Pre-Pitch Client Context Brief

Partners will not use a pack that last embarrassed them in the room. The job is a short, dated intelligence brief they can trust out loud, not an overnight research dump.

Speed is real: the 4 a.m. news scramble goes away. The cost of that speed is confident prose about a CEO who left, a plant that already closed, or a prior engagement the client would rather you never mention. Treat every bullet as something the partner might say in the first minute. If you would not stake that minute on it, it does not go in.

Do not build a brief for a pursuit you have not chosen. Run a bid/no-bid opportunity classifier first. A polished pack on a no-bid is wasted partner time and another chance to mix a different client's history into this one.

The brief is four blocks: who they are, what they just did, what we already know from prior work, and what not to say. Anything else is a file.

Who they are. Legal entity, ownership, leadership with as-of dates, and the business mix that will sit in the room. Financial and news research tools in the AlphaSense and PitchBook class feed ownership, filings, and transcripts. A title without a date is a rumor.

What they just did. Events in the last 90 days that change the conversation: filings, earnings language, leadership moves, announced network or portfolio changes, labor actions, material litigation. Older events belong only if they are still in force.

What we already know. Prior SOWs, submitted proposals, after-action notes, and CRM on this account. Enterprise retrieval over the firm's own files sits in the Glean and Hebbia class. Separate "we worked here" from "we know this from another client." The second category never enters this brief.

What not to say. Landmines: a loss this buyer still remembers, an executive who has left, a framing they rejected last time, confidential leftovers from another account, and any personal detail about an executive. Retrieval will skip this block unless you require it.

If the pack needs a contents page, it failed. Feed the cleaned brief into a RAG proposal draft generator later. Do not let the generator become the brief.

This is a worked example with made-up firms, written to show the cuts, not a case study with results.

Northshore Partners, a 400-person operations firm, is pitching Meridian Foods, a PE-backed CPG company owned by Oakridge Capital since early 2025. The oral is a 16-week supply-chain operating-model pitch, three days after Meridian announced it will close the Dayton plant and move volume to Savannah. The capture lead has the intern dump and a four-block brief.

Who they are, and what just happened

Public research showed Oakridge still owns them; COO Priya Shah started in May 2026 from a larger CPG competitor; Meridian frames Dayton as network optimization, not distress. The intern dump also carried a 2023 CEO quote and a clip that still named the prior COO. Both were cut. Saying either name would have told Shah the firm had not read this year.

What we already knew, and what not to say

Firm files showed a 2022 demand-planning diagnostic that closed mixed, a 2024 warehouse-network bid lost to a specialist, and CRM notes that the relationship partner last dined with the former CPO, who left in 2025. The brief kept one line: "2022 diagnostic, mixed close; do not cite findings as current." The 2024 loss sat under what not to say unless the client raised it. The former CPO was listed as do-not-name, not as a warm intro.

The same retrieval surfaced a 2021 plant-network study for a different food company, still full of that client's site names. That file is another account. It never entered Meridian's brief. A comparable past scope retriever can find that study for the proposal team's method spine. It is not this client's history.

The partner used Shah's start date and the Savannah move in the opening. Nobody mentioned Dayton as a crisis, the 2024 loss, or the former CPO.

Mixing news clips and old SOWs in one pile is how another client's site names land in this oral. Public facts can be said in the room. Firm files often cannot. Retrieval must show corpus, source, and date on every chunk before anyone copies a bullet.

Run two retrievals, then a human merge:

1. Public corpus: filings, earnings transcripts, press releases, trade press, ownership. Tools in the AlphaSense and PitchBook class belong here. They do not see your SharePoint.
2. Firm corpus: this account's SOWs, submitted proposals, after-action notes, CRM. Tools in the Glean and Hebbia class belong here. Firm knowledge assistants in the McKinsey Lilli class belong on approved sector notes and credentials, not on a partner's private thread.

Do not index drafts, competitor teardowns, or marked-up redlines. Do not retrieve another account's deliverables because the industry string matches. If a file has no date, it does not feed "what they just did."

Win/loss pattern synthesis belongs in the "what not to say" block when the same buyer or sector has killed a theme before. It is not a claim about this company.

A claim without an as-of date is how you walk in talking about last year's CEO. Keep the windows boring and enforced:

• Leadership and org: last 90 days, verified against an 8-K or equivalent press release, not a blog roundup.
• Financials and ownership: last reported period, labeled. A PitchBook-class round from three years ago is history unless you confirm nothing followed.
• "What they just did": 90 days, unless the event is still in force.
• Prior work on this account: last 36 months as relationship. Older work is credentials only if it sits on the approved case list.

Never put these in the brief:

• Unsourced rumor, banker color, or "we heard they might sell."
• Another client's confidential findings, site names, or process maps.
• Personal details about executives, and internal firm color about the account.
• Rates, fees, or commercial terms from an old card.
• Speculative win themes. Those are the pitch, not intelligence.
• Anything a stakeholder onboarding brief synthesizer would need after a win. That is a different artifact, with different readers and a different confidentiality bar.

The failure mode is a wrong name said with confidence. Force a source line under every leadership bullet. If the capture lead cannot find the filing, the bullet is out.

Score the trial on whether the partner used the brief and whether the client corrected a fact, not on pages produced. A page-count trial will scale the intern dump.

Pick one oral you are already staffing. Generate the four-block pack 72 hours out, in parallel with the usual dump. The capture lead checks names and dates against primary sources. The partner owns "what not to say" and cuts anything they would not say themselves. Bring both packs to prep. Watch which one the room actually quotes.

Judge it on:

• Zero client corrections on facts in the room ("we sold that division").
• At least one dated fact used in the opening that the dump had buried.
• Prep time spent on argument, not on hunting news the morning of.
• Wrong bullets traced to corpus mix or recency, then the index fixed before the next oral.
• The partner asking for the pack on the next pursuit.

Do not scale to every pursuit until that trust is real. If they still walk in with last year's talking points, the brief is too long or last week's error is still in memory. Effort on this use case is the hygiene: two corpora, dates, and a required landmine block. The retrieval is the cheap part.