Approach
Two spaces, three diamonds, one governed path to production.
Our method is a design process, not a delivery template. Three diamonds, each one a deliberate widening followed by a deliberate narrowing, sitting on a governance and security foundation. What comes out reflects your systems, your risk and your constraints.
The method
Most AI work fails because someone converged too early. A diamond widens the field first, then commits. Doing that three times, on the problem, the solution and the proof, is what keeps a decision defensible at each step.
Governance & Security
Designed in from Discover, not retrofitted after launch.
Access scope
Approval path
Audit trail
Problem space, Diverge
Discover
We widen the field before narrowing it. Current AI maturity, data and engineering readiness, the initiatives already running, the regulatory exposure attached to them, and where value actually sits in your operating model. Nothing is ruled in or out at this stage.
Problem space
Diverge
Discover
We widen the field before narrowing it. Current AI maturity, data and engineering readiness, the initiatives already running, the regulatory exposure attached to them, and where value actually sits in your operating model. Nothing is ruled in or out at this stage.
Converge
Define
We narrow to one evidenced problem statement and a ranked set of use cases. Leadership, risk and delivery all recognise the same starting position, which is what stops the work being relitigated three months later.
Solution space
Diverge
Design
We generate real options rather than defending the first one. Target architectures, integration paths, workflow redesign, the governance model, and the human checkpoints. This is also where decision rights get set: which decisions the system makes, which need human judgement, and who stays accountable for each.
Converge
Develop
We commit to one option and build it narrow. Evaluation criteria, risk tier and a named owner attach to the build before any code does, so there is something to measure against later.
Proof space
Diverge
Demo
We run the working proof against your real data, in front of the people who will live with it. Every stakeholder gets to try to break it. That is the point of the phase.
Converge
Evaluate
We measure against the criteria set in Define, not against the impression the demo made. Go, adjust, or stop. Stopping early is a result, and a cheap one.
Governance & Security
Designed in from Discover, not retrofitted after launch.
Access scope
Approval path
Audit trail
Why a third diamond
The classic double diamond ends when something is built. In AI work that is exactly the wrong place to stop, because a system that performs well in development can still fail on real data, real workflows and real oversight. The third diamond exists to prove that before anyone commits to scale.
Why you can trust the outcome
Each phase hands the next one evidence from your business rather than opinion. You always know what is being built, why, and what it is worth before it goes live.
After the third diamond
What happens once the proof holds.
The diamonds end at a decision. Two things carry on past it, and neither is a design exercise.
Deploy
We build and pilot in a narrow, instrumented slice of a real workflow, inside the systems your teams already use.
Value is measured against the criteria set in Design, not inferred from a demo. Guardrails and human checkpoints ship with the system.
Where the results hold, we scale the pattern. Where they do not, we stop early and say so. Your teams are trained to run what we build.
Govern
Once systems are live, oversight is standing work, not a project phase.
We maintain policy, the model inventory, risk tiering, approval paths and accountability, so the control environment tracks the portfolio as it changes. New use cases enter through the same gates.
Audit and regulatory questions are answered from existing evidence instead of reconstructed after the fact. That is what lets you scale with control.
Methodology foundations
What the work is grounded in.
Human-Centered Design
Change is designed with the people expected to run it, not delivered to them. Co-creation is what makes adoption stick after the engagement ends.
Systems Thinking
No AI initiative sits in isolation. Every recommendation accounts for the wider operating model it has to fit into, not just the use case in front of it.
Governance and Security by Design
Oversight, data protection and access control are treated as design inputs from the first diamond, not as a compliance review before launch. That is what lets you scale from a single pilot to a full portfolio without losing control of it.
That combination is what turns a workshop into something your business actually runs on.
Begin with Discover.
The AI enterprise readiness assessment is the short version of the first diamond. Eight questions, one for each pillar, and an initial banding.