Lesson 11 · The Grant Architect

11. AI Spotlight

30 min

By the end you'll be able to

  • Distinguish generative AI tools from analytical AI tools in grant work.
  • Identify the failure modes that disqualify AI-assisted submissions.
  • Apply the Human-in-the-Loop principle to a realistic drafting workflow.
  • Anticipate emerging funder disclosure requirements and prepare your practice.

AI is changing grant work, but not in the way the hype suggests. It is not writing winning proposals for you. It is compressing the time you spend on research, first drafts, formatting compliance, and prospect triage so you can spend more time on the strategic work that actually wins awards.

In this lesson we introduce the two categories of AI tools in the field: generative (large language models that draft and revise text) and analytical (tools that score prospects, summarize 990s, extract NOFO requirements, and flag eligibility issues). You will see realistic use cases for each, along with the failure modes that produce hallucinated funder names, fabricated statistics, and inappropriate boilerplate.

The non-negotiable principle is Human-in-the-Loop: every AI output must be reviewed, verified, and owned by a qualified human before it touches a proposal. We also preview the emerging funder disclosure policies (NIH, NSF, several large foundations) that will require you to declare AI use, and how to position your practice for that shift now rather than later.

Common mistakes

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

  • Letting AI write the funder-specific sections.

    Models cannot reliably reproduce a funder's stated priorities, RFP language, or recent grant history. Those sections need human research and human voice.

  • Treating disclosure as optional.

    NIH, NSF, and an increasing number of foundations are formalizing AI disclosure expectations. Building disclosure into your workflow now avoids retrofit pain later.

Practice problems

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

  1. Problem 1
    Audit one paragraph of AI-generated text for hallucinated facts. List every claim and mark each as verified, unverifiable, or fabricated.
    Show solution

    Sample paragraph yields five claims. (1) National prevalence statistic: verified against CDC source, keep. (2) Local prevalence statistic: unverifiable, replace with cited state health department figure or delete. (3) Named research study: fabricated, delete. (4) Quote attributed to a federal official: unverifiable, delete. (5) Programmatic outcome from a prior grant: verified against internal report, keep. Final paragraph keeps two of the five original claims, replaces one, and deletes two.

Practice quiz

  1. Question 1
    What does Human-in-the-Loop require in a grant context?
  2. Question 2
    Which is the most common failure mode of generative AI in grant proposals?
  3. Reflection 3
    Outline a Human-in-the-Loop drafting workflow for a needs statement, in three steps.

Lesson 11 recap

AI tools compress effort on research, drafting, and triage, but Human-in-the-Loop is non-negotiable. Hallucinated facts and missed disclosures are the failure modes to design against.

Coming next: Lesson 12 — Introduction to Prospect Research

Week 2 moves from foundational landscape into strategic prospect research, where you will build the prospect lists that drive the rest of the course.

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