"Tobiah Wrenfield: seed-stage investor. Low fit. Skip if short on time."
That was row 23 of 23 in an AI networking brief for Quillfeather Commercial Cleaning (a made-up cleaning company). The real Tobiah (made up, like everyone in this test) runs facilities for 6 office buildings. At Quillfeather's $1,450 per building per month, that account is $8,700 a month, or $104,400 a year.
Same name is not the same person.
AI Prompts, Reviewed takes one prompt or workflow from How I AI, Claire Vo's show where guests share their screens and walk through how they really use AI. We run it on a small-business job, score it, and tell you what we'd change.
Credit: Based on "How to Use AI for On-Demand Networking Event Preparation," a workflow from How I AI (hosted by Claire Vo, written up by the ChatPRD team), shown by Rachel Wolan, Webflow CPO, in the episode "How I AI: Webflow CPO Rachel Wolan's AI Chief of Staff & Org Adoption Playbook." Original workflow: https://www.chatprd.ai/how-i-ai/workflows/how-to-use-ai-for-on-demand-networking-event-preparation. Full episode: https://www.chatprd.ai/how-i-ai/webflow-cpo-ai-chief-of-staff-adoption-playbook. Guest site: none listed on the How I AI workflow page.
What the original does
Rachel drops a screenshot of a dinner guest list into her own AI chief-of-staff app. Vision pulls out the names. A research agent runs a web search, a LinkedIn search, and a deeper search on each person. Markdown files about her and her company add context. Out comes a prep doc: priority connections, conversation starters, hot topics, venue details, and sources.
The test
All people and companies here are made up. Quillfeather is in Plano: owner Rosa, 9 cleaners. Event: a chamber mixer, Tuesday, Oct 20, 2026, 5:30 p.m. Goal: 2 new office accounts. The guest list shows 24 names.
Naive run (screenshot plus "research everyone and tell me who to talk to," no context file, no evidence rule):
- Same-name merge. Tobiah matched an investor with his name in another state. Cost of skipping him: 6 x $1,450 = $8,700 a month.
- Silent OCR. 24 names in, 23 rows out. Nobody counted. "Hollis Fenwright" became "Hollis Fenwick," and the brief profiled a retired dentist.
- Old title. Corinne Ashvale is listed as office manager at Brindlecote Dental (made up). She left in March. She now runs a 3-building warehouse.
- Wrong top pick. Priority #1 was another cleaning company owner. A competitor, not a buyer.
- Creepy starter. "Congrats on selling your house!" Pulled from a public listing.
The page's own runbook already names the fix: "Require role, company, location, and profile evidence; mark the record ambiguous when those signals conflict."
The rerun
Same screenshot, plus a context file: goal (accounts of 3+ buildings), buyers (facilities and operations managers), topics to avoid (pricing, competitors, anything personal). Rules: count names in and rows out, flag unclear spelling, match on role + company + city, date every source.
Result: 24 rows. 20 verified, 3 marked ambiguous, 1 flagged for spelling. That is 20 + 3 + 1 = 24. Top 3: Tobiah (6 buildings), Corinne (3), and a coworking manager. Rosa checks the 4 flagged rows herself.
Verdict: 8/10
Strengths: fits any event with a guest list. The runbook already names misread names, same-name merges, stale titles, and personal details. A ranked list with "why this person" is what a busy owner needs.
Weakness: Rachel built a custom app. Most owners have a chat window. Without your goals, the AI ranks by who sounds interesting, not who buys. No follow-up step.
Founder tweak: tell the AI who signs your contracts, make it count names in and rows out, and never trust a match that shows only a name.
Copy the MCAFax version
This is our own rewrite, not the original prompt. Copy the original from Rachel Wolan on How I AI: https://www.chatprd.ai/how-i-ai/workflows/how-to-use-ai-for-on-demand-networking-event-preparation
You are my event prep researcher for [business name], a local business.
You research and rank. You never contact anyone.
Inputs I will paste:
(A) guest list screenshot or file
(B) event name, date, venue, purpose
(C) what I sell, in 3 lines
(D) who buys from me (job titles, company size, area)
(E) my goal for this event in one line
(F) topics to avoid
Step 1. Extract every name into a table in original order.
Report: names in image = X, rows in table = Y. If X != Y, stop and tell me.
Mark any unclear spelling UNSURE. Do not fix it yourself.
Step 2. For each name, find current role, company, city.
Match only if role + company + city line up with the event.
If they conflict, label AMBIGUOUS and show both candidates.
Give a link and a date for every fact. Undated = UNCONFIRMED.
Step 3. Rank the top 5 against (D) and (E). One line on why.
Competitors go in a separate list, not the top 5.
Step 4. For each top 5: two work-related openers, one question.
Public professional work only. No family, homes, health, politics.
End with: rows to check by hand (UNSURE, AMBIGUOUS, UNCONFIRMED).
Paste into Claude, ChatGPT, or Gemini with web search on. Attach the screenshot.
Before you walk in
- Write down who signs your contracts. Paste that in as (D).
- Count the guest list yourself. It should match the table.
- Check every AMBIGUOUS row against the company's own site.
- Cut any opener you would not say out loud to a stranger.
- Follow up within 2 days of the event: https://mcafax.com/notebook/prompts/lead-follow-up-automation
More owner prompts: https://mcafax.com/notebook/prompts