Make a 10-minute account research brief from a company's own website text

What this does

You paste text from a company’s own pages and a couple of recent news items. You get a one-page brief: what they do, who they sell to, what might matter to your contact, questions to ask, and a clear list of what the pasted text doesn’t tell you.

Who it’s for

SDRs and AEs about to reach out or take a first call, and owners researching a larger customer. Not for deep competitive analysis or anything that needs live data (headcount, funding, tech stack) that isn’t in what you paste.

What you need

• Text copied from the company’s About, Products, and Careers or News pages
• 1-3 recent news items or press releases (pasted text)
• Your product in one sentence
• The person’s job title (not name, if you want to keep it private while learning)
• Tool: Any AI chat tool (for example ChatGPT, Claude, Gemini, or Grok), Web browser
• Cost: $0 with a free AI chat tier (limits change); a web browser to copy text
• Setup: 2 minutes (estimate)

Steps

1. Open the company’s About, Products/Services, and Careers or News pages. Copy the main text of each (skip menus and footers). Paste into one document, with a heading before each chunk.
2. Copy 1-3 recent news items or press releases the same way. Note the date of each.
3. Open your AI chat tool. Paste the prompt below, then fill the [VARIABLES].
4. Paste your copied text where the prompt says [PASTED TEXT].
5. If your pasted text is very short (under about 100 words), the AI will stop after sections 1, 2, and 7. Paste more text and run it again for the rest.
6. Read the brief. For each claim, find the sentence in your pasted text it came from. Delete any claim you can’t match.
7. Copy the ‘Unknown / verify’ list into your notes. Those are your first questions or a quick manual check.

The prompt

You are a research assistant for a salesperson. Use ONLY the text I paste below. Do not use outside knowledge about this company, and do not guess. If something isn't in the text, write "Unknown."

NO-ADDITIONS RULE: Use only facts that are in my pasted text or my variables. Do not add facts, tips, statistics, scarcity or urgency, promises, references or testimonials, or the gender of any person that I did not supply.

My role: [YOUR ROLE, e.g. SDR, AE, owner]
What I sell, in one sentence: [YOUR PRODUCT/SERVICE]
Contact's job title: [TITLE]
My goal for this brief: [e.g. write a first email / prepare for a 30-minute call]

PASTED TEXT (each chunk starts with a heading and, for news, a date):
[PASTED TEXT]

If the pasted text is under about 100 words, say so at the top and write ONLY sections 1, 2, and 7. Do not write sections 3 to 6 from thin text.

Otherwise write a one-page brief with these sections:
1. What the company does (2 sentences, from the text)
2. Who they sell to (from the text; "Unknown" if not stated)
3. Recent changes or news (bullets, each with the date from the text)
4. Three things this contact's role likely cares about, labeled "Hypothesis, not fact," each tied to a quote or line from the text
5. Five questions I could ask
6. Possible reasons to reach out now (only if the text supports them; otherwise write "None found")
7. Unknown / verify list: things I would need to confirm before relying on this brief

Rules: no flattery, no invented numbers, no invented names. Under 350 words.

Sample input (fake data)

All names and companies are fictional. This is the input for run 4 (the output below), on the current prompt. Runs 5 and 6 used different inputs; see the test log.

My role: AE
What I sell: dispatch and route-planning software for small plumbing and HVAC companies
Contact's job title: Operations Manager
Goal: prepare for a 30-minute intro call

PASTED TEXT:
## About (fictional site)
Cobalt Ridge Mechanical is a 38-person plumbing and heating company serving the north suburbs. We handle residential repairs, new boiler installs, and light commercial service contracts. Family owned since 2009.
## Services
Emergency repair (24/7 line), boiler and furnace installation, annual maintenance plans for homes and small offices.
## News (Aug 19)
Cobalt Ridge adds a second dispatch office in Lakeview to cut drive times.
## Careers
Hiring two service technicians and a part-time scheduler.

Sample output

Real output from run 4 on 2026-10-14, using the current prompt, from an unspecified AI assistant (Grok Bot agent session; model name not shown). Copied as produced. Nothing was fixed after the fact; the problems we found in it are listed under “How it went”.

**1. What the company does**
Cobalt Ridge Mechanical is a 38-person plumbing and heating company in the north suburbs. It does residential repairs, boiler and furnace installs, and light commercial service contracts.

**2. Who they sell to**
Homeowners, small offices (maintenance plans), and light commercial customers. Unknown: the mix, or the size of the commercial contracts.

**3. Recent changes or news**
- Aug 19: added a second dispatch office in Lakeview "to cut drive times."
- Careers page (undated, so not confirmed as recent): hiring two service technicians and a part-time scheduler.

**4. Three things an Operations Manager likely cares about**
- Hypothesis, not fact: drive time between jobs. Text: "to cut drive times."
- Hypothesis, not fact: coordinating two dispatch offices. Text: "second dispatch office in Lakeview."
- Hypothesis, not fact: fitting emergency calls into scheduled work. Text: "Emergency repair (24/7 line)."

**5. Five questions**
1. How do the two dispatch offices decide who sends which technician?
2. What does an emergency call do to the rest of the day's schedule?
3. How are routes planned today?
4. What did the Lakeview office change about drive times?
5. Who would sit in on a look at dispatch tools?

**6. Possible reasons to reach out now**
- Second dispatch office announced Aug 19.
- Open part-time scheduler role (undated).

