How might AI agents be found and identified?
Over the past few months in AI Studio, we’ve carried out research with human “agents” – carers, teachers, charity workers and members of informal networks – to understand how agency already works in the real, non-digital world.
One commonality we found is that successful agents are ones that people can find, identify and refer to. In other words, agents need to be addressable.
Human agents have signals that help us to discover them: they stand behind a service desk, wear an official uniform, or can be reached by calling a given phone number or visiting a website.
But what might it mean for AI agents to have a real, recognisable presence that people can point to, share, and act on?
The ideas that follow are explorations intended to spark the imagination and to guide our future designs.
How might we give agents a presence in public spaces?
Our research found that help is recognised through trusted presence, not destinations. Familiar places like libraries and GP surgeries already act as informal findability hubs.
So, one way to make an agent addressable might be to give it a physical identifier similar to a QR code that could be used in such places.
We could imagine a code like this printed on letters, displayed on a screen, or placed on a lamp post – perhaps even using the Crown logo to give it a distinct and recognisable design.
When scanned, it could open content that is contextually relevant to your circumstances, location, and desired next step.
In this scenario, each marker becomes a starting point, tied to a campaign, a specific service, or a planned sequence of support. The places people already go become entry points, not just buildings you visit.
How might we use existing touchpoints?
Government already sends millions of letters, emails, and documents. Complex life events like moving house or dealing with probate generate a mountain of paperwork – PDFs, emails, receipts, confirmations.
What if you could ask an AI agent to turn artefacts like this into something structured and genuinely useful?
By extracting dates, names, and references, it could build a coherent timeline, turning a chaotic inbox into a clear project view. Think of it like a travel itinerary app, but for the big moments in your life.
Designing for addressability: what comes next?
Our research consistently found that people do not think in terms of departments or platforms – they think in terms of getting help, through whatever channel is closest to hand.
It’s common for people to discover government services through informal networks – librarians, community leaders, friends – and not just official channels.
That means that above all, any future solutions to making agents addressable must take this into account, and be emphatic about meeting people where they are.
The ideas shared here can help to shape and inform our work as we explore the future of human-AI interactions around government services, and identify the right questions to ask.