Lesson 160 · The Grant Architect

160. Accelerating Program Design

30 min

By the end you'll be able to

  • Use AI to surface implementation barriers before they appear in a logic model.
  • Generate alternative program models with different cost and risk profiles.
  • Map unintended consequences and inequities before they become real harms.
  • Treat every AI-surfaced idea as a hypothesis that requires partner and evidence review.

Program design is where most proposals quietly fail. The activities do not connect to the outcomes, the assumptions are unstated, and the implementation barriers were never surfaced. In this lesson you use generative AI as a structured brainstorming partner that stress-tests your program concept before it ever reaches a logic model, not as a replacement for the program officer, the practitioner, or the community voice that should drive design.

You will practice four prompting moves. The first is barrier surfacing: "Act as an experienced implementation scientist. Identify the ten most likely barriers to executing this program as described." The second is evidence scan: "List evidence-based interventions that have been associated with the outcomes I am targeting, with the caveat that I will verify each citation." The third is unintended consequence mapping: "Identify the ways this intervention could produce harm or inequity if implemented poorly." The fourth is alternative generation: "Propose three alternative program models that could achieve the same outcomes with different cost and risk profiles."

The discipline is treating every AI suggestion as a hypothesis, not a recommendation. You verify the evidence, you check the barrier against your operating context, and you let your community partners weigh in before any AI-surfaced idea enters the design. AI accelerates the divergent thinking phase. You and your partners still run convergence.

Common mistakes

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

  • Accepting AI-suggested interventions without verification.

    The model can name interventions and cite plausible studies. Every citation must resolve to a real source before it enters the design.

  • Skipping the partner voice.

    AI brainstorming does not replace community input. Design decisions still belong to the people the program will serve.

Practice problems

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

  1. Problem 1
    Write a single prompt that asks a model to stress-test a proposed after-school tutoring program for unintended consequences.
    Show solution

    Act as an implementation scientist with experience in K-12 community programs. The program is an after-school tutoring initiative serving 200 middle schoolers in a Title I district, four days per week, with paid tutors and a parent engagement component. Identify the five most likely unintended consequences or inequities this program could produce, and for each, suggest one design adjustment that would mitigate the risk.

Practice quiz

  1. Question 1
    Which prompt move is most useful in the divergent thinking phase of program design?
  2. Question 2
    What is the correct status of an AI-suggested evidence-based intervention?
  3. Reflection 3
    In one or two sentences, explain why unintended consequence mapping is worth a dedicated prompt during design.

Lesson 160 recap

AI accelerates divergent thinking in program design through barrier surfacing, evidence scans, unintended consequence mapping, and alternative generation. Convergent decisions still belong to humans and partners.

Coming next: Lesson 161 — Drafting Narratives - Iterative Process

Next, we apply the iterative drafting workflow that keeps the human voice in command of every narrative section.

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