PAL-EBM AI Prompt Library
Evidence Over Opinion
Most AI prompt libraries help you produce things faster. This one is built to do the opposite: it helps you interrogate what you already have. That distinction matters, because Evidence-Based Management is not a content problem. It is a judgement problem.
Every prompt here works the same way. You bring real evidence — a goal, a dashboard, a set of measures, an observed behaviour — and the prompt puts pressure on it. None of them will invent a strategy for you, generate an objective you did not ask for, or hand you a set of KPIs to adopt. That refusal is deliberate, and it is the point.
The prompts work at whatever altitude you operate at. A Product Owner can run them against one product and its Sprint Goals; a leader can run them against a portfolio. Where a prompt could go either way, it asks you which scope you mean rather than assuming you have an executive dashboard.
The Golden Rule
AI can accelerate how you gather and interpret evidence. It cannot decide what counts as value.
That decision stays with the accountable human. Every prompt below is written to keep it there.
How to use this library
The ten prompts form a single loop, not a menu. Work through them on one real problem rather than dipping in at random — the output of each one becomes the input to the next, and the last stage sends you back to the first.
- Check the input contract first. Each card opens with a yellow band telling you exactly what to paste in. A prompt without real evidence pasted into it will politely produce fiction.
- Read the evidence labels. Every prompt ends with the same instruction: mark each claim VERIFIED, INFERRED, or UNAVAILABLE. Anything marked INFERRED is the model reasoning, not your organisation's reality. Never let an INFERRED claim into a goal-setting conversation without checking it.
- Works in any capable assistant. Nothing here is tied to a specific tool. Paste into ChatGPT, Claude, Gemini, or whatever your organisation has approved.
Prefer to watch? I walk through the thinking behind these prompts and run three of them live.
Stage 1 — Diagnose Your Context
Start here if you are not yet sure whether the problem in front of you is a goals problem, a trust problem, or both. The same goal behaves very differently depending on the environment it lands in. This works for a single team as well as for a department.
1. Goals & Trust Diagnostic
Describe what people actually do — not how you rate yourselves — and locate your team or organisation on the goal-clarity and trust axes.
Stage 2 — Set the Goals
Complex problems are better solved with outcome-based goals. Activity and output goals hide the real purpose of the work, from the customer and therefore from the organisation. These three prompts build the goal ladder, connect it to the goals your teams already hold, and check what kind of goals you have actually written.
2. Backward Goal Ladder
Derive Intermediate and Immediate Tactical Goals by working backwards from your Strategic Goal — never forwards from the work you already have queued.
3. Goal Onion Mapper
Connect the leadership goal ladder to the Product Goal and Sprint Goal your teams actually hold — and find the layer that is missing.
4. Five-Category Goal Sorter
Sort every objective you have into Input, Activity, Output, Outcome, or Impact — and find out how few of them are actually outcomes.
Stage 3 — Choose the Measures
Measures do not simply report reality. They change what people believe the goal is, and therefore what they do. These four prompts check whether you are looking at a balanced picture, help you pick measures that answer a real question, and predict how the ones you choose will backfire.
5. Key Value Area Balance Scan
Find out which of the four Key Value Areas your goals actually reflect, which you are blind to, and what that blindness exposes you to.
6. Key Value Measure Selector
Pick measures that answer your chosen Key Value Area's questions, and size the Satisfaction Gap between what customers want and what they get today.
7. Cobra Effect Predictor
Quick single-measure check. Before you deploy a measure, find out how people will game it and who will pay for that.
8. Goal to Measure to Behaviour Tracer
Full audit of one goal. Derive the measures the goal deserves, compare them to the ones you have, and trace the behaviours back to see whether they still serve the goal.
Stage 4 — Experiment, Inspect and Adapt
A goal you cannot test is an opinion with a deadline. These two prompts close the loop: the first turns a tactical goal into a falsifiable experiment with the threshold written down before it runs, and the second handles what happens when the result tells you a goal further up the ladder was wrong.
9. Experiment Canvas Builder
Turn a tactical goal into a falsifiable hypothesis with a pre-registered threshold, a decision rule, and an honest place to record an inconclusive result.
10. Goal Invalidation Navigator
The experiment came back and a goal further up the ladder is no longer reachable. Get the real option set instead of quietly reframing the failure.
What This Library Deliberately Leaves Out
You will not find a prompt here that writes your OKRs, generates a set of KPIs for your department, or produces a strategy from a one-line brief. Those prompts would rank well and they would undermine everything the rest of this page is for.
The reason is simple. A measure is not a neutral instrument. It changes what people believe the goal is, and therefore what they do — which means choosing one is an act of leadership, not an act of drafting. Handing that choice to a model that has never met your customers produces something fluent, plausible, and quietly wrong. By the time the behaviour it drives becomes visible, it has been driving that behaviour for two quarters.
So the prompts above will help you interrogate a measure, predict how it will be gamed, and check whether it still serves the goal it came from. They will not pick it for you.
Where to Go Next
These prompts assume you already know the Evidence-Based Management framework well enough to supply real inputs. If the Key Value Areas, the goal ladder, or the experiment loop are new to you, the framework itself is the place to start — the prompts will be far more useful afterwards.