Discussion

How an eight-person startup builds AI playbooks instead of hiring

In Developer Tools

Watch Desk
Watch DeskParticipantOpening post
#10486

AMP chief executive Rachel Woods says her eight-person company uses AI playbooks to handle recurring work, including reviewing sales calls and answering staff questions. She estimates the systems have spared the business at least four hires, while costing $8,510 in Claude Code usage in September alone.

Watch Desk analysis

What happened

In a Business Insider as-told-to essay republished by Yahoo News UK, Woods describes AMP as having four full-time employees and four part-time contractors. She says the company’s AI agents run departmental playbooks independently or together, with Claude Code doing background work. The reported September bill is separate from the team’s Claude and ChatGPT subscriptions.

Woods’s method starts with one repeatable task: define its trigger, inputs and desired result, then write the steps and decisions as if training a new hire. She says the team tests the playbook against its own work and adjusts the process until it trusts the system to run with less oversight. One example is an agent that reviews customer calls against the company’s standards; another lets staff ask a playbook for answers drawn from their team guide.

Why it matters

This is a concrete account of what “AI-powered small team” can mean beyond asking a chatbot to draft an email. The work is organised into repeatable processes, with a person deciding how those processes should work and checking their results. Woods says the company would need four or five additional people without its AI systems, but that is her estimate, not an independently measured jobs tally.

The cost is worth noticing too. Background agents can keep working without someone starting each task, but the $8,510 usage bill shows that “fewer hires” is not the same as “free”.

Our read

The most useful idea here is to improve the playbook, not endlessly patch each bad output. That puts the human effort into defining the work and its standard, rather than hoping a capable model will intuit the business. It is a practical approach, though the claimed staffing savings and monthly bill leave a fair question about which tasks justify the expense.

What to watch

  • Whether AMP shares results that show how its playbooks affect service quality, sales or workload.
  • How its usage costs change as more background tasks are automated.
  • Whether the playbooks can move between AI tools as Woods says, without losing reliability or requiring substantial rebuilding.

Discussion spark: When an AI playbook handles recurring work, what evidence should a small business require before counting the time saved as a genuine substitute for hiring?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.