Do You Need an AI Consultant — or an Afternoon?
Why does "get an AI consultant" sound like the obvious move?
Because it matches how every other business technology arrived. When the internet showed up you hired a web firm; when accounting got complicated you found a bookkeeper. So when AI lands on your desk, the instinct is the same: this is big and unfamiliar, therefore find an expert and pay them to handle it.
That instinct isn't wrong so much as mistimed. The hard part of most AI setups isn't technical — it's knowing your own business well enough to describe it. And nobody knows your business better than you do. The consultant would spend their first few billed hours interviewing you to learn what's already in your head. For a first setup, you can skip straight to writing it down.
What can one focused afternoon actually deliver?
More than most people expect, because the ingredients are things you already own:
- A one-page business brief. Who you are, who you serve, what you sell, what you refuse to do, and how you want to sound. This single page does more work than any clever prompt. Our guide on giving your AI business context walks through exactly what goes in it.
- A folder of real examples. Five to ten pieces of your actual best writing — proposals, replies, a listing you were proud of. Examples teach an AI your voice faster than any adjective can. More on that in training AI on your voice.
- A short rules list. The sharp, specific things: "Never quote a price without a site visit." "Answer in the language the customer wrote in."
- One small automation. Not a robot workforce — a single job you'll genuinely use, like a first-draft reply to a real email in your inbox.
By the end of the afternoon you have a working starting point you own, sitting in a folder on your own machine. That's not a toy. It's the same foundation a paid install would build for you — you just built it yourself.
So what's the catch — is an afternoon really enough?
Here's the honest limit: an afternoon gives you a real setup, not a finished one. The first version will be rough. It gets genuinely good over the following weeks, as you use it, catch what it gets wrong, and write the corrections down where the AI can read them next time. That's a habit, not a project with a finish line — and it's the part no consultant can do for you, because it depends on your daily judgment about your own work.
Anyone who tells you a single afternoon produces a polished, autonomous system is overselling. And anyone who tells you the basics are too hard to attempt without them is underselling you. The truth sits between: the start is easy, the polish takes patience, and a person always stays between the AI's draft and the customer.
When is paying a consultant actually the right call?
Real experts earn real money for real reasons. Pay for help when:
- The stakes are high. Regulated customer data, a decision that's expensive to reverse, or output that carries legal or financial weight. Get a second set of hands.
- The complexity is real. Several systems that must connect, a whole team to train, or a workflow with a dozen moving parts. That's beyond an afternoon and beyond a solo start.
- Your time is worth more elsewhere. This is the most honest reason of all. If a focused afternoon is possible but you'd genuinely rather buy it back and have someone set it up alongside you, that's a fair trade — as long as you know that's what you're paying for.
What you should not pay for is being told the fundamentals are a mystery only they can unlock. If a pitch leans on making you feel that AI is too complicated to touch, that's a sales tactic, not a service.
Consultant, installer, or afternoon — how do I tell them apart?
| The afternoon | An installer | A consultant | |
|---|---|---|---|
| Who does the work | You | Someone, with you | Someone advises, you decide |
| What you keep | Files you own | A working setup you own | A plan or strategy |
| Best when | Starting simple | You want it done together | Complex or high-stakes |
| The question to ask | Did I actually use it? | What do I own when you leave? | Is this advice, or a working thing? |
The distinction that matters most isn't the title — it's what's in your hands when they walk out the door. A slide deck ages. Files and a working tool on a machine you control keep earning. That's the same ownership test we apply to any AI decision.
Our bias, stated plainly
We're an installer, not a consultancy — we sit down with owners and set the thing up in a day, and we install Claude Code because it reads a folder of plain files directly and can act on your machine, not just chat about it. So of course we think a done-with-you install is worth it for the right business. But we'd rather tell you the truth: a lot of you can do the first version this afternoon, free, and never need us. The ones who genuinely benefit from paying are the complex cases and the folks who'd simply rather have company doing it. If that's not you, keep your money and open a folder.
Questions people ask
Can I really set up AI for my business myself?
For a first useful setup, usually yes — write a one-page description, gather five real examples, give an AI a small real task. That's an afternoon. Paid help is for complex or high-risk cases, or when you'd rather have someone do it with you.
What's the difference between a consultant and an installer?
A consultant advises and hands you a plan; an installer sets the thing up with you and leaves you a working setup you own. Ask either one what you actually keep when they leave.
When is hiring a consultant worth it?
When the stakes or complexity are real — regulated data, many staff, connected systems, a costly-to-reverse decision — or when your time is genuinely better spent elsewhere. Not for being told the basics are beyond you.
How much should a first setup cost?
Doing it yourself costs your time plus your AI tool's own subscription (check the vendor's official pricing). A done-with-you install is a paid service — ours is a one-time $1,500–$5,000 depending on scope, current as of July 2026; confirm on our site. Be wary of open-ended retainers for what's really a setup task.
What can I realistically finish in one afternoon?
A brief, a folder of examples, a rules list, an AI pointed at it, and one small automation you'll use. A real starting point — not a polished system, which comes from using and correcting it over weeks.