Personal assistant for a consulting business

Context

The client was doing consulting work that required repeated synthesis, organizing incoming materials, drafting deliverables, and reviewing outputs before anything went back to the client. Much of that work was valuable, but the process was still slowed down by too many manual handoffs and too much founder attention going to coordination rather than final judgment.

What we built

We built a set of internal agents focused on consulting work. One agent handled inbox triage for morning briefs, meeting notes, follow-ups, and recurring internal tasks. Another agent focused on ingesting and organizing client materials in a way that matched and improved the client's existing workflow. A third agent focused on generating deliverables across the client's book of business. A fourth agent critiqued those outputs, pushing for clearer structure, better reasoning, and more useful final results before anything was surfaced to the team.

The system was not intended to replace the client's judgment. It was designed to reduce the amount of time spent getting from raw inputs to a strong draft. The final output of the workflow was still presented to the client for review and editing, which kept quality control with the human while dramatically shrinking the amount of operational lift required to get there.

Agent workflow

• Inbox triage agent. Prepared morning briefs, captured meeting notes, tracked follow-ups, and surfaced recurring internal tasks.
• Materials organization agent. Ingested documents and inputs from clients, then organized them into a structure that matched the existing consulting workflow.
• Deliverable creation agent. Produced first drafts and supporting materials for the client's book of business.
• Critique agent. Reviewed generated outputs and suggested stronger framing, clearer arguments, and more polished deliverables before handoff.
• Human review layer. Presented the result to the client for final editing and approval before anything client-facing was sent.

Results

• Average client time required for deliverables dropped by 80%.
• Monthly expense only increased by 2.6% from token usage.
• Revenue increased by 20% because the team could handle a broader scope of work.
• The client spent less time coordinating inputs and more time on final judgment, relationships, and higher-value work.

In practice, the value came from combining specialization with a clean handoff chain. Each agent handled one part of the workflow well, and the human stayed in the loop at the point where taste, context, and accountability mattered most.