Projects: Deal-Level AI Context for Gavel Exec Chat - Gavel

Documentation Index

Fetch the complete documentation index at: /llms.txt

Use this file to discover all available pages before exploring further.

Projects are Gavel Exec’s customization layer. When you work inside a Project, you give the AI your firm’s actual context: the prior drafts you rely on, the positions your client will not move from, and the instructions that reflect how you practice. Every chat you run within a Project draws on that context automatically, so you do not have to re-explain your client’s situation or your preferred approach with every prompt. A Project turns Gavel Exec from a general legal AI into something that understands how your firm handles a specific deal type, client, or practice area. Small law firms can use Projects to build AI that actually reflects how they practice law — no engineering team required.

The two components of a Project

Every Project has two building blocks.

Project files

Documents and data you want the AI to draw from: prior drafts of the agreement, deal notes, templates, prior redlines, legal background memos, or any other files relevant to the matter. Gavel Exec indexes these files and uses them as reference material in every chat within the Project.

Project instructions

The core positions, rules, and preferences you want the AI to follow. For example: “The client will not accept uncapped indemnity exposure,” “Governing law must be Delaware,” or “Use the tone and defined terms from the attached template.” Instructions are applied to every prompt you make inside the Project.

How Projects make chat context-aware

Once you open a chat inside a Project, every prompt you send is already informed by your Project Files and Project Instructions. You do not have to re-state the client’s positions or paste in prior language. Gavel Exec operates with that context built in.

• If your Project Instructions specify that the client will not accept uncapped indemnity exposure, and your Project Files include a prior draft with approved carveout language, Gavel Exec takes both into account when redlining an indemnification clause. It will not suggest language your client has already rejected, and it can replicate the carveout structure your team has already agreed to.
• If you upload a prior executed agreement as a Project File and instruct the AI to draft in the same style, it uses that agreement’s defined terms, sentence structure, and clause architecture as a reference when drafting new provisions.

This is the difference between asking a generic AI to redline a clause and asking an associate who has read the entire file.

Setting up a Project

1. Click on Sources: No context added
   In the Gavel Exec panel, navigate to the top of the tab and click on the Sources block. Click on the ”+ Create” to create a Project. Give it a name that identifies the matter or client.

2. Add Project files
   Upload the documents you want Gavel Exec to reference: prior drafts, templates, deal notes, term sheets, or any background files relevant to the matter. You can add multiple files.

3. Add Project instructions
   Write out the positions and rules you want the AI to follow for this Project. Be specific: define the client’s hard limits, preferred language for key provisions, and any stylistic requirements.

4. Select your new Project as your "Source"
   Begin your chat session. Every prompt in that session draws on your files and instructions automatically.

Project files are permanent reference material available in every chat within that Project. This is different from uploading a file in a standard Chat session, which applies only to that single session. See Upload reference files for more on ad-hoc uploads.

Playbooks vs. Projects

• Use Projects when... You need the AI to act like a senior associate: handling complex judgment calls, drawing on multiple documents, and performing deep review or negotiation with deal-specific context. Projects are best for matters where you have significant existing work product and a nuanced set of client positions that differ from deal to deal.

• Use Playbooks when...
  You want the AI to act like a junior associate applying a checklist: enforcing standard language, flagging deviations from policy, and running a consistent first-pass review across a class of contracts. Playbooks trade depth for consistency.