A framework, given away free

The Honest Programme

A programme fails when it can no longer tell the truth about its own state. Long before the budget runs out, it loses the ability to see itself clearly. This is a free, research-grounded way to check whether yours still can.

The underlying discipline is what I call delivery epistemics: how a programme knows what it knows, and where that knowledge quietly breaks down. It draws on five research traditions, each pointing at a different way a programme stops seeing itself. The diagnostic below scores all five in under ten minutes. No email, no wall, nothing stored.

Five ways a programme stops telling the truth about itself.

Shared Reality01

Weick, on sensemaking

Do the delivery team, the client, and the sponsor share one honest model of what is happening? Crises are preceded by a breakdown in shared sensemaking. When everyone holds a different picture, coherent action becomes impossible, and energy goes into pulling in different directions.

Structural Foundations02

Schwaber & Sutherland, CHAOS

Does the programme have the minimum instrumentation to inspect and adapt? A committed backlog, a shared Definition of Done, an honest reporting cadence. Their absence is not a style preference. It is a structural failure mode, and it is the most common one.

Signal Integrity03

Flyvbjerg, on optimism bias

Are the programme's signals honest, or smoothed upward on the way to the sponsor? Estimates drift optimistic by default, and status gets polished to protect support. A programme that cannot see its own true position cannot correct course.

Safety to Surface04

Edmondson, on psychological safety

Can bad news travel upward without career risk? High-performing teams report more problems, not fewer, because they can. The absence of reported problems is rarely evidence of health. It is usually evidence that problems cannot be raised.

AI Readiness05

McKinsey, Deloitte, NIST

On AI programmes, the same failures apply, plus a few of their own: unclear data ownership, no governance for model outputs, no audit trail for AI-assisted decisions, and adoption that was assumed rather than designed. The model is rarely the thing that fails.

Score each category honestly and the shape of the failure shows up before the budget does. The red zones are where to start.

Score your own programme.

Twenty questions, four per category. You get a five-axis profile, a health band, and ranked, specific next steps for your weakest areas. Under ten minutes. No email, nothing stored.

Take the diagnostic→

Free templates, the artefacts the diagnostic points to.

The instruments a programme needs to see itself, ready to use and adapt. No sign-up.

RAID logExcel ↓

A tab per log — risks, assumptions, issues, dependencies, actions — with owners, dropdowns, a probability×impact risk score, a live dashboard, a 5×5 risk matrix, and dates that turn red when they slip. Worked examples throughout.

Right-sized governance checklistExcel ↓

Governance that scales with the work. Filter the 'Applies from' column to your engagement size so you only see what applies. A 'Done?' dropdown and a filled example sheet.

Decision logExcel ↓

A written record of what was decided, why, and — the column that matters most — what you gave up. Every honest decision closes a door; this names it.

RACI matrixExcel ↓

Who is Responsible, Accountable, Consulted and Informed for each activity, with a built-in check that enforces exactly one accountable owner per row.

Stakeholder register + mapExcel ↓

Register every stakeholder with influence, interest, current and target stance — and a power/interest grid that sorts them into the four engagement strategies automatically.

Lessons learned logExcel ↓

Capture lessons as they happen, not in a rushed closure session: what you expected, what happened, why, and the one thing you'll do differently. A lesson isn't learned until it's embedded.

Change log + request formExcel ↓

Every change to scope, time or cost — logged, assessed three ways, and explicitly decided with a name against it. Plus a request form to work up a single change in full.

Definition of DoneWord ↓

A checklist to agree with the client, not just the team, with worked examples for agile, hybrid and waterfall delivery. The one artefact that ends 'I thought it was finished' disputes.

Team charterWord ↓

How the team works before it is under pressure: roles, cadence, decision rights, communication do's and don'ts. A filled worked example follows the blank.

One-page statusWord ↓

An honest weekly status that shows real delivery and names the gap, instead of a reassuring RAG colour — including the line for what you're not yet saying out loud.

One-page business caseWord ↓

Enough to make an honest decision, on one page: the problem, the options, the recommendation, the cost, the measurable benefit — and what you'll stop doing to make room.

Post-implementation reviewWord ↓

The honest look back almost nobody ships: benefits promised versus realised, what worked, what didn't, and what you'd do differently. With a worked example.

The research behind it

Flyvbjerg, B. (2021). Top Ten Behavioral Biases in Project Management. Project Management Journal, 52(6).
Edmondson, A.C. (1999). Psychological Safety and Learning Behavior in Work Teams. Administrative Science Quarterly, 44(2).
Weick, K.E. (1988). Enacted Sensemaking in Crisis Situations. Journal of Management Studies, 25(4).
Schwaber, K. & Sutherland, J. (2020). The Scrum Guide.
Standish Group (2024). CHAOS Report.
McKinsey & Company (2024). The State of AI.
Deloitte (2026). State of AI in the Enterprise.
NIST (2023). AI Risk Management Framework (AI RMF 1.0).