Skip to content
Phase 4: Program Design

AI Prompt for a Theory of Change

By Angel Reyes, DrPH(c), PMP, PMI-ACP, MPH, MCHES · Updated · Free to copy and use
Quick Answer

This AI prompt for a theory of change works backward from the long-term change you name. It maps the preconditions that must be in place, where your program contributes, and the "if, then, because" reasoning at each step. It returns a short narrative, a pathway table and a list of the weakest links for your team to review. Where evidence would be needed, it says so and does not supply research from memory.

The prompt

Copy and paste
Act as a structured program design facilitator. Help me develop a theory of change using only the program facts and evidence I supply in this conversation. Do not invent needs, partners, results, statistics or research, and do not supply research findings from memory. Where support is missing, write "evidence needed" and say what kind of source would settle it.

Organization and program: [NAME OR GENERAL DESCRIPTION]
Population and setting: [WHO AND WHERE]
Problem and the barriers people face: [PROBLEM AND BARRIERS]
Long-term change we want to see: [LONG-TERM CHANGE FOR THIS POPULATION]
Timeframe: [PLANNING HORIZON]
What we do and deliver: [ACTIVITIES AND OUTPUTS]
Resources and partners: [RESOURCES AND PARTNERS]
Evidence and community input we have, with sources and limits: [EVIDENCE AND INPUT]
Constraints, risks and outside conditions: [CONSTRAINTS AND RISKS]
How to work: [INTERVIEW ME FIRST or USE THE FACTS ABOVE]

If I chose the interview, ask me focused questions one at a time about anything missing before you draft. If I cannot answer, leave the point unresolved instead of guessing.

Work backward from the long-term change:
1. Identify the preconditions that must be in place for it to happen.
2. Identify the medium-term and short-term outcomes that lead to those preconditions.
3. Connect our activities and outputs to the outcomes they could plausibly influence.
4. Separate what our program can contribute from what depends on others.

For every link in the pathway, write the reasoning as IF [this happens] THEN [this follows] BECAUSE [reason]. Label each "because" as a supplied fact, an assumption or evidence needed.

Return:
A. A narrative theory of change of 150 to 200 words.
B. A pathway table with columns: outcome or precondition | what leads to it | if-then-because | basis (fact, assumption or evidence needed) | who else it depends on | risks.
C. The three weakest links, each with a question to put to staff and community members.

Keep intended results separate from demonstrated results, and keep conflicting evidence visible.

Works in ChatGPT, Claude, Gemini and other AI chat tools. Replace each bracketed placeholder with your own details before you send it.

What to fill in

[LONG-TERM CHANGE FOR THIS POPULATION]
A changed condition for a defined group of people, not something your organization does. "Families in our county have stable housing" is a change. "Provide housing counseling" is an activity.
[PROBLEM AND BARRIERS]
What stands between the population and that change, as you understand it from your own experience and from what the community has told you.
[ACTIVITIES AND OUTPUTS]
What the program does and what it delivers. Keep this factual. The prompt works out how these connect to the long-term change.
[EVIDENCE AND INPUT]
What you know and how you know it: past results, an evaluation, community feedback or research you have read, with the limits of each. Use summaries, never individual records.
[INTERVIEW ME FIRST or USE THE FACTS ABOVE]
Choose the interview if your thinking is still loose. Choose the second option if you have already filled in every field.

Why this prompt works

Most programs can list what they do. Fewer can explain why those activities should lead to the change they want. A theory of change is that explanation, and the prompt is built to draw it out.

  • Backward mapping. Starting from the long-term change and asking what must happen first reveals preconditions that a list of activities would never surface, including ones your program does not address.
  • If, then, because. Delivering a service does not automatically produce participation, new behavior or lasting change. The "because" clause forces each link to be stated so a reviewer has something specific to challenge.
  • A label on every "because." Fact, assumption or evidence needed. Most first drafts turn out to rest mainly on assumptions, and it is better to find that out now.
  • Contribution versus dependence. Step 4 separates what you can influence from what requires other organizations or conditions. Funders notice when a program claims credit for change it cannot produce alone.
  • An interview option. If you choose it, the AI asks questions one at a time, which is useful when the theory exists in people's heads and not on paper.
  • No research from memory. AI tools can describe studies that do not exist. The prompt asks for the kind of source to look for, not for findings.

Why it is appropriate to use

Develop a theory of change during program design, before the work plan, evaluation approach and budget are fixed. If the pathway has a missing step, this is when you can add an activity, find a partner or scale back the claim.

AI is a reasonable aid here because the task is articulating and testing your own reasoning. The AI plays the colleague who keeps asking "and why would that follow?" It cannot tell you what your community needs or whether a link holds in practice.

Use it to prepare questions for your team. Do not use it to produce evidence, to set community priorities, to confirm that one thing causes another or to turn hoped-for outcomes into claimed results.

A theory of change and a logic model are companions. Once the pathway is agreed, the logic model prompt turns it into the table most funders ask for. For background on both, see the logic model and theory of change guide.

How to check and revise the output

  1. Trace the facts. Anything labeled as a supplied fact should be something you actually told it. Relabel the rest.
  2. Test every link. Look for jumps from an output to an outcome. Ask whether the activity can plausibly influence the outcome and what else must happen.
  3. Take the weakest links to people. Ask staff, partners and community members which links they doubt, whose experience is missing and what could prevent the change.
  4. Gather evidence where it matters most. For the links your case depends on, find real sources and read them yourself.
  5. Decide what to do about each weak link: revise the pathway, change the program, add a partner or narrow the long-term claim.
  6. Revise the narrative and table together so they tell the same story before either goes into a proposal.

Frequently asked questions

How does a theory of change differ from a logic model?
A theory of change explains why and under what conditions change is expected, including the assumptions behind each step. A logic model summarizes resources, activities, outputs and expected outcomes in a compact table. They should agree with each other, but they emphasize different things.
Can I use the prompt before I have evidence for every link?
Yes, for a planning draft. Leave uncertain links labeled and use them to decide what evidence to gather and what to ask your stakeholders. An assumption is not a finding.
What should I do with the weakest links?
Bring them to staff and community discussions. Decide whether to revise the pathway, change the program design or gather evidence. The prompt starts that discussion. It does not settle it.

More prompts for program design

Related prompts and guides

← Back to all AI prompts for grant writing

Ready to Master Grant Writing?

The Complete Grant Architect is a free 16-week course that transforms you from grant writer to strategic grant professional. Learn proposal engineering, federal compliance, budgeting, evaluation design, and AI-powered workflows.

Enroll Free in The Complete Grant Architect