DIY, part-time admin, or consultant: who should do your AI work
Three real paths for getting AI going in a small business, what each involves week to week, and the way each one usually goes wrong.
- The task, not the tool, decides who should own the work
- DIY wins when the owner likes tinkering and the task changes often
- A part-time admin wins when the tasks repeat the same way every week
- A consultant or concierge wins when the owner wants advice and accountability
There is a moment most owners hit by now. A task repeats every week, maybe quote replies or meeting notes, and a chat window is open on the screen. The tool works. The question is who owns this work: you, someone on your team, or someone you bring in. The honest answer depends on the task and on you, and each path has a different way it fails.
The answer depends on the work
Three paths cover almost every situation. Do it yourself with a tool like ChatGPT or Claude. Hand it to a part-time admin who already works in the business. Bring in a consultant or a service like AI Concierge, where a real person learns the business and works alongside the owner. The choice turns on two things: how repetitive the task is, and whether the owner wants to learn the tool or wants someone else to carry it.
Path 1: Do it yourself
This path suits an owner who likes tinkering. You open the tool, try the task, keep what works and drop what does not. Nothing here requires a budget beyond the subscription itself, and you learn the limits of the tool firsthand, which matters later when you hire help. It also builds judgment: once you have drafted a few replies yourself, you know what a good result looks like.
The failure mode is drift. The owner tries it for two weeks, gets busy, and the tool sits unused while the task goes back to the old way. DIY also caps out when the task needs a setup bigger than a prompt, such as a connection into a CRM or a system the whole team uses. If no one is accountable, the work quietly stops.
Path 2: Hand it to a part-time admin
This path wins when the task repeats the same way every week. Quote follow-up, intake, meeting notes, cleaning up a spreadsheet. A part-time admin or an existing team member can run those with AI as part of the job, once someone shows them the routine. The owner stays out of the tool and the work happens without their screen time. This is the cheapest path in owner attention, which is usually the scarcest thing in the building.
The failure mode is a broken routine. If the admin invents their own prompts with no ground rules, the output drifts, or something sensitive goes into a tool it should not. And when that person quits or changes roles, the routine walks out the door with them unless it was written down. Training helps here: Ohio AI Lab runs hands-on sessions, 90 minutes and up, built on the team's approved real tasks, and people leave with take-home routines and safe-use ground rules. Sessions cover one owner, a small group or a whole team.
Path 3: Bring in a consultant or concierge
This path suits the owner who wants advice and accountability, not just setup. A consultant looks at the business, picks the tasks worth doing first, and stays while the plan happens. Ohio AI Lab's own version is AI Concierge: a real person, never a chatbot, who learns the business, holds two 1-hour working calls a month on real tasks, and answers voice messages and email between calls with no limit on questions. They help with small setups, give straight advice when the owner is deciding, and keep the plan moving.
The distinction that matters: a concierge is a person, and the work stays in the owner's accounts. Ohio AI Lab states that everything built lives in the client's own accounts and is theirs to keep. In the first month, the fit call and intake happen in week 1 with a 60-minute interview, a personal AI context file goes into the owner's own AI account in week 2, and weeks 3 and 4 are practice and fixing friction. The failure mode for this path is different from the others: it fails when the owner treats it as a class to sit through rather than working sessions. The calls only pay off if the owner brings the real task.
Bigger builds sit outside all three paths. API connections, CRM integrations, hosted agents or dashboards get scoped and quoted before anything starts, so the owner knows what a build costs before it begins.
Pick your path this week
- 01List two repeating tasks
Write down two tasks that show up in your week the same way every time.
- 02Try the task yourself
Open ChatGPT or Claude and run the task once by hand to see what a good result looks like.
- 03Decide who owns it
If you enjoyed it and the task changes often, keep it. If it repeats the same way, hand it to a team member or part-time admin with a written routine.
- 04Get help where you need judgment
If you want advice on which tasks matter and someone to keep you moving, book a free 25-minute fit call on Zoom.
- 05Write the routine down
Whatever the path, record the steps and the prompt so the work survives a staff change.
Owners usually need one AI subscription, such as ChatGPT or Claude, paid directly to the provider. Any other cost gets named before anything is bought. Ohio AI Lab does not guarantee savings or revenue; it picks useful work, measures what changed and reports honestly. That stance is worth demanding from whichever path you pick.
