A Claude Project is a persistent workspace with memory — it retains your uploaded context and instructions across every conversation — and that persistence is what makes AI useful across a long, complex deal instead of just within a single prompt. Most reps use Claude as a one-shot tool: open a conversation, get an output, close the tab. That approach leaves the compounding value on the table, and compounding is exactly what big deals reward.
Why one-shot usage caps out on big deals
Enterprise deals run months and accumulate context: discovery findings, stakeholder politics, competitive dynamics, pricing history, the champion's internal constraints. In one-shot usage, every conversation starts from zero, so every output is only as informed as what you re-pasted that day — and nobody re-pastes four months of context. The result is AI that helps with tasks but never understands the deal. A Project inverts this: context loads once, persists, and every subsequent output draws on the full accumulated picture.
Setting up a deal project
Create a Project per major opportunity and give it two things. First, project instructions — a standing brief that frames every conversation: who you are, what you sell, what this deal is, and how you want outputs shaped. Second, project knowledge — the documents that constitute deal reality: discovery call extractions, the org chart as you understand it, your MEDDPICC assessment, relevant proposal drafts, and competitor notes.
Prompt: "Write project instructions for a deal workspace. Context: I'm an enterprise AE at [COMPANY] selling [SOLUTION]. The deal: [ACCOUNT], [DEAL SIZE], currently at [STAGE]. Key stakeholders: [LIST]. Primary competitor: [COMPETITOR]. Instructions should tell the assistant to: always ground outputs in the uploaded deal documents, flag when I'm asking for something the documents contradict, use the buyer's verbatim language from the discovery extractions, and keep all drafts under my stated word limits."
The contradiction-flagging instruction earns its place within a week — a workspace that pushes back when your ask conflicts with what the buyer actually said is a second set of eyes on the deal.
What a deal project makes possible
Cumulative drafting. The follow-up email in month three automatically reflects the pricing sensitivity surfaced in month one, because the workspace never forgot it. Every draft is deal-aware without re-briefing.
Stakeholder-specific everything. Ask for the business case reframed for the CFO, then the CISO, and the workspace adapts using what it knows about each from your uploaded notes — their stated concerns, their language, their role in the decision.
Deal-state interrogation. The workspace becomes queryable: "What did the champion say about the security review timeline?" "List every commitment we've made this quarter." "What's changed in my MEDDPICC assessment since the last upload?" Questions that used to mean re-reading four months of notes get answered in seconds.
Pre-meeting simulation. With the full context loaded, "act as their CFO and challenge this proposal" produces objections grounded in this deal's actual economics and politics rather than generic pushback.
Prompt: "Using everything in this project's knowledge, produce a deal-state snapshot: current MEDDPICC assessment with evidence per element, every open commitment on both sides with owners and dates, the two biggest risks based on [STAKEHOLDER] activity, and the three questions I most need answered before [NEXT MILESTONE]."
Keeping the project current
A deal workspace is only as good as its freshest upload. Wire maintenance to your existing post-call routine: the same extraction that writes your CRM note gets uploaded to the project. Weekly, replace stale documents rather than stacking versions — five current documents outperform twenty layered ones, because contradictory vintages of the same information degrade every output. When the deal closes, run one final session: extract the win/loss patterns, then archive.
Where the bigger-deal effect actually comes from
The compounding shows up in three places big deals are won and lost. Consistency: month-four communications carry the same framing and facts as month one, which committees read as competence. Responsiveness: when the champion asks for a CFO-ready version by tomorrow morning, the workspace produces it in minutes because it already holds the economics and the CFO's stated concerns. And memory under pressure: in late-stage negotiation, "list every commitment either side has made" retrieves in seconds the record that decides who is renegotiating history. None of these is a feature; all of them are what persistent context does to a long deal.
The portfolio pattern
Beyond per-deal workspaces, keep one standing project for your territory — ICP definition, industry knowledge base, competitive battlecards, your best-performing outreach — and use it for everything pre-deal: prospecting research, first-touch drafting, hypothesis building. New opportunities graduate from the territory project to their own deal project at qualification. Two tiers, clean handoff, nothing re-explained twice.
Frequently Asked Questions
How many deal projects should I run at once?
One per genuinely complex opportunity — typically your top five to eight deals. Transactional deals don't accumulate enough context to justify the setup; the territory project covers them.
What should I upload first when starting a deal project?
Discovery call extractions, in the buyer's verbatim language. They are the highest-density deal context and the source every other output should draw from; org charts and MEDDPICC notes come second.
Is there a data-security concern with uploading deal documents?
Follow your company's AI tool policy, and strip anything policy requires — customer identifiers, contract specifics. The workflow's structure survives redaction; check policy before uploading, not after.
Put It to Work
The project-instruction and deal-extraction prompts here are part of the AI setup collection in the Promptifi library. Browse the library and stand up a workspace for your biggest live deal today.