How to Choose an AI Automation Agency (Without Getting Burned)

The AI automation agency market exploded in the last 18 months. Most of it is wrappers and hype. Here's how to separate the few that ship from the many that don't.

Questions to ask

• Show me a workflow you shipped in production — not a demo.
• Who owns the code and the accounts at the end of the engagement?
• What's the handover plan and runbook when something breaks at 2am?
• How do you handle evals, regressions, and prompt drift?
• What's your stance on building vs. buying for this specific workflow?

Red flags

• Pitch leads with the model name, not the business outcome.
• Locked-in proprietary "platform" you can never export from.
• No production references they'll let you talk to.
• Flat retainer with no scope and no deliverables tied to it.
• Promises of "autonomous agents that replace your team" — that's not where the tech is.

Pricing models

Fixed-scope pilots ($5k–$25k) are the cleanest way to start. Monthly retainers make sense after the first workflow ships and you want continuous iteration. Avoid % of savings — it always gets ugly to measure.

How to scope the first engagement

Pick one workflow that's high-frequency, rule-based, and currently painful. Set a baseline (time, cost, errors). Agree on the success metric in writing. Ship, measure, expand. That single discipline filters out 80% of bad agency relationships before they start.