Consulting Firm Triples Research Output With Multi-Agent Claude Workflows

A boutique strategy consulting firm — 14 consultants, no dedicated engineering team — wanted to deliver more work without hiring. Their core product was research-intensive: competitive analysis, market entry assessments, technical due diligence. Each engagement required 40-80 hours of research and synthesis before any actual strategic thinking could happen.

Their consultants were spending most of their time doing work that felt like it could be automated.

The Problem

They tried a simple Claude setup: give consultants access to Claude, let them use it for research. Adoption was uneven. Some consultants used it for everything; others didn’t trust the output. Quality varied. The firm had no visibility into how Claude was being used across engagements, whether outputs were consistent, or what the actual time savings were.

More importantly, the client isolation problem was real. Consultants working on multiple engagements might inadvertently mix research from one client into the context of another. With sensitive competitive information flowing through the research process, this was a genuine risk.

They needed a structured system, not individual consultants with ad-hoc Claude access.

The Solution

Sentrely with per-client project isolation. Every client engagement is a separate project in the gateway with completely separate agent identities, separate audit trails, and separate token budgets.

The firm built a three-agent research pipeline:

Research Agent: Given a research brief, the agent systematically gathers information from specified sources. Policy scopes it to approved research databases and public web access. No access to other client project data.

Analysis Agent: Takes the research agent’s outputs and structures them: themes, evidence, counterarguments, gaps. Hands off to the writing agent.

Writing Agent: Produces first drafts of deliverable sections in the firm’s standard format. Access scoped to the current engagement’s document storage only.

Human consultants review and refine the output, add the strategic layer that requires their expertise, and handle client interaction. The agents handle the 60-70% of work that was previously pure research and structuring effort.

A2A messaging through the gateway means the pipeline runs with minimal coordination overhead. The research agent completes a section and the analysis agent picks it up automatically. The consultant monitors progress through the dashboard and steps in where judgment is needed.

The Results

3x deliverable output in the first quarter. The firm took on 40% more engagements without adding headcount.

Report generation from 3 days to 4 hours. The research and first-draft phases, which previously consumed most of the engagement time, are now largely automated. Consultants spend their time on the strategic work that justifies their rates.

Zero client data cross-contamination incidents. Project isolation in the gateway makes this structurally impossible — agents on Engagement A literally cannot access data from Engagement B.

+40 NPS points from clients in the quarter post-deployment. Faster turnaround and more comprehensive research were the primary drivers.

Managing Partner: “We were skeptical this would work for strategy consulting — our clients pay for judgment, not research. But what we found is that automating the research actually improves the quality of the judgment. When the research is done in 4 hours instead of 3 days, we have more time to think about what it means.”