Guide · Hiring · AI automation

How to hire an AI automation consultant

Most AI automation projects fail before the first workflow is built — not because the tools are wrong, but because the wrong problem got automated. Here's how to evaluate a consultant or agency so you end up with a system that earns its place, not a faster version of something broken.

Process matters more than tools

Zapier, Make, n8n, and every LLM API are commodities. A consultant who leads with tools ("I'm a certified n8n expert") is optimising for the wrong variable. What actually decides whether an automation survives contact with your business is the method used before a single node is dropped on a canvas.

The method I use — the First Principle approach — runs in a strict order: Question, Delete, Simplify, Accelerate, Automate. Automation is the last step, not the first. Any consultant who skips the first three phases is selling you speed for a process that shouldn't exist in its current form.

Questions to ask on the first call

  1. "Walk me through a project you killed." A good consultant has told a client "don't build this" at least once. If every project they mention shipped, they're order-takers, not specialists.
  2. "How do you decide what not to automate?" Listen for a repeatable framework — cost of manual work, frequency, error rate, edge-case density. Vague answers mean they'll automate whatever you point at.
  3. "What's the maintenance model?" Every automation breaks. Ask who fixes it, how it's monitored, and whether your team can read the workflow without them.
  4. "Show me a workflow you built end-to-end." Case studies with real screenshots, tool choices, and the trade-offs behind them beat a portfolio of logos.
  5. "What does the handover look like?" You want documentation, a video walkthrough, and access to every account. No handover plan is a retention trap.

Red flags

  • They quote before they've mapped the process. A fixed price on day one is a fixed price on a fantasy.
  • They lead with AI. Not every automation needs an LLM. If the pitch is "AI agent for X" before they've seen your data, they're selling a hammer.
  • They use only one tool. Zapier maximalists, n8n purists, and Make-only shops all pick the tool that matches their skillset, not your constraints.
  • No talk of failure modes. Ask what happens when the API rate-limits, when the LLM hallucinates, when the webhook fires twice. Silence means they'll find out in production.
  • They won't share a scoped pilot. A two-week paid pilot on one workflow is the cheapest way to test fit. Consultants who refuse it are protecting a retainer, not you.

Consultant vs. agency vs. freelancer

An AI automation agency is right when you have 5+ workflows, need parallel delivery, and want account management. Expect $8k–$30k/month retainers and a project manager between you and the builder.

An independent AI automation consultant is right when you want the person doing the diagnosis to also do the build. Faster loops, no telephone game, usually $150–$300/hr or fixed-scope projects.

A freelancer on Upwork or Fiverr is right for a single, well-defined workflow you've already scoped yourself. Bad fit when you need someone to tell you what to build.

What a good engagement looks like

  1. Audit (week 1). The consultant sits with your team, watches the actual work, and writes back what should be deleted, simplified, or left alone.
  2. Pilot (weeks 2–3). One workflow, shipped end to end, with metrics on time saved and error rate before/after.
  3. Rollout (month 2+). Prioritised backlog, weekly demos, documentation as it's built — not at the end.
  4. Handover. Video walkthroughs, an ownership matrix, and a support SLA if you want ongoing coverage.

Want a First Principle audit?

I run a free 30-minute audit where we look at one workflow you do by hand, strip the unnecessary, and map what a leaner version looks like. You'll leave knowing exactly what to cut, what to keep, and what the manual version is costing you every week.

★ Book review