The people you want in the room when you think about AI.
Most AI vendors will sell you a tool. Most consultancies will sell you a strategy deck. Neither will sit with you while you think, push back on the obvious answers, and help you spot the problems worth working on.
That is the room we want to be in. In practice: we connect what your organisation already knows – documents, data, decisions, conversations – into a living company memory, and put AI to work on top of it, inside the tools you already use.
Where to start
Start with one recurring task.
Your team gathers information, prepares briefings, and answers questions from material scattered across the business. Often it is a senior person redoing the same piece of knowledge work by hand every day or week.
We start with one piece of that work. Together we investigate what AI can do, test it against real examples, and decide whether it is worth building into your day-to-day operations.
In practice
The same memory, in your sector.
The recurring tasks differ from one kind of organisation to another. Four pages describe what a company memory changes for each, built around work we have actually done.
Public affairs PR agencies Think tanks Corporate communications
The problem
The AI doesn’t know your business.
Most organisations have tried AI chat tools and hit the same wall. The tool is clever; it just knows nothing about you. It has never read your archive, doesn’t know your clients, and forgets everything between sessions.
Your organisation already produces everything AI needs – data, documents, decisions, conversations. We turn it into AI that does real work: answering, drafting, monitoring, and remembering, inside the tools you already use.
What we do
Your knowledge, put to work.
-
Years of survey data become an analyst
One that answers in seconds, cites its sources, and never tires of follow-up questions.
-
A parliament’s daily output becomes a briefing
Both chambers, every sitting day, in the inbox before the morning meeting. Read the case study.
-
A day’s coverage becomes a story map
Not clippings counted but narratives tracked as they develop – running daily for a communications team at a major bank. How we transformed media monitoring.
-
A discovery meeting becomes structured knowledge
Captured, connected, and available to every future conversation. The agent that changed our discovery meetings.
How we think
Efficiency is the entry point, never the destination.
Bolting AI onto an existing process buys you months of advantage. The deeper prize is the problems you couldn’t touch before – too expensive, too slow, or simply unimagined. Finding those is the real work, and it starts with thinking, not building.
Every design we ship is a conjecture. We state what would prove it wrong before we build it, and we let production evidence settle the argument.
The method
The intelligence cycle.
Listen
Map how your organisation actually operates – people, processes, systems – into a real-time blueprint.
Research
Design automation pathways that reflect how the business truly works. Each one is a hypothesis with a test.
Launch
Deploy pathways as working agents and automations. When no existing tool fits, we build one.
Measure
Look for evidence it doesn’t work, not just metrics that flatter. Retrain, revise, or retire.
Working with us
You hire a lab, not a licence.
We work in monthly R&D slots: a dedicated team of engineers, designers, and prompt specialists focused on your problems for set periods each month. We start by mapping how your organisation actually operates, connect your knowledge into a living memory – technically, a knowledge graph we call AIP – and build the AI that works on top of it, reviewing the results with you every few weeks.
Nothing we ship is a fixed-feature product. Each capability is built for how you work, on a platform that lets us release quickly and change course when the evidence says so. Most clients see the first working agent within the first cycle.
The people
Everything beyond the technical skills.
A weather engineer who retrained as a lawyer. A playwright who learned to talk to machines. Computer scientists who grew up across languages and countries. The broader the experience, the better the work – and ours could hardly be broader.
Research blog
We publish what we learn.
New · with John Zealley
What makes a better decision? The data you exclude matters most
AI without context is a deluge of data trying to pass itself off as information. Start from the decision you are trying to improve, then be disciplined about everything you leave out.
- What makes a better decision? The data you exclude matters mostPerspectives
- How to evaluate a bespoke AI agent providerPerspectives
- What is a company memory?Perspectives
- Generic solutions for specific problemsHow we build
- Knowledge graphs for communications teamsHow we build
- What do we mean by “AI agents”?Perspectives
- Whatever happened to iterative prompting?How we build
- What our own company memory looks like after a yearCase studies
- The What and the WhyHow we build
- Design-led AI: how we actually get things donePerspectives
- Everything you write should be a promptPerspectives
Contact
Bring us a problem worth thinking about.
A first conversation costs nothing and usually surfaces something useful – whether or not we end up working together.