The Handoff Problem

Here is the scenario that breaks most teams on their first attempt at multi-person Claude workflows. Alice spends 45 minutes with Claude researching a prospect. She builds deep context — the prospect's tech stack, recent funding round, org chart, pain points, and competitive situation. She produces a solid research brief. Then she hands the brief to Bob to write the proposal.

Bob opens a new Claude session and pastes the research brief. He asks Claude to write a proposal. Claude reads the brief and produces something that looks like a proposal but is missing all of the nuance Alice built up during her 45-minute session. Why? Because the brief is a lossy compression of Alice's session. It captures the conclusions but not the reasoning. It captures the facts but not the context about why those facts matter. It captures what was found but not what was considered and rejected.

The handoff problem is not unique to AI workflows. Every team that has ever passed work from one person to another has experienced it. The meeting notes that leave out the discussion. The design brief that omits the rejected alternatives. The code PR that does not explain why this approach was chosen over the obvious alternative. What makes AI handoffs worse is that Claude starts with zero context. A human colleague inheriting work at least has organizational context — they know the client, the industry, the company's past proposals. Claude knows nothing that is not in the handoff document.

The fix is structured handoff artifacts. Not prose summaries. Not conversation exports. Structured documents that capture the five categories of context a receiving Claude session needs: findings (what was discovered), reasoning (why it matters), decisions (what was decided and what alternatives were rejected), constraints (what the receiving session must respect), and open questions (what still needs to be resolved).

1. 1. Findings What was discovered during the session. Facts, data points, observations. This is the "what" section — the raw information that the receiving session needs to work with.
2. 2. Reasoning Why the findings matter. What patterns were identified. What connections were made between data points. This is the context that turns raw facts into actionable intelligence.
3. 3. Decisions What was decided during the session and what alternatives were considered and rejected. "We decided to position against Competitor X on price, not features, because their features are stronger but their pricing is 3x ours." The receiving session needs to know what was already decided so it does not re-litigate.
4. 4. Constraints What the receiving session must respect. Budget limits, timeline requirements, tone guidelines, stakeholder preferences, legal restrictions. These are the non-negotiable boundaries the next person's Claude session must operate within.
5. 5. Open Questions What was not resolved. What needs more research. What decisions are pending stakeholder input. The receiving session should know what is settled and what is still in flux.