Back to all work
AI AgentsAutomation

65 AI agents, zero full-time staff — how I run a consulting firm alone

Claude + n8n + MCP + 65 specialized agents

I run a consulting business with 65 AI agents. Here's what that actually looks like on a Tuesday.

Most AI consultants talk about what AI *could* do for your business. I built 65 specialized agents that run mine. Not as a demo. Not as a proof of concept. As the actual operating system behind every client engagement, every proposal, every invoice.

What Happens Without the System

Without JARVIS, here's my week as a solo consultant managing 4-5 active projects:

  • 8 hours reviewing documents, writing proposals, checking contracts for errors
  • 5 hours processing client messages, extracting action items, updating project status
  • 6 hours on market research, competitive analysis, pricing decisions
  • 3 hours tracking finances, deadlines, cash flow across projects
  • 2 hours on repetitive coordination: scheduling, follow-ups, status updates
  • That's 24 hours/week of work that doesn't directly create value for clients.

    The 3 Agents That Changed Everything

    1. The McKinsey Panel — 15 agents that run a full market research study in a single session. Instead of two weeks of research, 15 specialized agents (market sizer, trend analyst, pricing analyst, ICP builder, channel analyst, risk assessor) ran in parallel and produced an 825-line report with validated data. Time: one afternoon. Manual equivalent: 40+ hours.

    2. The Document QA Team — Before any proposal, contract, or deliverable leaves my desk, it goes through automated review. The QA agent checks for inconsistencies in numbers, dates, and pricing. The compliance checker flags regulatory issues. The tone checker ensures professionalism. Three agents that catch what a tired human eye misses at 11 PM before a deadline.

    3. The WhatsApp Pipeline — Clients send voice notes, screenshots, and scattered messages across multiple threads. The pipeline transcribes audio, extracts text from images, identifies action items, and routes them to the right project tracker. What used to be 45 minutes of manual processing per client interaction now takes zero minutes of my time.

    The Architecture

    `

    Claude (AI backbone) + n8n (orchestration) + MCP (tool connections)

    ↓

    65 specialized agents organized by function:

    Operations (11) → scanning, deadlines, finance, contracts

    Strategy (6) → pricing, planning, mentoring

    Legal (5) → compliance, liability, review

    QA (1) → multi-mode document review

    Dev (7) → code review, testing, automation

    + 35 more across integrations, communications, career, audit

    `

    No proprietary platform. No vendor lock-in. Standard tools, custom orchestration.

    The Number That Matters

    Conservative estimate: JARVIS saves me 12-15 hours per week. At market rate, that's $62,000-78,000 per year in operational capacity redirected entirely to client work.

    But the real value isn't the hours — it's the consistency. Agents don't forget to follow up. They don't skip QA because they're tired. They don't send a proposal with a pricing error.

    The One That Failed

    Honesty moment: the first version of my automated outreach agent was terrible. It generated technically correct but painfully generic messages. I scrapped it and learned that AI works best when it augments specific, high-stakes decisions — not when it replaces human judgment on things that require context and empathy.

    That failure shaped how I build for clients: automate the 80% that's repetitive, keep humans on the 20% that requires judgment.

    What This Means For You

    You don't need 65 agents. You need the 5 that eliminate 80% of your operational overhead. The ones that run your back-office while you focus on the work that actually grows your business.

    Most businesses need 3-5 well-designed agents, not 65. Let's find yours.