A free 15-minute Zoom call. No cost, no pressure.What usually goes wrong
The pattern is pretty consistent. Someone signs up on a personal email, so the account belongs to a person instead of the business. A few people get added, nobody else does. The tool is never connected to email, files or the calendar, so it only knows what gets pasted into it. No one was shown how to prompt it well, so the first few answers were flat and people gave up.
Then there are the quieter problems. Staff start using their own personal accounts for work, and client details end up in places you cannot see. Someone sets up a shared login and now the password is in a group text. A few automations or saved prompts get built by one enthusiastic person and only that person understands them.
None of this is a disaster. But it does mean the AI is costing you money while doing very little, and the team has already decided it is not that useful. That last part is the real thing to fix.
First we take stock, then we clean up
We start by finding out what actually exists. Which accounts, on whose email, who has access, what got connected, what prompts or automations were built, and where personal accounts are being used for work.
From there the cleanup is straightforward:
- Move the account onto a business email so it belongs to the company, not a person.
- Get the right people on and the wrong people off, with proper roles instead of one shared password.
- Connect it to email, shared files and the tools your team already uses, so it works from real context.
- Pull the useful prompts and automations into one place the team can find.
- Write down the simple rules: what goes in, what does not, and who to ask.
You keep your own account with Claude, ChatGPT or Gemini and pay them directly. We do not resell it and we do not move you somewhere new for the sake of it. If the account you already have is the right one, we stay on it.
Then we train people properly
Often a failed setup was never really a setup problem. It was a training problem. Someone was told "here is AI, use it" and left to figure it out.
So the first training session is not a tour of buttons. It is your actual work. We sit with each person and run their real tasks through it: the quote they are writing, the email they keep putting off, the notes from a site visit they need to turn into something readable. People leave with two or three things they will genuinely use that week.
We also cover what not to do. Do not paste a full client file into a public chatbot if the office rules say otherwise. Do not trust a number without checking it. Do not let the AI send anything to a client without a human reading it first. Simple habits, said out loud, once.
What a normal week looks like after
The point is that AI becomes boring. It sits in the background and helps.
A normal week might look like this. Someone drafts a follow-up email in a couple of minutes and edits it before sending. Site notes get tidied into a summary that goes into the job file. A long thread gets shortened so you can read it on your phone. A first draft of a quote or a job post gets written and then corrected. The phone messages that came in overnight get sorted.
And when something breaks, a login stops working or someone asks "can AI do this?", there is a real person to text. That is the part most do it yourself setups never had. You are not back on a help forum at nine at night.
Sensitive files: law, medical and finance offices
If you are a law firm, a medical practice or a financial advisor, the cleanup has an extra step. Client files should not be going into a public chatbot by default.
There are two sensible routes. Zero data retention, where the AI provider does not keep your prompts or files after it answers. That is not the same thing as "not used for training", and it is worth understanding the difference before you rely on it. The other route is private AI on a machine inside your office, so files never leave the building at all.
We can talk through both. Your office should still check its own rules, and we will not tell you a product is compliant on your behalf.
What to watch out for while you fix it
A few traps worth naming.
- Do not delete the old account before you know what is saved in it. Prompts and history are easy to lose.
- Do not let everyone keep a personal account "just in case". That is how client details end up in two places.
- Do not buy more tools to solve a training problem. One tool, used well, beats three used badly.
- Do not turn on automations before the basics work. An automation running on a half set up account just makes mistakes faster.
- Do not skip the written rules. A one page company AI policy saves a lot of guessing.
If you are unsure whether to keep what you built or start cleaner, that is a normal question and we will give you a straight answer.
Start with the free AI game plan
The easiest way in is the free AI game plan. It is a free 15 minute Zoom call, no cost and no pressure. We look at what you already have, what went wrong, and say plainly what needs cleaning up and which level fits. Nothing gets set up on that call.
Every client starts with Setup, which is where the cleanup happens. From there you can move to Support for a real person to text, Coaching if you want help working a goal, or Done for you if you would rather we build and run it. Move up or down whenever you like, and cancel anytime and keep your AI account.
You will get your price in writing after the free AI game plan, before anything starts. If your setup went sideways, book the free AI game plan and let us put it right.