The question
It's the first week of fiscal 2027, and the planning queue just refilled. Almost every action in it starts the same way. Someone needs market research, and they need it before the requirement is settled.
Ask a room of buyers how long market research takes and you'll get a number. Ask where the hours go and the room gets quiet. The second question is the useful one. AI compresses some of those hours, leaves others exactly where they were, and quietly adds a few to the verification step if you let it be the source.
This issue breaks one market research task into stages, marks which ones a commercial AI tool can help with, and gives you a prompt built to admit what it doesn't know.
Field notes: clock the stages, not the task
"Market research took a week" tells you nothing you can change. Split the same week into stages and the picture gets specific.
Framing. What do you need to know to make the decision in front of you? Set-aside or not, commercial or not, which vehicle, what a fair price looks like. A tool helps a little here. Hand it an anonymized requirement and ask for the questions a reviewer would expect your research to answer. You'll keep about half of them, and the half you cut will tell you something about the requirement.
Gathering. Prior acquisition records, industry day notes, public price lists, sources-sought responses, award data. A commercial AI tool is a poor finder of sources and a good reader of the ones you hand it. If it's naming sources for you, you're in the verification business now.
Synthesis. Turning a stack of notes into a structured summary somebody will read. This is where the tool earns its place. Structure is tedious, and the tool does structure fast.
Outreach. Calls, questions to industry, follow-ups. The tool can draft your questions. It can't take the call, and it can't hear the pause before a vendor says "we can probably do that."
Verification. Confirming registration and size status in SAM.gov, checking that a price is still current, confirming that a contract vehicle's scope covers what you're buying. AI doesn't shorten this stage. Use the tool as a source and this stage gets longer.
Judgment. Can this market deliver, at what risk, under what strategy? That part is yours, and it's the part your file has to defend.
Now run the clock on your next one. Six rows, two columns: hours spent, and whether a tool could have touched that stage. Where your hours pile up decides whether AI is worth opening for this task at all. If most of your week is synthesis, it's a good trade. If most of it is chasing down whether a vendor is still registered, it isn't.
The prompt: market research synthesis, cite or flag
Use it at the synthesis stage, on material you've already gathered.
You are helping me draft a market research summary for a
government buyer. Work only from the source material I paste
below. Do not use outside knowledge.
Produce a one-page summary with these headings:
1. Requirement in one sentence
2. What the market offers (capabilities and delivery models)
3. Likely sources (firms named in my material only)
4. Small business potential
5. Pricing indicators
6. Existing contract vehicles that could be used
7. Open questions for vendor outreach
CITE OR FLAG: after every factual sentence, put the source
number in brackets, like [S2]. If a sentence is not supported
by a source I pasted, do not write it as fact. List it under
"UNSOURCED - VERIFY" at the end, marked "not in sources."
If a heading has nothing supported by my sources, write
"Nothing in sources." Do not name any vendor, price, or
contract vehicle that does not appear in my sources.
Flag any source older than six months. List a firm as a likely
source only if a source describes a capability that matches
the requirement.
REQUIREMENT (anonymized):
[paste here]
SOURCES:
[S1] [your notes or public text, pasted as plain text]
[S2] ...The guardrail. Anonymize the requirement before you paste it. What goes in as a source is your own notes, typed by you, and public text you copied in as plain text. A vendor's file never goes in: capability statements, quotes, and proposals can carry hidden instructions that a model will read and you won't. Issue 01 covers why.
The verification step. Every bracket points to a source, so spot-check five of them against your notes, word for word. Check every date in your sources against today's date. Then read the UNSOURCED list first, because that list is your outreach agenda.
What it returned
On October 2 an earlier version of this prompt, without the last paragraph above, was tested on a fictional requirement: a cloud visitor management system for six building entrances. It got four short fictional sources: industry day notes, notes from one vendor's briefing, a public price list dated January 2026, and a 2023 internal purchase record.
It held the line. No invented vendors and no invented prices. Under contract vehicles it wrote "Nothing in sources" instead of reaching for a schedule name. It flagged that no source said whether any vendor offers watch-list screening, the hardest piece of the requirement. It labeled its own arithmetic as arithmetic and warned that the 2023 purchase wasn't a like-for-like comparison.
