The USP for the EU Digital Identity Wallet, and AI Agents are the new users
Plus: Your AI Agent will probably blackmail you, and when will the digital ID market take off?
Hi everyone, thanks for coming back to Customer Futures.
Each week I unpack the disruptive shifts around Empowerment Tech. Digital wallets, Personal AI and the future of the digital customer relationship.
If you haven’t yet signed up, why not subscribe:
Hi folks,
Welcome to the great AI Agent hype cycle.
Where sky-high expectations are now colliding with the limitations of the tech.
Gartner now predicts that over 40% of current AI agent initiatives will be scrapped by 2027. Due to cost overruns, disappointing returns, and mounting operational risks.
At the same time, new research from Carnegie Mellon has laid bare today’s technical ceiling. Even the most advanced agent models struggle with routine tasks like browsing the web, writing code, or managing applications. The best of them manage only 30% accuracy. Gemini-2.5-Pro tops the list, followed by Claude-3.5-Sonnet (26%), and GPT-4o and Llama barely getting to single digits.
That’s quite a reality check.
None of it is stopping the headlines, though.
This week Salesforce CEO Marc Benioff claimed that half of the company’s work is now being done by AI. And with a 95% success rate. And over at Microsoft, Mustafa Suleyman (head of Consumer AI, previously a founder at Deepmind) suggests their new model can outperform doctors by 4x.
So what’s marketing spin, what’s real, and what’s actually making a difference in the increasingly crowded AI Agent arms race?
Despite the froth, something deeper is happening.
While the early numbers might be unimpressive, the broader shift is unmistakable. AI Agent systems are starting to appear across every layer of business, and investors are betting on consumer applications for everyday life.
The real question isn’t whether they’re coming, but how we choose to adopt them, and at what cost.
AI agents alone don’t make a system empowering. They’re just one component in a broader shift towards Empowerment Tech. AI agents will need data, and lots of it. It’s why they’ll need to be paired with digital wallets and personal data stores. So that AI Agents can be trained on trusted personal data sets, and so that they - and the customers they serve - can prove who (or what) they are, what they’ve done, and what they’re entitled to do.
I’m still optimistic.
Why? Because the two original sins of the web were no payments and no identity. And right now, AI Agents have neither. Certainly not at scale, or trusted enough to be let out of the lab.
But both are coming. They have to.
The hype cycle is inevitable. As is the coming crash into the ‘trough’. The question then becomes: how long will it take? And where and how is value going to be created?
I’m betting that much of the value is going to show up on the side of the customer. And soon.
As ever, it’s all part of exploring the future of being a digital customer. So welcome back to the Customer Futures newsletter.
In this week’s edition:
Is this the USP for the EU Digital Identity Wallet?
Personalisation isn’t just about the message - it’s about the timing
Yes, but when will the digital ID market take off?
Digital trust isn’t owned, it’s rented
Agents are the new users - and like us, they’ll need their own digital ID
… and much more
So pick up a Chai Latte or smoothie, cosy up in your favourite corner, and Let’s Go.
Is this the USP for the EU Digital Identity Wallet?
For over 15 years, I’ve been working on and around the ‘commercial case’ for customer data sharing. From data stores to digital wallets, from ‘next generation intermediaries’ to Personal AI Agents.
And I can see 11 different revenue streams when customers are in control.
Andy Tobin and the team at EWC, the Large Scale Pilot group investigating the EU Digital ID Wallet, are now deeply exploring this very thing: The commercial model implications and opportunities for EU businesses and individuals.
Researching it with 10,000 citizens, and in 18 countries, no less.
Aside from agreeing the technical standards, and avoiding accidental state-wide surveillance, the commercial models around the EUDIW might just be the most important conversation right now.
Because without incentives, without understanding the value created, the EU’s digital identity ambitions will fade. Or perhaps worse, they will be abused like the cookie consent fiasco that came out of the GDPR.
Not only does the digital wallet, the identity information and the data flows need to be secure and scalable. They need to be valuable.
Andy Tobin writes:
“The eIDAS wallet needs a unique selling point (USP). Our research shows that if it just does what other wallets do, adoption will be very low.
“It WILL be possible to verify the authenticity of organisations that are asking you for data.
“Let’s extend that to person-to-person verification as well. There are so many use cases. Is this person on Facebook Marketplace a bot? Is this person I am dating who they say they are? Is this person I am buying a used car from the same person registered as the owner? This would be a real USP and would drive huge uptake and usage.
“Three simple person-to-person challenge responses could be built into every eIDAS wallet:
I am a real person (not a bot)
I am a real person and my name is first name + last name
I am a real person and my age is >18 (or 21 etc)
“Each request for this data comes with proof from the requestor about their data. So if you ask if the other person is a real person, you send proof that you are also a real person. Reciprocity.”
