think the last five or 10 years have been about data stores and wallets and me having my own data. But there’s been a challenge getting adoption. Like why will people have it? They don’t want to manage their data. A single digit number of people care about managing their data in a spreadsheet, using conditional formatting because that’s the only tool that people have for those things. But once there’s utility, then we can create value for the data. We can actually do something with the data. And that’s where AI and AI agents specifically are going to help. The punchline, I suppose, is the thing I’m excited about is AI agents becoming the killer app for digital wallets. Because now I can actually not only have the data, but I can act on it. And it’s not just a pie chart that tells me how much I spent on coffee last month.
Please give me recommendations on where my family should go on holiday next year. Please cancel all the subscriptions that I’ve forgotten about. Please prioritise my inbox and my messages, not just for things that are important, the people I last spoke to, but the people that are contextually, the things that are contextually relevant. You know, I heard a great expression that your inbox, your email inbox is your to-do list organised by someone else. You know?
And I need an AI agent just to make sense of my WhatsApp, my LinkedIn, my Telegram, my Signal, my email, right? And I miss stuff. We all do. We’re busy. And I want an AI agent just to help orchestrate that stuff. And that’s where I’m most excited and interested.
Sergio Maldonado
It is true. Five years ago, we thought about control and the best breakthroughs we had, a few apps that did work for a number of years, they were all about visualizing data. And now we finally have a purpose. So if we look at identity before we move into what the agent can do for you, then there’s this key piece of identity. And we’ve had an interview here with Adrian Dirk a few months ago, the case for using a blockchain based identity had not been strong enough as he explained. Then I spoke recently to Dan Stone from Icebreaker and he made a very good case for it. Where are you?
Jamie Smith
I think there’s, you know, the decentralized technology has been a natural counter argument or counterweight to having things centralized. You know, there’s too much centralized, there’s too much control. And the Web3 maxis, right, the people who, you know, not your keys, not your crypto have swung entirely the other way. I want to be my own bank, which is crazy, right? Because as society, we need governance, we need rules, we need to understand what you can or can’t do.
We need liability models, right? All that stuff. And so I think there was a, in 2014 to 2022, there’s kind of this real swing the other way of like, let’s decentralize all the things. I know, even in the identity world, like, well, if my identity is not on my device, under my control, you know, someone can turn me off, all that stuff. And I think there’s starting to be pragmatism creeping back in that says,
What does minimum viable centralization look like? Like how can we have things that are pragmatic and scale, but that have some of the benefits of decentralization? So that’s one angle. The other is people are also recognizing that blockchains arguably are also more lock-in, right? I’ve got to be on this ledger or that ledger. We’re building another ecosystem. And guess what? There’s an explosion of ledgers and tokens and coins and these little ecosystems where I’ve got to use your signature scheme. So you know, there still needs to be, even though things are decentralized, I’m still having to move, you know, value or my interaction across these different ledgers or these different blockchain DLT ecosystems. So there are some, I’m going to say bleeding edge, they’re not there yet, but there’s some new protocols coming out that you can get all the benefits of decentralization without needing a blockchain at all.
In fact, some of those that were championing the original SSI, so self sovereign identity movement, saying, no, blockchains are more locked in. That’s a bad idea. What we need is a protocol that means we can get all the key management and all that kind of complicated stuff and the benefits of that without needing another single blockchain. So that’s another angle. But one of the last things I’ll point to, which I think is interesting is when you look at, you if you zoom out for the last 10 years, all the blockchain labs, studios, garages, accelerators, there are usually two, three or four use cases that kind of resolves. If you stand back and close one eye, the first one was a cross border settlement. It’s reasonable. The second one was supply chain. And the third one was identity. But if you look at cross border settlement and supply chain, there are identity problems. So really, is this whole thing of decentralised trust and not having a central intermediary give you the identifier or the reputation of who’s involved? So again, back to the first thing we talked about, like directionally. Centralization is needed for some things, for administration, for liability, for auditability. But there might be some new technologies coming that mean we can get some of those benefits without needing a centralized controller. So that’s the little story, but there’s a really interesting thing going on around the post web. It’s a movement that’s being kind of pushed forward by a company called Outlier Ventures, a guy called Jamie Burke. And they started in 2014, 15 saying, it’s all going to be blockchain and AI, like proper buzzword bingo, right? Blockchain and AI.
But what they’re now saying is, you know, web one was read, web two was write, and web three was own. Well, with AI agents, what happens next? And what we get is they would say delegate is the word. So read, write, own, delegate. That needs identity and it needs portable data and it needs all that good stuff. But that means the web was designed and built for people filling out forms, in big centralized, what’s become centralized platforms. And the idea of maybe we move to the post web.
