Intelligence is your competitive advantage - amplify it!
Better decisions faster is the relentless requirement.
How will AI impact my business and how should I adapt?
Where AI rollouts usually fail
People deploying LLMs are running into problems Ashbyan Cybernetics described decades ago: checking work they cannot fully see, goals that drift between the board's intent and what the system does, work that nobody independently checks, and organisations that cannot absorb the change. The vendors cannot tell you why their systems work, so the firm has to know what to check, and who checks it.
What we do
Evaluation Protocols, built into the firm as it adopts agents: its people deciding what the agents may draw on, who signs what they produce, and how a decision made by humans and agents together is verified.
Diagnose
A written diagnostic of one workflow or team: where your people sit on the evaluation loop, what the stakes require, and what is worth doing. Usually the first step.
Build
Working code on a real workflow, with an evaluation suite that tells you when the model, the prompt or the data drifts, before your clients do. Owned by your team.
Embed
Senior attention alongside leadership through a longer rollout, until your team can run the system without us.
Try the evaluation loop
Check too little and polished output sails through; check everything by hand and the people checking wear out and make more mistakes. Our interactive instrument shows where the sustainable zone sits for your people and what your work demands.
Our bet on AI agents
On 6 October 2026 we made a dated prediction in two parts: that the firms deploying AI agents will name and solve the problems Ashbyan Cybernetics described, most likely under new names, and that the work of Ashby, Beer and Pask will be cited more often as agents spread. We keep the evidence as it arrives, including the evidence against us.
The lineage
Three British Cyberneticians, and our own paper.
W. Ross Ashby
An Introduction to Cybernetics, 1956. A regulator can only match what it can see, and an intelligence amplifier selects against a goal held outside it.
Stafford Beer
Brain of the Firm, 1972. Five systems any viable firm must have, and the same five again inside every part.
Gordon Pask
Conversation Theory, 1975. Understanding is an agreement between two actors, each able to teach the concept back.
O'Rourke, Kybernetes, 2026
The large language model read as a variety transducer, framed by Ashby's intelligence-amplifier and Beer's viable system model.
Recent writing
Notes from the work, on the kind of AI questions that come up in partner-led practices.
Workslop and brain fry are one curve. An interactive instrument for reading where your team sits on it — and what a Diagnose engagement measures.
Considering AI for your practice?
Drop a short note about the work you're thinking about. What you've tried, what's stuck, and what "shipped" would look like.