Diligence.

For most firms, diligence conviction is built inside a single partner's head — slowly, quietly, and without a trail. The problem is not just opacity to the team; it is opacity to every AI system the firm has invested in. If the thinking never leaves someone's skull, no model can sharpen it, no automation can act on it, and no teammate can challenge it. The first step toward an AI-native diligence process is to externalize the work.

Two surfaces, one process

We recommend doing the bulk of the intelligence-grappling work across two tools. An LLM — the model layer the firm standardizes on — for reasoning, synthesis, and adversarial questioning. A collaborative workspace — Notion is what we use — for the persistent, structured, team-editable record of what the firm actually believes and why. Presentations, formal memos, and partner-facing materials still live where they always have: SharePoint, Google Drive, the file drive. The distinction matters. Most people conflate “where we store documents” with “where we do our thinking.” They are not the same thing, and grasping the difference is what unlocks AI assistance in the messy middle of diligence.

The workspace pipeline

When a company enters active diligence, it gets its own page in a dedicated workspace pipeline. The template mirrors the memo generator's sections for continuity, and adds a standard set of headings designed to force specificity:

• Status
• How and when did we meet the company?
• What are the company's ambitions for its next fundraise?
• What is the expected timing and valuation?
• Key Questions — What are the key issues that will make or break this investment? What resources are or should be involved in evaluating the potential? What are next steps to push forward to a decision?
• Pros — Strengths of the opportunity
• Cons — Weaknesses of the opportunity

More details on how to structure this page can be found in the Collaborative Workspace blueprint.

Drill into the exact questions

The highest-leverage exercise is spelling out the specific questions the firm needs answered — not vague themes, but precise, answerable interrogatives. A call with the team produces a flood of impressions; the AI's job is to extract the top diligence questions to wrestle with, land them in the workspace page, and let the team update them as evidence arrives. This keeps the ball moving. It gives the whole system — human and machine — visibility into what the firm is actually trying to resolve. From there, the LLM can pull context and data to help make an informed, evidence-based decision on each question, one way or another.

Messy work needs the right container

This kind of work is inherently messy: writes and rewrites, tagging a teammate for input, pivoting when a new fact arrives. The rigid paragraph structure of a traditional document fights against that motion. A collaborative workspace is built for it — blocks move, pages nest, properties filter. Just as important, it integrates cleanly with the AI layer. Modern AI assistants can read from and write to workspace pages via API in ways that feel native, which means the headless automations that power the firm's diligence stack can push complex, unstructured data in and pull structured signal out without friction.

Clickable automations

The default company template is structured with checkbox button attributes that trigger automations directly from the page. A team member clicks a button and a workflow runs — competitor scan, contact finder, memo draft — without leaving the surface where the thinking lives. This is the headless automation layer: invisible to the user, powerful in the background, and only possible because the work surface is both structured and API-accessible.

The bar

Every company in active diligence has a live workspace page. The key questions are written out explicitly within 24 hours of the first team call. The page is updated as evidence arrives, not left to stale. The team treats the workspace as the source of truth for the state of conviction — not a partner's memory — so that AI, automation, and teammates can all see the same picture and push it forward together.