Section 01 · the model

What the work is, and what counts as progress.

Progress is movement in two things only: the solution, and the confidence we have in it. Activity that moves neither feels like progress and is not. Everything in this guide serves those two.


What the work is

Two kinds of guide.

The other guides

How to do the work. The step-by-step practice for problem-led discovery and solution-led risk scans: playbooks, methods, templates.

This model

What the work is. The thinking the playbooks rest on: the quality of the solution, and the confidence we have in it. Nothing else is the goal.

In a Solve cycle

We start in one of two places.

Inside Solve, the practice depends on where the initiative begins. Reading this wrong is the most common mistake teams make.

Start with a problem

A pain point, gap, or opportunity

No solution yet. Do not jump to one. Work out the real problem worth solving, then develop and test concepts. Skip this and you build a strong solution to a problem nobody has.

Start with a solution

A tool, feature, or process change

It arrives with a hypothesis attached. Extract what must be true for it to deliver, then test the riskiest of those beliefs first. We do not run the solution to see what happens; we surface the beliefs it rests on.

Fig. 02
Start with a problem No solution yet First move: diagnose the current state Start with a solution Hypothesis attached First move: extract the beliefs it rests on Solve find the solution worth building A solution + hypothesis
Same destination, different doorway. Whether you start from a problem or a proposed solution, the goal is a solution worth building. What changes is the first move: diagnose, or extract the beliefs. Picking the wrong door is the most common mistake teams make.
Classify the starting point

Building to learn

Often, building is the test.

Developing confidence does not mean avoiding building. It means building the right thing, at the right fidelity, for where you are. There are two kinds.

Throw-away prototypes

Cheap, fast, disposable

Used early in Solve while searching for the right solution. They prove one thing, then get binned: for example, whether an AI can hit an outcome consistently before you commit to anything.

Working prototypes

Production-faithful, in a sandbox

Once committed to a real solution, build incrementally, as close to production patterns as possible, off production. Each increment proves an assumption and contributes a real component. We never build it twice.

The build cycle has a deliberate shape

01

Build the foundation

Stand up the production-faithful sandbox everything else gets built on.

02

Build by criticality

Build the feature for the most critical assumption, move it to Demonstrated, then take the next. Order follows priority, not a roadmap.

03

Stitch it together

Pieces accumulate instead of getting thrown away, leaving a coherent prototype ready to test at scale.

Fig. 03
Build by criticality · pieces accumulate · never built twice Foundation+ assumption 1 + assumption 2 Coherent prototype
One build, growing. Stand up the sandbox, then add one increment at a time, in order of criticality. Each block proves an assumption and stays in. Nothing is thrown away, so the prototype that ends the phase is the one you trial.

Setting the target

Confidence is not pursued to infinity.

Not every assumption needs Confirmed at scale. That would be slow, expensive, and pointless. Agree the target with the sponsor up front, on two levels:

1 · The level the decision requires

A small, low-risk change may need only Demonstrated. A large, costly, network-wide commitment needs Confirmed at scale.

2 · The level you target next cycle

Which assumptions move, and to what level, in the time you have. A concrete commitment for the cycle ahead.

This keeps the work honest both ways: no under-testing a big bet, no over-testing a small one.

Fig. 04
A small, low-risk bet A network-wide commitment AssumedIndicated Demonstrated Confirmed at scale
How far to climb depends on the bet. A small change is safe to ship at Demonstrated. A network-wide commitment needs Confirmed at scale. Agreeing the target up front stops both under-testing the big bet and over-testing the small one.

The decision rhythm

The recommendation happens at the end of a cycle.

We work in time-boxed cycles to validate or invalidate a set of critical assumptions. At each boundary, the team makes one of three calls.

Continue

Testing confirms the direction, but critical assumptions remain. Keep investing. Resolves into another cycle, or a trial once testing supports in-field validation at scale.

Pivot

Learnings show the solution needs rethinking. The problem may still stand; this solution, in its current form, does not.

Stop

Something is not desirable, feasible, or viable, and it cannot be solved or it kills viability. Off-ramp, with the learning banked.

At the end of a trial

Scale, do not scale, or send back to innovation. A trial is not designed solo: it needs the future owner, commercial finance, and operations governance, against the operations governance model.

Fig. 05
6-week cycle test critical assumptions boundary call Continue Pivot Stop keep investing rethink the solution off-ramp, learning banked
Three doors at every boundary. A cycle ends in one call. Continue loops into another cycle or a trial; Pivot keeps the problem but rethinks the solution; Stop off-ramps with the learning banked. At a trial's end the calls become Scale, Don't scale, or Send back.

The one thing to remember

Develop solutions worth backing, and an honest recommendation on whether to keep investing.

The five categories, the prioritisation, the prototypes, the confidence levels, the phases: each one exists to serve that single outcome.