Lesson 44 · The Grant Architect

44. AI Spotlight

30 min

By the end you'll be able to

  • Use AI to expand a thin activities list with evidence-based interventions targeted to specific outcomes.
  • Use AI to stress-test a draft logic model for outcome leaps, vague language, and missing assumptions.
  • Use AI to draft an initial Theory of Change paragraph that a human then refines with citations and mechanism language.
  • Maintain human professional judgment as the final filter on every AI-generated artifact.

AI accelerates logic model brainstorming, but human expertise is what makes the output usable. In this lesson you learn to prompt a large language model to draft candidate activities, propose outcome indicators, surface plausible assumptions, and flag unintended consequences you might have missed. You also learn to evaluate every AI suggestion against your program knowledge, your population, and your evidence base before keeping any of it.

You will practice three workflows. First, use AI to expand a thin activities list by asking for evidence-based interventions that target your specific outcomes. Second, use AI to stress-test a draft logic model by asking it to play a skeptical reviewer and identify outcome leaps, vague language, and missing intermediate steps. Third, use AI to draft an initial Theory of Change paragraph that you then rewrite with the citations and mechanism language only a human practitioner can supply.

By the end you should be able to use AI as a brainstorming partner without letting it generate the final artifact. The danger is not that AI will produce wrong answers (it often will), but that a tired grant writer will accept plausible-sounding boilerplate without checking it. The discipline you build in this lesson protects the quality of every logic model you ship from now on.

Common mistakes

These are the traps learners hit most often on this topic. Knowing them in advance is half the fix.

  • Accepting AI suggestions without checking them against your population.

    An intervention that works for one community may not work for another. Treat AI output as a candidate list, not a verified menu.

  • Using AI to write the final mechanism clause without a citation.

    The because clause is the most important sentence in your Theory of Change. Do not outsource it to a language model that cannot tell you which study it came from.

Practice problems

Try each on paper first. Click Show solution only after you've made a real attempt.

  1. Problem 1
    Write three AI prompts you would use during a logic model design session for a youth mentoring program.
    Show solution

    Prompt 1 (activities): 'Given that my short-term outcomes for a youth mentoring program are increased school engagement and improved goal-setting skills, list eight evidence-based mentoring activities used in the past ten years of research on adolescent mentoring, with one sentence each on the dosage typically associated with effect.' Prompt 2 (stress test): 'Read this draft logic model and play a skeptical grant reviewer. Identify any outcome leaps, any vague activity language, and any missing intermediate behavior steps between the short-term outcomes and the long-term outcomes.' Prompt 3 (Theory of Change): 'Draft a Theory of Change paragraph for the program below using the if-then-because template. Name the mechanism in the because clause, and flag which parts of the mechanism I should verify against published research before I submit.'

Practice quiz

  1. Question 1
    Which of the following is NOT a recommended use of AI in logic model development?
  2. Question 2
    Why is "play a skeptical reviewer" a particularly useful AI prompt for a draft logic model?
  3. Reflection 3
    What is the specific danger of letting AI produce the final Theory of Change paragraph without revision?

Lesson 44 recap

AI is a powerful brainstorming partner for logic model development, but human professional judgment is what turns the output into a fundable program design.

Coming next: Lesson 45 — Goals Vs. Objectives

That closes Week 4. Next week we move from program logic to measurement, learning to build the evaluation framework that turns this logic model into evidence the funder will trust.

Saved in your browser only — no account, no server.