Andy is doing a couple of clever things here.
First, he’s pointing out that, as currently framed and proposed, these new eIDAS2 wallets won’t get adopted by citizens or the market. Not as hoped.
Second, he’s reminding us that business ID is going to be critical. And that the EUDIW will - and must - be able to verify businesses too, not just people. That’s going to be central to building trust, and to driving adoption.
But third, he’s opening up a new front on the battlefield for valuable wallet use cases. The important idea that these EU identity wallets can also work peer-to-peer. Citizen-to-citizen.
Like I said, there are multiple ways to think about how digital wallets and customer-controlled data can create value. And that there are 10+ new revenue streams to explore.
But as with so many of these things, we must follow the money. Not because the wallet must be a commercial object (“if the product is free, you’re the product” etc.). But because without incentives for people to use the new EU digital ID wallet - and ‘logging into government services’ isn’t going to be one of them - it’s hard to see how this thing will take off.
Many of us have been saying it for years.
The EWC report on the EUDIW commercial models will be out later this year.
And it’s going to be important reading.
Personalisation isn’t just about the message - it’s about the timing
For years, the ad industry has followed one big assumption. The more targeted the ad, the better. Relevance drives results. Precision drives conversions.
But what if that’s wrong? What if hyper-targeted ads are actually putting people off?
The American Marketing Association is now pointing to new research that asks a simple question. Does showing someone an ad for the exact product they just looked at increase the chance they’ll buy it?
It turns out, not really.
In fact, across the board, generic brand ads outperformed personalised product ads. The only time ‘hyper-targeted’ retargeting worked? It was after someone had already done the work. After they had browsed a bunch of comparison sites, or read the reviews.
In other words, after the customer had already narrowed down the options themselves.
Before that point, generic ads were more effective. More useful. More welcome. So maybe personalisation isn’t just about being right. It’s about being right at the right time.
And that means understanding the customer’s context.
Not just preferences, and certainly not just clicks. But understanding where the customer is in their journey, what matters to them in that moment, and what kind of message they’re open to hearing.
To get that right, brands need better data. Smarter data. Data that’s timely, accurate, and permissioned.
Which is exactly where Empowerment Tech comes in.
Digital wallets. Verifiable data. Personal AI. Tools that don’t just collect information, but help individuals share it. On their terms, at the right moment, with the right guardrails in place.
Not every brand interaction should be a pitch. And not every message needs to close a sale.
But if a brand wants to be relevant? Truly relevant? Messaging fast becomes about when and why, not just who and what.
Yes, but when will the digital ID market take off?
It might be the question I get asked about the most.
When - or indeed if - the decentralised ID market, and Empowerment Tech more broadly, will take off.
Michael Cooper at UNCX has written an excellent piece on how and when digital ID will evolve. Specifically calling out the evolution of Digital ID across a number of stages.
Here’s his full list:
Present Day (2025): We're currently in a transitional phase with:
Fragmented identity solutions
Password managers and biometric authentication
Single sign-on options (Google, Apple, Facebook)
Early decentralised identity experiments
Near Term (2025-2026):
Digital identity wallets from major platforms
More government-issued digital IDs
Integration of verifiable credentials standards
Early adoption of personal AI identity management
Medium Term (2027-2028):
Cross-platform identity portability
Wider acceptance of decentralised identifiers
AI-managed identity permissions becoming common
Physical-digital identity convergence
Longer Term (2029-2030):
Mature digital identity ecosystems
AI-to-AI identity negotiation
Seamless authentication across contexts
Identity as fundamental digital infrastructure
I think he’s making a pretty good bet. And a more sober - and I more suspect realistic - view than Gartner, who believes there will be 500M digital ID wallets being used monthly by 2026.
Read that again. 500M digital wallets offered? Maybe. Actually issued? Perhaps. But, to use their words, “…used monthly to present verifiable claims”…? That’s a tougher prediction.
I’d love to be wrong. But Dec 2026 is only 18 months away.
Here’s where I’m at:
1. We won’t see a ‘big bang’ adoption cycle
Rather, it will be in fits and starts. There will be a number of little tech explosions appearing on the customer side. Chaotically, in lots of different places, for lots of different reasons.
Like the web (and consumer browsers) before it, Empowerment Tech will be driven by a wonderful, diverse and fragmented set of use cases for individuals.
Then, over time, these hotspots of value will grow and merge. And we’ll unleash the strong network effects that many of us can see coming with Empowerment Tech.
Join for the identity check, stay for the personalisation. Join for the discount, stay for the ability to prove your status elsewhere.