So, but the reason I’m linking this back to blockchain is like, depending on who you pay attention to, know, there’s either one million active wallets, Web3 wallets, the wallets are 30 million, but it’s like a tiny, tiny number and it’s not had the business impact, it’s not the economic impact. But if you reframe blockchain as a backend transformation, then we can have AI agents acting within foreign individuals and organizations and settling stuff, managing supply chains, thinking about identity in this new way. there are some threads there you can pull together to say, where am I on blockchains? I think there’s an identity movement that’s moving post blockchain, post DLT. I think for some things like some of the public key cryptography, blockchain is not a bad answer for some public registries rather than doing things centralised using public key infrastructure, PKI. But the post web is where I’m starting to think Jamie Burke and Outlier Ventures might be right that agents plus the ability to settle payments can be a kind of a new wave of what might be coming in the next 10 years.
Sergio Maldonado
It is true. Everything we’ve built, our interfaces, our phones, the shape of our phones, it’s all built around websites and the web and the internet and the way we network has been built around our limitations, our senses. So if you were to go behind the scenes, then you will be using all these APIs. But someone said, or whichever systems for interoperability and instead of data machine to machine, so you just use an agent. But someone said the other day that this was even harder than the human-built world. And I really liked that. The fact that you’ve got agents that can cut through that and be used to dealing with the human built UI, it’s actually better because they struggle even more when we try to find governance and compatibility and interoperability at the data layer. if we combine all of this and we set ourselves in this future, in five years, we’ve got a new kind of interface with the world and the world is made of other people that we interact with, for example, as you were saying, email, and the world is made of data systems, the environment and the agent is for us to deal with that world. It’s to be productive in the work environment, to socialize. What kind of device, what kind of agent is the best possible agent in your view?
Jamie Smith
I mean the honest answer is I don’t know, I think things are going to get adopted when they create value. And wallets on their own weren’t creating value, except for people who are wanting to manage cryptocurrency. Wallets on their own, having, you know, even having an app with a proof of doctor on it, it’s kind of useless unless I can present it somewhere. Unless I can use it, right? So we need acceptance, just like in payments. Having a credit card is useless, having the acceptance network is the important part. So part one of your question is, where the agent’s gonna show up and which one’s gonna be valuable. I think it’s where they solve problems for individuals, but that won’t happen until we have governance. Anthropix got a product called Claude and they released 3.5 Sonnet, which has an AI agent, which can run in the browser. So it can see the page, it can click on a button, it can type the text, and you can give an instruction, make a booking, at a restaurant in New York on Thursday, right? So it can, the best restaurant for me. So it can go in and type in restaurants in an area. It’ll then look at the Google search results. It’ll click on it. It’ll then pull up the address, to the website, pull up the address, take that into a map, work out the distance. It can do all these things, right? It breaks the task into small, bite-sized chunks that one agent can go and fix. So Google just at their big announcement they announced Google Mariner, which is their version of this. And in the release they’re going to great lengths to say this is experimental, it’s a research prototype, they’re playing around, but they limit, you cannot enter billing information, payment details, right? It can’t sign contracts, it can’t make a transaction for you. And that’s because there isn’t governance yet, which really means that they say it’s about control, the individual needs to be in control.
But I think that’s really just about liability and governance, right? can’t, because fraud is going to balloon once these agents can start making payments. So where I’m going with this is, what am I excited about in agents? There’s going to be natural limits, right? So technical limits, speed limits. For example, in the AI agent that Google released, it can only act if the browser is front and center. If you change tabs or change programs, the agent won’t be working anymore. And they say that’s for control and visibility. It’s good. These are baby steps. So we need the UI UX to be sorted. We need the liability and governance model to be sorted about what an agent can and can’t do. So people are saying these dreams of, you know, it can go and make the right recommendation of the flight and plan my whole holiday and book and not until we’ve cracked the UI UX, the commercial model, the governance layer, how liability is going to work. And my answer to some of that is, if it’s being run by a big centralized organization, that will be good because we’ll have bounds to it, so it’ll work within the Apple interface or within Google. But it’ll also be prevented because their compliance teams are going to say, no way will I let my agent run these things. So back to the very first point, directionally, it makes sense for the individual to have agents that work on their behalf. But we’ve got a really big problem around governance and around identity. Again, I wrote in the customer features newsletter yesterday. I pointed to a comment, there’s an identity company that does a wallet decentralized on blockchain called Privado ID. And one of the co-founders, Evan McMullen said, a great thing is AI is going to need accountability, but accountability needs identity.
