01/04/2026
Weekly signal from the field: use AI to accelerate Theory of Change development, not to outsource the thinking.
More organisations are being asked to show not just what they did, but what changed, for whom, and how solid the evidence is.
That pressure quickly exposes a familiar weakness: a Theory of Change that’s too vague to stand up. And when the logic is fuzzy, AI won’t fix it. It just makes the ambiguity sound more convincing.
This week’s 12 Steps focus is Step 3: Developing a Theory of Change.
A strong ToC makes everything downstream easier (indicators, tools, analysis, reporting). A weak one creates a chain reaction.
A useful anchor here is the UK Evaluation Society AI good practice guidance, which emphasises that evaluation purpose should drive AI use, human control must remain central, risks must be managed, and outputs must be verified and transparent.
Here are the key rules we are using (aligned to those UK Evaluation Society principles) when applying AI to a ToC:
1. Purpose first, then AI: Be clear what decision the ToC must inform. Only use AI if it helps serve that purpose.
2. Human control at the critical points: AI can draft and organise, but evaluator judgement stays in charge of causal claims, assumptions, and what is credible.
3. Break it into components, then stress-test: Work step by step (outcomes, pathways, assumptions, risks, what must be true). This beats generating a full ToC in one prompt.
4. Ground it in evidence and context: Use real programme documents and known context (proposals, MEL plans, indicators, reports). Without grounding, AI will invent plausible but wrong logic.
5. Verification and validation are not optional: Treat outputs as draft material. Challenge assumptions, verify internally, then validate with stakeholders.
6. Document what AI did and what humans checked: Keep a lightweight record of inputs, outputs, and what was verified or changed.
The rule stays the same: IMM is the architecture. AI provides the acceleration. Human judgement provides the verification.
We’ll unpack practical workflows at the AI Impact Clarity Summit (free, live). 13-15 April).
Save your seat: https://lnkd.in/etJnJRpw
