5 min read
57% of U.S. small businesses have now invested in AI, up from 36% in 2023, yet only about 30% of their employees use it every day. Those figures, from a Business.com study cited in a CIO.com analysis published October 2, describe a very expensive kind of shelfware. Small businesses bought the software. Almost nobody showed them how to make it part of the job.
The same piece cites Bluevine’s 2026 small business report: 74% of owners say they use or are testing AI. So adoption is not the problem anymore. Setup is. And that gap has created a fast-growing market of people who will happily charge you to close it.
My position is simple, and it will annoy a few consultants: most solo owners should not pay for an “AI transformation.” They should pay, if at all, for a few hours of very specific setup, and only after trying to do it themselves. Here is why, and how to tell good help from expensive theater.
The gap is real, but it is smaller than the pitch decks say
The CIO article frames small business AI adoption as a ladder: shadow AI (people using whatever free tool they like), then licensed but unstructured, then structured enablement, then fully governed and embedded. That is a useful map for a 40 person company with a team to coordinate.
For a business of one, most of that ladder collapses. You do not have rogue employees pasting customer data into random tools. You do not need a governance committee. Your “structured enablement” is you, writing down three workflows and actually using them. The distance between “I pay for ChatGPT” and “AI handles real work every week” is, for most solo owners, a long weekend of focused setup, not a six month engagement.
We have argued before that most solo owners use AI but very few run on it. The question this piece answers is narrower: when you decide to close that gap, should you hire someone to do it?
Why I am skeptical of paid AI setup for solo businesses
The tools now ship with the setup built in
A year ago, connecting an AI assistant to your accounting and CRM meant fiddling with automation platforms. Today, vendors are packaging the setup themselves. Anthropic’s Claude for Small Business ships with ready-to-run workflows for payroll planning, month-end close and campaign management, plus a free on-demand AI Fluency for Small Business course. When the vendor gives away the playbook, paying a third party to rewrite it needs a strong justification.
Nobody knows your business better than you
The valuable part of AI setup is not clicking “connect.” It is deciding which jobs to hand over, writing down how you want them done, and noticing when the output is wrong. A consultant has to learn all of that from you anyway, and you pay for the lesson.
Expertise in this field is still very thin
The tools changed dramatically in the past six months alone. Someone who became an “AI consultant” last year may be selling you yesterday’s workaround. Certifications in this space are new and uneven, and a confident demo is not evidence of results.
Retainers create dependence
The worst outcome is not wasted money. It is a setup you do not understand, built in someone else’s accounts, that breaks the day the consultant stops answering emails. If you cannot explain how your own AI workflows work, you do not own them.
When paying for help genuinely makes sense
Skepticism is not a blanket no. Help is worth buying when:
- The cost of a mistake is high. Anything touching payroll, taxes, client confidential data or regulated information deserves a second set of expert eyes. Our piece on mistakes solo owners make with business data in AI shows how easy it is to get this wrong.
- You have tried and stalled twice. If you have started the same setup two times and abandoned it, a few hours of guided help is cheaper than a third false start.
- The job involves custom connections your tools do not offer out of the box, such as an old booking system or a niche industry app.
- Your time is genuinely worth more than the fee. If one billable day covers the whole engagement and frees up five days a month afterward, the math works.
How to hire AI help without getting fleeced
If you do hire, structure the deal so the risk sits with them, not you:
- Buy an outcome, not hours. “Set up an assistant that drafts my weekly invoice reminders from QuickBooks and I approve them” is a scope. “AI strategy and implementation” is not.
- Insist everything lives in your accounts. Your logins, your subscriptions, your documents. Never let the setup live in theirs.
- Demand a written runbook. A one page document explaining each workflow, what it connects to, and how to turn it off. If they cannot write it, they do not understand it either.
- Include a handover session. The last hour should be you doing the workflow while they watch, not the other way around.
- Ask for one recent example with a client of your size. Enterprise case studies do not count for a business of one.
- Avoid long retainers at the start. A fixed small project first. Monthly support only if the first project delivered.
Try the free version of this first
Before you pay anyone, give yourself one focused session. Pick the single task you most want off your plate. Write half a page on how you do it today. Ask your AI assistant to turn that into a step by step workflow and test it on last week’s real work. If you are new to this, our guide to the first three jobs to hand AI gives you a starting list.
If that session works, you just saved a consulting fee. If it does not, you will walk into any paid engagement knowing exactly what you need, which is the best protection against paying for the wrong thing.
Have you paid someone to set up AI for your business? Was it worth it? I would genuinely like to hear the good and the bad in the comments.