It’s totally true. So my bet is that agents are naturally going to be decentralized things because they will be limited in what they can do unless they’re decentralized. But we can’t have a centralized identity infrastructure, which means that wallets and credentials are going to be, you know, the icebreaker thing is going to be more important. And you can start to see how it comes together. You can start to see our agents that are running on my device run by me that have a fiduciary responsibility to me. Because Apple’s agents and Anthropics agents are going to have a fiduciary responsibility to the platform, not to me and to the stakeholders, right?
So there’s a guy in California called Dazza Greenwood who’s writing about fiduciary responsibility in agents and he is a legal expert and a technical expert and AI expert. I’d recommend the listeners and the watchers to check out Dazza Greenwood who’s writing papers on this stuff. This is where I think it’s going to happen. It’s going to create value, but only when we take these barriers out of the way. And that’s going to, you know, it’ll take time, but I think, you know, in two years we won’t make much progress, in 10 we will.
Sergio Maldonado
So the agent is going to be agreeing to terms and conditions on your behalf.
Jamie Smith
Well, so I mean, again, it’s baby steps. So the analogy I’ve been thinking about is, you know, autonomous cars, self-driving cars, they still have a steering wheel. Why? Because we have retrofitted the autonomous capability to an existing vehicle. Right, we’ve not redesigned a car to be completely autonomous yet, mainly because of regulation that’s possible today, right? Or governance liability. And so, agents are going to be retrofitted to the things we have. That’s why we have those browser extensions. They don’t need to be a browser extension. But we have them as a browser extension. We have them on our devices, on our laptops, because of UI, UX. But it won’t yet be able to sign contracts because the compliance teams won’t let them. Right? They won’t let… How do we know it was Jamie? How do we know it’s… They have a problem right now saying, how do we know it’s Jamie who signed the contract? Let alone if it was Jamie’s AI agent and not someone else’s AI agent. So we’re going to need wallet and identity infrastructure to make it possible before the compliance teams will accept the contract. However, some baby steps towards it. One is that my AI agent can look through the terms and conditions and make sense of it for me before I sign. So, you know, there’s a great thing. We’re at peak T’s and C’s at the moment, right? I think something like I’m going to make the number up, like Terms and Conditions, Ts and Cs. I think it takes something like 13 hours to read, normal reading speed, because they’re written for lawyers by lawyers. We’re reaching peak Ts and Cs, peak consent fatigue, because I have to swat away the consent, which I’m not, it’s not really meaningful consent, not really meaningful Ts and Cs. So my agent’s going to start making sense of those things.
Now that’s like having a steering wheel in the car. It’s kind of a little baby step towards having an agent sign it for me, but I’m going to be a bit more aware. But the second point I was going to make about consent and contracts is again, this guy, Dazza Greenwood, who I encourage you to go and check out, he set up the law program at MIT, so law.mit.edu. He drafted or helped draft some regulation that is already in place in the 52 states that has language that permits an agent or a computer system to sign a contract that is already in law.
Sergio Maldonado
Finally, we’re going to be able to have a distillation. What is it that you really want of me? You know, what kind of liability you want to take off your back and pass on to me and let my system see what I am ready to accept. Contracts are going to be much more efficient.
Jamie Smith
Totally right. Better for transparency, better for accountability. And the thing I was looking up just while you that I wanted to remember was this thing called the Uniform Electronic Transactions Act, the UETA. And it’s an existing legal framework and rules for electronic signatures, automated transactions and the attribution of acts of electronic agents or autonomous agents. So this stuff, like I said, we’re putting these pieces slowly together and I think it’s going to make the foundations are being put in place. The car with the steering wheel with the autonomous software, it starts to just help me with the lanes and then it helps me reverse and then it can start helping me navigate across the city. So right now we’re putting an agent into a browser.
Right now we have basic regulation around electronic transactions and document signing. We can start to see how they build together, which is why we come full circle to where I think we’re going, which will be wallets and agents acting on behalf of the individual. If you just share responsibility to me, probably a commercial model that’s around freemium or I’m paying for the service. that organizations are going to have to get whiplash because they have spent the last 10 years, 20 years stopping the bots coming in. Right? They have been putting up all the barriers, all the alligators and the barbed wire and the checks and the mother’s maiden name and all that stuff to stop the bots coming in. And now they’re going to do a 180. Because now they have to check if it requires a bot that’s official? Is it a bot that’s been approved by Jamie? Does it have consent to act on Jamie’s behalf? So guess what? We need wallets and credentials and authorization and these things are, we have the raw material for it, but that’s what’s going to be the challenging next phase.
Sergio Maldonado
I love that. Thank you, Jamie.
Jamie Smith
You’re very welcome. It’s always a pleasure to talk.