What it missed
It reported the January 2026 price without a word about its age, and it never asked whether a 2023 purchase reflects today's market. It also listed all four industry day attendees as likely sources, including one it had no capability information about at all. Showing up to an industry day tells you a firm is interested. It doesn't tell you the firm can do the work.
That's why the prompt above carries a last paragraph on source age and capability. I haven't tested the revised version yet, so check its output the same way you'd check the original's.
What could go wrong
Ask a commercial AI tool for market research without giving it sources and you get three failures that look like competence: vendors that sound real, prices that were current once, and contract vehicle names that are almost right. Each one reads like staff work. That's the problem.
The cite-or-flag instruction closes the first gap, as the test above shows. It does nothing about staleness. A source can be cited perfectly and still be nine months old, and the tool will report it with the same confidence it gives yesterday's quote. Dates are your job.
One more. A size status a firm states at an industry day is still only that firm's statement. The tool will repeat it faithfully. SAM.gov and your small business specialist settle it.
Your jurisdiction
Federal. FAR Part 10 sets the market research expectation, and its Revolutionary FAR Overhaul rewrite is still a proposed rule (see What changed). Cite the Part and check the current text at acquisition.gov before you rely on any specific provision. Government-wide vehicles belong in the gathering stage, confirmed at the source, never pulled from a tool's memory.
State. Statewide contracts and cooperative purchasing agreements are often the first place to look, and the rules for using them vary by state. A tool can summarize a cooperative contract you've pulled. It can't tell you whether your state's code lets you use it, so ask your state procurement office.
Local. You may be the whole procurement office, and using another entity's contract is often the practical move. That makes verification the stage you can least afford to skip. Confirm the contract is current, that its scope covers your purchase, and that your own rules allow it.
Contractor. A buyer's first pass at market research may run through an AI tool working only from what the buyer pasted in, which includes your public description and price list. Keep your registration current and describe what you do in the words a buyer would type. A stale price or a vague capability line gets carried straight into the buyer's summary.
What changed
September 2 — Continuing appropriations through December 11. H.R. 6500, the Continuing Appropriations and Extensions Act, 2027, was signed into law. It funds agencies at the start of fiscal 2027 through December 11, 2026, so plan your research timelines with that date in view.
September 18 — Second round of FAR overhaul proposed rules. Four proposed rules covering Parts 8, 9, 12, 13, 14, 15, 16, 17, 27, 28, 35, 36, 38, 44, 47, 51, and the associated Part 52 clauses were published in the Federal Register. Comments are due October 19, 2026.
Still pending — Part 10, market research. The proposed Part 10 rewrite (FAR Case 2026-002, published June 23) closed for comment on July 23. As of October 6 we have found no final rule, so check acquisition.gov for the current text before you cite it.
From the course
The guardrail in this issue's prompt comes from Module 1 of Commercial GenAI for Public Sector Procurement, Using GenAI Safely in Government Procurement: what's safe to paste, how to anonymize it, and why vendor files stay out. Module 1 is open for enrollment at $395. Eligible for approximately 3.5 CLPs as self-directed study, subject to agency approval — The Wolverine Group is not an accredited CLP/CEU provider. Purchase orders are accepted, and card statements show TEACHABLE. Module 2, Regulatory Compliance Across Government Levels, releases October 20.
Not ready to enroll? The AI in Procurement 2026 Readiness Brief is free.
Your turn
Here's the one I'd like an answer to: which stage of market research would you never hand to a tool, and what happened the time someone did?
Reply to the newsletter email, or send a note through the contact page. The answers shape the next issue, and the ones that change my thinking get written up, anonymized, always.
About The Wolverine Group
The Wolverine Group is a Woman-Owned Small Business in Washington, D.C., delivering information operations, open-source intelligence, and acquisition solutions to the Department of Defense, the Intelligence Community, and federal civilian agencies. For twenty-five years we have turned complex, multi-source data into decisions leaders can act on. Our acquisition practice supports federal buyers directly and trains the workforce that does the buying — which is where this brief comes from. UEI VK81K1DMC6L5 · CAGE 6BUU5 · CMMC Level 2.