2. We’ll see an evolution of ET in our lives, not a revolution
There are many, many dependencies here. Several things need to be true for decentralised ID and Empowerment Tech to get adopted. From the user experience and business models, to the tech standards and liability frameworks.
Stitching all that together, and the designs and network building, is going to take time.
3. The market will snap into place following a breakthrough ET innovation
We must remember that Facebook was the 7th social network. And that market disruption usually happens when a new business creates a new interface or new channel to use existing tech.
Over the next year or so, one or two providers will put together existing capabilities in new ways. And we’ll wonder why it hadn’t been done before.
Maybe that will be at the payment checkout with AI Agents. Maybe it will be with a new Digital Staff Passport. But something unpredictable will open our eyes to what’s possible. And we’ll see the early steps of a new $bn market.
I’m also betting that it will surprise us, because it will come from somewhere outside our current field of vision. Maybe we’re looking in the wrong place.
Maybe it’s a disruptive new way to do KYC, invented in Nigeria (230M people). Or a new peer-to-peer vouching network comes out of Brazil (211M people).
Are you paying attention?
Unknown unknowns.
It’s an exciting time. But it’s going to take time.
For all the excitement about the tech, know that ET is about people. About being human. It’s an age-old story about human progress.
Yes, we might be able to see what’s coming. But understanding when it will happen is, of course, the hard question that everyone’s now asking.
UNCX PREDICTION, GARTNER’S 500M
Digital trust isn’t owned, it’s rented
A lovely short piece from Alex Walz on why trust is temporary, and must be earned.
“Every click, every “yes,” every “I’ll share my data” is a decision. Because trust isn’t owned. It’s rented. And the terms are rewritten with every interaction.
“Betray it, and they’re gone. Ignore them, and they’ll forget you. Try to manipulate them, and they’ll never forgive you.
“The companies that win tomorrow will not be the ones with the largest marketing budgets. They’ll be the ones with the clearest intent. The cleanest experiences. The courage to give control back to the customer.
“Because in the end, you’re not selling a product. You’re selling a relationship. And relationships are built on respect.”
Agents are the new users - and like us, they’ll need their own digital ID
Hot topic of the moment is how agents can prove what they are, and who they work for.
I came across this idea of ‘agentic ID’ from Strata. Showing why existing ID approaches like MFA, passkeys, IAM with RBAC etc. are going to struggle when agents are temporary, dynamic and flexible.
Their table below compares Human Identity, ‘Non-human’ Identity (NHI) and Agentic Identity:
The point of ‘agent lifespan’ is really important. Because many AI Agents are going to be spun up in real time, for a specific customer need. Just In Time.
And they are going to need ‘delegated authority,’ to work On Behalf Of (or ‘OBO’) people, corporations, and indeed other agents.
I like Eric Olden’s view on how it’s going to work, with the “Six A’s”:
Authentication: Prove identity via SPIFFE, JWT, PKCE
Access Control: Limit actions via scoped API policies
Authorization: Support for OAuth On-Behalf-Of (OBO) delegation
Auditing: Trace every agent interaction and decision
Administration: JIT register agents in any IDP
Availability: Ensure identity continuity even if IDP fails
Here’s the thing. We are going to need lots of different perspectives on agent ID to map all this out. Right now, we’re only in the foothills of this stuff, and it’s going to take time.
Some folks are working on using ‘OAuth’ identity tech for AI Agents. Some are betting on wallets and credentials. Either way, we not only need to understand how agent ID will work, but also why, when and where.
Agents, like people, will be naturally decentralised. So we’ll need new ways to verify who is who and what is what, without having to wait 15 mins to perform another ID check, by another centralised intermediary.
Agent tech is moving fast. The use cases for consumer co-pilots are now everywhere. And investment is already flowing.
But AI Agents won’t be trusted unless and until we can verify them just like we verify people. But we can’t do it the same way. And we can’t use the same mental models and tech.
More coming on all this soon. But I’m more excited than I’ve ever been for the ID community.
This is our time.
OTHER THINGS
There are far too many interesting and important Customer Futures things to include this week.
So here are some more links to chew on:
Post: The GOV.UK App Is Not A Digital Wallet, and Is Just a Bunch of Web Pages READ
Article: The Business Benefits of The European Business Wallet READ
Research: Your AI Agent will probably blackmail you READ
Post: Agentic Commerce: Shopping Assistant or Brand Assassin? READ
Article: The Privacy Americans Are About to Lose READ
News: CloudFlare is now allowing websites to charge AI engines for crawling their sites READ
And that’s a wrap. Stay tuned for more Customer Futures soon, both here and over at LinkedIn.
And if you’re not yet signed up, why not subscribe: