You pay for the training. Everyone nods. A week later, two people use the tool and the rest are back to the old way.
That is the usual story, and it is not a people problem. It is a design problem. Training that sticks is short, hands-on, built on the work your team does every day, and followed up. This post lays out what that looks like for a Tampa team, whether you do it in person, on video or both.
For the wider view, see our guide to AI implementation for small business. If you want the one-day plan itself, we wrote that up in how to train your team on AI.
Why training is the part that breaks
The numbers say owners know this. In a March 2026 Goldman Sachs 10,000 Small Businesses survey of its program alumni, 73 percent said they would benefit from more access to training and implementation resources. And only 14 percent said AI was fully embedded in their core operations (Goldman Sachs). That sample is alumni of a business education program, not every small business, so treat it as a signal, not a census.
Another look: the U.S. Chamber of Commerce Foundation and Ipsos surveyed workers at small businesses in May 2026. About half said they already use AI at work. And 19 percent said use at their company was driven mostly by employees exploring on their own, against 11 percent who said it was driven mostly by organizational guidance (U.S. Chamber of Commerce Foundation).
Put those together and you get a familiar picture. Some of your people are already using AI, without rules and without telling you. Others are afraid to touch it. Training closes that gap. It gets everyone using the same tools, the same way, with the same guardrails.
In person, on video, or both
None of these is always right. Here is how to choose.
In person
Best when:
- The team is in one place, like a shop, a clinic, a restaurant or an office.
- People need to work on their own screens with someone looking over their shoulder.
- You want to practice on your real tasks in the room: the actual inbox, the actual booking calendar, the actual quote template.
- Trust is low or fear is high. A person in the room, answering questions, calms a team faster than a recording.
Tampa is a good place for this. A trainer who can come to your shop in Ybor, your office in Westshore or your site in Brandon sees how the work really happens. In-person sessions also surface the stuff nobody puts on a form: the clipboard by the register, the group text everyone actually uses.
On video
Best when:
- The team is spread out, remote or on different shifts.
- You need to train new hires later without paying for another session.
- The topic is narrow, like "how to write a good prompt" or "what never goes into the tool."
Live video sessions let people ask questions. Recordings let people rewatch. Both are fine. Keep each session short, around an hour, and keep it to one skill.
Both
For most teams of 5 to 30, this works best:
- One in-person session to start: hands-on, built on real tasks, with the rules explained face to face.
- Short video follow-ups for specific skills, plus recordings for new hires.
- A check-in (see below) at 30 days, in person or by video.
That is also how we run it. Di-Hy's Train step is in person in Tampa Bay, on video, or both, and always comes with a written playbook.
What a session should actually cover
A good first session is not a lecture about how AI works. It is practice on tasks that cost your people time today. Roughly:
1. The two-minute basics. What an AI assistant is (a tool that writes, sorts and answers when you give it instructions), what a "prompt" is (the instruction you type), and what it is bad at (knowing facts about your business, doing math without checking, being right every time).
2. The rules. What can go in and what never does. Customer financials, health information, passwords, anything under an NDA. If you do not have a written policy yet, use our free AI use policy generator to draft one, then read is your business data safe in ChatGPT.
3. Hands-on with their own work. Each person picks one task from their week. A reply to an upset customer. A job quote. A schedule. A summary of a long email thread. They do it with the assistant, then check it, then fix what is wrong. The "fix what is wrong" part is the lesson.
4. How to check output. Read it like a new hire wrote it. Check names, numbers, dates and promises. If the tool states a fact, verify it. If it cites a source, open it.
5. What stays human. Complaints, pricing decisions, legal language, health or safety questions, anything involving a person's job. The tool drafts, a person decides.
6. Who to ask. Name one person on the team as the go-to. Give them an hour a week for the first month.
The written playbook (the part that outlasts the day)
If training lives only in people's heads, it disappears when they quit or forget. A playbook is the written version: short, plain and specific to your business. It should include:
- The approved tools and which account each person uses.
- The rules in one page: what goes in, what does not.
- Your top five tasks with the exact instruction that works for each, copy and paste ready. A saved instruction like this is called a prompt template.
- Checking steps for each task: what to look at before sending.
- What stays human.
- Who to ask when something looks wrong.
- How to report a problem or an idea.
Aim for 3 to 6 pages. If nobody reads it, it is too long. Put it somewhere people can find in one click, and add it to the new-hire checklist.
The 30-day check-in
This is where most training fails and where most of the value sits. At about 30 days, sit down with the team (30 to 45 minutes) and ask:
- Who is using it, for what, how often? Honest answers, no blame.
- What saved real time? Keep that and write it into the playbook.
- What did not work, or went wrong? Fix the instruction, or drop the use.
- Any near misses? A wrong number that almost went out, or something sensitive pasted into the wrong tool. Update the rules.
- What do people want next? One new task, not five.
Then update the playbook and set the next check for 90 days. A good result at 30 days is not "everyone loves it." It is "five people use it every week for two specific things, and nothing sensitive leaked."
Tools also change. The MIT NANDA preliminary report from 2025 found that most generative AI systems in its sample do not retain feedback or improve over time on their own (MIT NANDA via mirrored copy). Your team has to do the adjusting, and the check-in is where it happens.
Common mistakes
- Training before you pick a problem. Start with a task people hate, not a tool.
- Training everyone on everything. A front desk and an estimator need different sessions.
- No policy. People will use consumer tools with customer data if nobody says not to.
- Leadership sits it out. If the owner does not use it, the team will not.
- No follow-up. One session without a check-in is a presentation, not training.
- Punishing mistakes. You want people to say "this went wrong," not hide it.
How to start this month
- Pick two or three tasks that eat the most time across the team. Not sure which? Use the free AI time savings calculator or read find your 3 biggest time-sinks.
- Write your one-page AI rules.
- Book a hands-on session of 2 to 3 hours, in person if the team is local.
- Write the playbook the same week, while it is fresh.
- Put the 30-day check-in on the calendar before you leave the room.
Where Di-Hy fits
Di-Hy Digital is an AI implementation firm in Ybor City, Tampa, senior built and founder led. Training is step three of our four: a free audit, an implementation of about 30 days, training (in person in Tampa Bay, on video, or both, with a written playbook), and maintenance, monthly or yearly. We do not publish a price for training; it depends on the size of the team and what you are rolling out, so we scope it on a call.
Look at our services, see what we do locally on our Tampa page, or book a strategy call. If you want to start with a free look at where your team's hours go, take the AI Readiness Audit.