Writing

    What a Fractional CTO Actually Does for Agentic AI

    The engagement model, deliverables, and when SMBs should hire fractional technical leadership for AI systems.

    Fractional CTO4 min read

    A full-time CTO costs $250k before benefits. Most SMBs don't have that problem yet. What they have is one person with enough scar tissue to design an agentic AI architecture, build the first version, and leave behind something the team can run without a specialist on call. Every engagement I've run has gone the same way: one workflow at a time, with someone accountable for whether it works.

    What the engagement covers

    Four areas, usually in this order:

    • Strategy: which of your workflows are agent-shaped, and which are just automation wearing a costume
    • Architecture: MCP servers against the tools you already pay for, orchestration, which model goes where and why
    • Implementation: hands-on in Claude Code, or steering your team while they build it
    • Operations: the part everyone underestimates. Structured logging, cost ceilings, and a plan for the day the agent is confidently wrong

    When it makes sense

    • You've proven AI can help, but nobody owns the technical roadmap
    • Your team shipped a demo agent that nobody trusts enough to leave running
    • You need MCP integrations with the tools you already pay for
    • You're deciding build vs. buy and want an opinion from someone with no license to sell you

    When it doesn't

    If you need someone in standups five days a week, hire full-time. If you want a ChatGPT wrapper with your logo on it, hire a contractor for a week. The fractional model earns its price on judgment and delivery rather than seat time, so it only works when the decisions are the expensive part.

    What it costs

    Fractional CTO runs $4,000–8,000 a month, depending on depth and on how much I'm building versus advising. Most engagements start smaller: a single workflow audit. One process, one agent, one number that either moves or doesn't.

    If that range isn't what you expected, tell me what you were expecting and I'll tell you whether it's doable or whether you should be talking to someone else. Some of these conversations end with me pointing you at a SaaS tool that already does the job. That's a fine outcome.