If your company runs on Microsoft 365, the strongest account research tool you have is probably one you have never opened. Copilot Researcher is an agent inside Microsoft Copilot that runs multi-step research and hands back a cited report, and unlike a plain chat window it reads your own work content alongside the web. This is a working method for using Copilot Researcher for account research: what to ask, how to steer it, how to check what comes back, and how to turn it into a call plan.
One constraint shapes everything below. Microsoft caps Researcher at 25 queries per user per month, so point it at the accounts you would otherwise prepare for by hand.
What Researcher does that a chat window does not
Microsoft describes it plainly: "Researcher agent is an intelligent assistant within Microsoft Copilot, designed to handle complex, multi-step research tasks. It delivers comprehensive, source-cited reports to help you make informed decisions faster. Researcher draws insights from both the web and your work content - including files, emails, meetings, and chats you have access to."
That last clause is the reason to use it for account research rather than a general assistant. Web research tells you what an account said in public. Your work content tells you what it said to your own company: the thread a colleague had with their VP of Operations eighteen months ago, the notes from a deal that stalled. Most reps re-research accounts their employer already knows, because that history sits in mailboxes nobody searches.
You will find it under Agents in Chat in the Microsoft Copilot app, on desktop or the iOS and Android app. Microsoft states it is "currently available to Microsoft 365 Premium and Pro subscribers and to Microsoft 365 business and enterprise users with a Microsoft 365 Copilot add-on license."
Two Microsoft products sound alike. Researcher is a general research agent inside Copilot; Copilot for Sales is a separate CRM-connected licence in Outlook and Teams. You do not need the second to use the first, and the Copilot setup guide covers the wider stack.
Start a query and walk away: simple ones return in under five minutes, complex ones take 10 to 45. Researcher "respects the same permissions, policies, and compliance you already rely on," so it reads only what your account can already open. That settles the governance question, not the judgement one, and what not to paste into a chatbot applies.
Write the prompt that earns the query
Microsoft's guidance is to be specific and to define your search scope, deciding "whether you want data from workplace resources, the web, or both." A vague request produces a long report with thin conclusions and costs the same as a sharp one. The prompt below puts internal history first on purpose.
I am preparing for a first meeting with [ACCOUNT]. I sell [WHAT YOU SELL] to [BUYER TITLE]. Search both my work content and the web. Produce a research brief with these sections, in this order: 1. Our history with this account. Every prior contact anyone at my organisation has had, with dates, names and what was discussed, from my emails, meetings, chats and files. If there is none, say so explicitly rather than dropping the section. 2. What changed at the account in the last 12 months: leadership, funding, structure, products, regulation. 3. Their stated priorities, quoted from their own materials, each with a source and a date. 4. The likely buying committee, with each person's role and the evidence they are involved. 5. Three hypotheses about where [WHAT YOU SELL] fits, each with the evidence for and against it. 6. Five questions I should ask on the call that this research cannot answer. Cite every claim. Mark any claim that rests on a single source with [single source]. Mark anything you are inferring rather than reporting with [inference].
Answer the clarifying questions instead of skipping them
Researcher may ask follow-up questions before it runs. Microsoft calls this phase "important to help you steer the research," and it is the cheapest thirty seconds in the process. Answer with constraints: which business unit, which region, which competitor you are displacing, and what you already know.
One caveat: Microsoft notes the model "was trained on web data available up to 2024," with recent web information arriving through the Bing index. Date-check anything time-sensitive before you say it out loud.
Pressure-test the report before it reaches a call
A cited report reads as more authoritative than it is. A citation proves a source exists; it does not prove the source supports the claim, is current or is representative. Run this in the same conversation.
Review the brief you just produced. Do not rewrite it. 1. List every claim you marked [single source], and what a second source would have to show to confirm each one. 2. List every statement that is an inference presented as fact. 3. Name the three most important things about this account your sources do not cover at all. 4. Say which sections rest on our internal history and which on the web, and point out anywhere the two disagree. Answer those four questions only.
Question four is the one that pays. Where your own record contradicts the public story, that gap is usually the most useful thing in the report.
Turn the report into something you use in the room
A research report is not a call plan. This converts it:
Turn the brief into a one-page call plan: - An opening line referencing something specific and recent. No flattery, no "I was impressed by". - The single hypothesis I am testing on this call, in one sentence. - Six discovery questions in order, each tied to a hypothesis. - Two things our history with this account suggests I should avoid raising in a first meeting, and why. - What I must learn before I can qualify the opportunity. Write it so I can read it in ninety seconds before the call starts.
Common mistakes
Spending a query on an account you would not have prepared for anyway. Save them for first meetings, renewals worth defending, and deals that went quiet unexplained.
Clicking past the clarifying questions. That is how a run gets spent on the parent company's consumer division.
Letting it search the web only. Then you have a slower free tool. The internal half is what no competitor can reproduce.
Treating citations as verification. Check the two or three facts you intend to repeat to the buyer; leave the rest as context.
Asking for everything about a company. Scope the question to the decision in front of you.
Blaming your prompt when the web section comes back empty. Web access is a tenant-level setting, and Microsoft is explicit that there is no per-site control: "if web search is disabled at the tenant level, Researcher agent won't use any web data." If every report is internal-only, ask your administrator rather than rewriting the prompt.
Frequently asked questions
How many Researcher queries do I get, and how long does one take?
Microsoft's Researcher documentation states a maximum of 25 queries per user per month. Simple queries return in under five minutes; highly complex ones take 10 to 45 minutes. Both figures come from Microsoft documentation last updated 18 August 2026, read 24 September 2026.
Can Researcher actually read my company's emails and meetings?
Yes, the ones you personally have access to. Microsoft says Researcher draws on "files, emails, meetings, and chats you have access to" and respects "the same permissions, policies, and compliance you already rely on." It can also pull from Graph connectors, which is how systems outside Microsoft 365 reach it once an administrator configures them.
Does this workflow still work if my company uses Gemini or ChatGPT instead?
The structure transfers: brief, clarify, pressure-test, convert to a call plan. What does not transfer is the internal-history section, which depends on the assistant being connected to your company's own mail and files. For the Google version, see Gemini Deep Research for account research; without any Copilot licence, the 15-minute account research brief is the tool-agnostic version.
The three prompts above are deliberately generic so you can paste them today. For versions built for stakeholder mapping, buying signals, filings and competitive intelligence, the Account Research & Buyer Intelligence category has around twenty, starting with "Build a Comprehensive Account Brief with Deep Research".