Kickoff Deck Auto-Assembler

The job is first assembly of a standard kickoff, not a client-ready story. The cheaper path is already familiar: duplicate last Thursday's deck, find-replace the client name, and hope a leftover workstream is not still on slide 9.

That path is why rooms open on the wrong logo, a legal name from a different entity, or an agenda item nobody sold. Building from a signed SOW, current bios, and the house agenda template is faster than that copy. It does not write the partner's first ten minutes.

Speed is the only outcome this use case is for. You get a skeleton that matches the contract and this week's staffing plan. If the model writes "why this matters now," delete it and leave those slides blank.

Do not run this on a SOW that is still a discovery draft. Finish the commercial document first, including with a SOW drafter from discovery transcript if that is how you get to signed scope. A kickoff built from an unsigned draft will present open points as agreed.

Feed four sources. Forbid a fifth.

1. The signed SOW. Legal entity, workstreams, deliverables, exclusions, timeline, governance. If a bullet is not in the SOW, it is not on a slide.
2. Current team bios. Names, roles, and one relevant line for people staffed this week, not last deck's bios or the proposal team that rotated off.
3. Client context from a stakeholder onboarding brief synthesizer. Who is in the room, how they name themselves, what not to say. Not a news dump, and not another client's case.
4. The house kickoff agenda template. Welcome, team, objectives, scope, timeline, working norms, next two weeks, questions. Labels and order come from this file only.

The prompt is an instruction to fill that template from those files, not "make a beautiful deck for a supply-chain kickoff." Beauty prompts invent workshops, extra workstreams, and a title slide that reused last job's brand kit.

Put the legal name and logo path at the top: entity as on the SOW cover, logo from this engagement's brand folder. Give it the empty house template, not last week's file. Another account's filled deck, offered as "an example of tone," is how leftover slides get reused.

Tools in the Gamma, Plus AI, Microsoft Copilot in PowerPoint, and Beautiful.ai class produce a first draft of slides from prompts, documents, and templates. Treat them as first-assembly tools. Do not rank them or expect a client-ready story. Export into the firm's master and edit there.

Leave the opening slides as stubs. The partner writes why we are here, what changed since the proposal, and how they want the room to feel. An assembled paragraph on "our shared ambition" reads as assembled.

Generate only the assembly sections: title (legal name and logos from the locked files); agenda (house template, same order, same labels); objectives and scope (SOW language, including exclusions); team (current bios); timeline (SOW milestones, not a new plan); working norms (template plus the named cadence from the SOW); next two weeks (only activities the SOW or staffing plan already named).

Do not generate relationship history unless the partner supplies a sentence, and do not generate "why this matters now." Do not generate a redesigned agenda: design sprint, opportunity workshop, extra steering forum. If access, sign-offs, or environments are still open, that belongs in a pre-kickoff readiness gap detector shared with the sponsor, not on a slide that says the team is ready. A short list of risks the team has already accepted can come from a risk register bootstrapper. Do not let the model invent new ones for the room.

If the model adds an agenda item that is not in the template, delete it. Cut a third of whatever comes back. Completeness is not the goal at kickoff.

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

Hale & Rowan kicks off a 14-week network-planning engagement with Northline Freight LLC on Friday. The signed SOW names three workstreams (lane-cost baseline, hub scenario model, 90-day implementation plan); staffed are Marcus Chen, Tomas Nguyen, and Asha Patel; partner Elena Voss opens against the house eight-item agenda.

Thursday, Marcus does not duplicate last week's Vesper Retail kickoff. He feeds the signed SOW, three current bios, a one-page Northline brief, and the empty house template, with the legal name and brand-folder logo locked, opening slides left blank.

The draft also does three things that would have landed in the room. The title slide uses "Northline" and a generic freight mark the tool already had, not the LLC name and not the logo in the brand folder. Slide 11 is still headed "Week 3 Riverside DC walkthrough," leftover structure from Vesper, which Marcus had attached as "an example of tone." The agenda has grown a ninth item, "Opportunity workshop," which is not in the house template and not in the SOW.

All three go. Elena's opening stays blank. Scope is checked line by line against the SOW: hub scenario model stays; proposal language that was not contracted comes out. Bios match this week's staffing plan. Friday, Elena talks for ten minutes without a generated paragraph. The rest of the deck is the contract and the team.

Folders and channels come later, via a project setup automation agent, not from the deck tool.

Review the exported file, not the chat.

• Legal name matches the SOW cover on every slide; nicknames only if the client uses them in writing.
• Logos: this client's current mark and this firm's, never a generated "close enough" icon or a prior-account color.
• Search for the last three client names your team served. Any hit is a leftover slide.
• Agenda labels and order match the house template; extra items are deletions.
• Every scope and timeline bullet has a home in the SOW; open points stay labeled open.
• Team slide matches this week's staffing plan and current bios.
• Opening slides are stubs, or only sentences the partner will actually say.

If a check fails, fix the sources and regenerate those sections. Hand-edits on a polluted draft miss the second leftover.

Trial this on one kickoff you are already running, in parallel with the usual copy-from-last-week file. The engagement manager owns the four-file pack and the mechanical review. The partner owns the first ten minutes.

Keep using the assembler if the partner did not rewrite the opening, the client did not correct a name, logo, or workstream, and the agenda stayed on the house list. Spend the saved time on the SOW line-check and staffing, not on fonts.

Drop it if the partner still walks in with last week's file. This use case is cheap. Trust is not. Do not scale until the logo and leftover-slide failures have actually stopped. Do not treat a generated deck as proof the engagement is ready. Readiness is access, sign-off, and environment. The deck is a skeleton that matches what you sold.