**7. Unknown / verify**
- Current dispatch or routing tool
- Whether the scheduler posting is still open
- Who owns tool decisions
- How many trucks are on the road
- Whether the Lakeview office is already running

Check before you send

• Every item in sections 1-3 appears in your pasted text. Delete anything that doesn’t.
• Section 4 is labeled hypothesis. Don’t state it to the prospect as fact.
• Dates on news items match the source.
• Names and titles are spelled right before you use them anywhere.

How it went

• Test date, tool, model: 2026-10-14. Unspecified AI assistant (Grok Bot agent session; the model name is not shown to the agent). One tool only.
• Runs logged: 6, on 6 different fake inputs (see the test log). Runs 1-3 used the original prompt. Runs 4-6 are a second round on new fake inputs after the prompt was fixed, as listed below. Every run was done by the same AI assistant session that wrote the fake input, and in round 2 the same session also edited the prompt, so these are not blind or independent runs. A script counted words and characters against the prompt’s limits. No second person re-ran it from the written steps.
• Why the label is “Tried in an AI tool,” not “tested”: we reserve “tested” for plays with at least 3 logged runs in at least 2 different AI tools, and we ran one tool. We also have no independent reproduction of the steps and no timed manual baseline. The runs are real, but this is one tool’s behavior, and 3 runs per round is too few to say a problem is gone.
• Time saved: Not measured. We did not time a manual baseline or an assisted run.
• What we changed after round 1: Added: no-additions rule; ‘If the pasted text is under about 100 words, say so at the top and write ONLY sections 1, 2, and 7.’
• Round 2 (current prompt, runs 4-6). What worked:
  • Thin-text stop rule worked: with a 6-word input (run 5) the output was only sections 1, 2, and 7 and no invented hypotheses or questions. In the earlier round the same kind of input produced a padded brief.
  • Runs 4 and 6 stayed at 258 and 216 words (limit 350) with all 7 sections. A script found no number in the output that wasn’t in the input and no gendered pronouns.
  • The undated Careers posting was labeled ‘undated’ in run 4, which the original prompt never asked for; that was the model’s own choice and may not repeat.
• Round 2: what failed or needed fixing:
  • The ‘under about 100 words’ cutoff is fuzzy. We only ran one clearly thin input (6 words) and two clearly normal ones. A 90-to-130-word paste has not been tried.
  • Section 4 hypotheses are still guesses about what a role cares about. They are labeled and tied to a line of text, but a rep could still read one as research.
  • Not fixed (was not on the fix list): nothing in the prompt forces the model to mark undated items as undated.
• Checked against “Check before you send”: Items 1-3 passed in runs 4 and 6; run 5 correctly produced no hypotheses. Names and titles were not invented.
• Edge cases: Thin input (run 5) now behaves. Not tried: text just over the 100-word line.
• Round 1 (original prompt, runs 1-3): what failed, kept for reference:
  • Run 3 (one sentence of pasted text): the prompt does not stop the model from writing hypotheses and questions anyway. They were generic and cited no real text (it wrote ‘Text: none’), so the brief looked like research but wasn’t.
  • Run 2 section 6 reasoned past the text (‘second laser cutter added, so more parts to inspect’). It was labeled inference, but ‘only if the text supports them’ was bent.
  • Undated Careers postings were listed under ‘Recent changes or news’. Easy to misread as new.
• Not yet observed: how a different model or tool handles the same prompt (a second-tool run is still to do).

Watch-outs

Privacy: only paste public company text, not private CRM notes, into a consumer AI tool unless your company allows it. Don’t present the AI’s hypotheses to a prospect as things you ‘saw’.

See the Safe Use page: AI drafts, you decide, you send.

Variations

• Owner version: swap ‘contact’s job title’ for ‘what this customer buys from vendors like me’, and use it before a bid.
• Manager version: ask for ‘three questions to coach the rep on before this call.’
• Short version: sections 3, 4, and 7 only.

Sources

None. Original wording. (Add any public framework or post used, with link and checked-on date.)

Retest

Re-check by: 2026-12-13 (60 days after the 2026-10-14 runs), or sooner when run in a second tool.

Test log

Changelog

• 2026-10-14: Draft written. Not run yet.
• 2026-10-14: Ran 3 times in one AI assistant on fake data. Filled in sample input and output, how it went, and the test log. Front matter set to runs_logged 3, last_tested 2026-10-14, retest_by 2026-12-13. The 2-tool bar for the word “tested” is not met, so the play is not labeled tested. The prompt text was not changed after the runs.
• 2026-10-14: Prompt revised after round 1. Added: no-additions rule; ‘If the pasted text is under about 100 words, say so at the top and write ONLY sections 1, 2, and 7.’ Re-ran 3 fresh runs (4-6) on new fake inputs in the same AI assistant. Updated the sample input and output to run 4, the “How it went” section, and the test log (original runs 1-3 kept, marked). Front matter set to runs_logged 6, last_tested 2026-10-14, retest_by 2026-12-13. The 2-tool bar for the word “tested” is not met, so the play is not labeled tested.
• 2026-10-15: Published with the evidence label “Tried in an AI tool” (6 runs, 1 tool, fictional inputs). Earlier wording that implied more evidence than one tool’s runs was replaced with the label; the runs, results, and limits above are unchanged.