159. AI Assisted Prospect Research
By the end you'll be able to
- Sequence analytical discovery, primary source collection, and generative synthesis correctly.
- Build a structured funder profile from documents you provided to the model.
- Verify every funder profile against the IRS Tax Exempt Organization Search and the funder's website.
- Identify the points where AI shortcuts produce hallucinated funder data.
Prospect research is the workflow where AI offers the largest time savings and the largest hallucination risk in the same task. In this lesson you build a defensible AI-assisted research process that uses analytical tools for discovery, generative tools for synthesis, and human judgment for the final go or no-go decision on each prospect.
You will work through a layered approach. Start with analytical platforms (Instrumentl, Candid, Foundation Directory) to surface candidate funders against your organization's profile. Pull the underlying primary sources for each candidate: the most recent 990, the foundation's annual report, recent funding announcements, and any press coverage of leadership changes. Then use a generative model to synthesize those documents into a structured profile covering median grant size, geographic focus, funding priorities, application norms, and any signals of strategic shift. The model never invents a funder. It only summarizes documents you provided.
The verification step is non-negotiable. Before any AI-generated profile drives a go decision, you confirm the foundation's legal name and EIN against the IRS Tax Exempt Organization Search, confirm grant ranges against the 990 itself, and confirm contact details on the foundation's own website. By the end you can describe an AI-assisted research workflow that is faster than manual research and more accurate than naive AI use.
Common mistakes
These are the traps learners hit most often on this topic. Knowing them in advance is half the fix.
Asking a generative model to "find foundations."
The model will invent names, EINs, and grant amounts. Discovery is an analytical task.
Skipping the IRS verification step.
A misidentified foundation wastes a cultivation cycle and damages credibility with the actual funder if the error reaches them.
Practice problems
Try each on paper first. Click Show solution only after you've made a real attempt.
- Problem 1Outline a three-step AI-assisted research workflow to evaluate a candidate foundation for an early childhood program.
Show solution
First, query an analytical platform such as Instrumentl or Foundation Directory for foundations funding early childhood in your geography under your target grant range. Second, pull the most recent 990, annual report, and funding announcements for each candidate and instruct a generative model to produce a structured profile covering median grant size, geographic focus, priorities, and application norms. Third, confirm legal name and EIN on the IRS Tax Exempt Organization Search and confirm contact details on the foundation's own website before any prospect enters the active pipeline.
Practice quiz
- Question 1Which step belongs to the analytical AI layer of an AI assisted prospect research workflow?
- Question 2What is the role of the generative model in a defensible prospect research workflow?
- Reflection 3In one or two sentences, explain why you confirm a foundation's legal name and EIN against the IRS Tax Exempt Organization Search before any go decision.
Lesson 159 recap
AI-assisted research is a layered workflow (analytical for discovery, generative for synthesis, human for verification). Each layer has a job, and each layer has a failure mode.
Coming next: Lesson 160 — Accelerating Program Design
Next, we apply AI to program design, where the model serves as a structured brainstorming partner rather than a designer.
Saved in your browser only — no account, no server.