category: training-and-enablement
Your team is spending hours every week on tasks that should take minutes. Training them on AI should not mean signing up for another subscription, sitting through generic modules, or paying a consultant to read YouTube tutorials back to you.
You can build practical AI skills in-house, using free tools, in four weeks. No vendor lock-in. No irrelevant theory. Just your team solving problems they already have.
The reason most AI training fails
Most training programmes teach the tool first. They show your team how the software works, then leave them to figure out where it fits. That is backwards.
Your team does not need to know everything a tool can do. They need to know how to use it for three specific tasks in your business. That is a different kind of training, and it takes a fraction of the time.
Take Rachel. She works in operations at a professional services firm. Every Monday she pulls data from three systems, copies it into a spreadsheet, and formats a summary for the management team. It takes two hours. She has done it the same way for two years.
When her MD ran an AI pilot, they did not start with a course. They sat with Rachel for thirty minutes and asked: "What would you need this tool to do?" She described the Monday report. They tested it. Within a week, the report took twenty minutes. Rachel now uses that time to review the output and flag anything unusual, which is the part that actually needs her.
That is what working AI training looks like. You start with a named task and work backwards to the tool.
Pick one problem before you pick a tool
Before you train anyone on anything, choose one repetitive task that meets three conditions. It happens at least weekly. It follows the same steps every time. And it takes longer than it should.
Good candidates:
- Drafting routine client emails from a set of notes
- Extracting figures from supplier invoices and entering them into a system
- Summarising meeting notes into action points
- Pulling numbers from multiple sources into a weekly report
Bad candidates:
- Anything that requires professional judgement on every case
- Tasks that change significantly each time they are done
- Anything that requires real-time data the tool cannot access
Once you have the task, write down every step your team currently takes to complete it. This becomes your training brief. The goal is to identify which steps the AI can handle and which steps still need a person.
Free tools worth testing first
You do not need to spend anything to start. These tools are free at the level you need for initial training:
Claude (claude.ai): Strong at drafting, summarising, and extracting structured information from text. The free tier is enough for testing most business writing and data tasks.
ChatGPT (chatgpt.com): Good general-purpose tool. Useful for drafting, research, and workflow ideas. Free tier has limitations on speed and volume, but it is enough to test an approach.
Perplexity (perplexity.ai): Useful when your task involves finding and summarising external information. Cites sources, which makes it easier to verify accuracy.
Ollama (ollama.com): For businesses where data privacy matters, Ollama runs AI models locally on your own machine. Nothing leaves your system. Free to install, free to run. Requires a reasonably capable computer but no technical expertise beyond the initial setup. This is the right choice if you are working with sensitive client data, financial records, or anything covered by GDPR obligations.
Start with one tool, not all of them. Test it against your chosen task for one week before making any decisions.
A four-week training structure that actually works
Week 1: Test it yourself first
Before you involve your team, spend two hours with the tool and your chosen task. Work through it step by step. Prompt the tool. Review the output. Try a different prompt. Note what works, what does not, and what surprised you.
Write down three things: the prompt that produced the best output, the main limitation you noticed, and the step where a person still needs to check the result. This is your training foundation.
Week 2: Introduce one person
Bring in the team member who does the task most frequently. Do not give them a manual. Sit with them and walk through what you found. Let them try the tool themselves on a real example from their actual work.
The questions they ask during this session are valuable. Write them down. Their confusion points to the parts that need clearer guidance. Their discoveries often improve on what you found in week one.
Record the session if they are comfortable with it. A five-minute screen recording of a real task being done the right way is worth more than any written guide.
Week 3: Expand to the rest of the team
Run a thirty-minute session. Show the recording from week two. Let each team member try the task on their own. Keep the session focused on one task only. Resist the temptation to show everything the tool can do.
After the session, set one action: each person uses the tool for this specific task at least twice before the next meeting.
Week 4: Review and document
Run a second thirty-minute session. Ask three questions: what worked, what did not, and what would you do differently. Collect the answers and turn them into a simple one-page guide.
The guide covers: what the tool does, what task it handles, the prompt that works best, what to check in the output, and what to do if the result looks wrong. Keep it short. One page. Plain English.
This document goes into a shared folder alongside any recordings. It is your internal training resource. Update it quarterly.
Make it safe to get it wrong
The biggest barrier to AI adoption in small teams is not skill. It is fear. People worry they will produce something incorrect and embarrass themselves, or that they will break a process that already works.
Address this directly. Tell your team explicitly: the first few weeks are learning time. Getting a bad output is useful. It tells us where the tool has limits. There is no wrong attempt.
Create a shared testing space, a folder or document where people can paste prompts and outputs without consequence. Encourage sharing when something goes wrong as much as when something works.
One MD we work with introduced a ten-minute slot at the end of their Monday team meeting called "what I tried." Each person shares one thing they tested with AI that week, whether it worked or not. Within a month, the team was generating their own automation ideas rather than waiting to be trained.
That shift, from passive recipients to active contributors, is when adoption becomes permanent.
When to go deeper
After four weeks with one task, you will know whether the tool is worth expanding. Signs it is working: the task takes less time, the output is consistent, and your team is using it without prompting.
At that point, you have two options. Add a second task and run the same four-week process. Or go deeper on the first task by integrating the tool into your actual workflow so it runs automatically rather than being triggered manually each time.
The second option is where real time savings compound. A task your team does manually with AI support saves time every cycle. A task that runs automatically saves the same time every cycle without anyone needing to think about it.
That is a different conversation and a slightly different kind of build. But it starts here, with one task, four weeks, and free tools your team already have access to.
What to Do Next
Pick one task this week. Write down every step. Spend two hours testing it with a free tool before you involve anyone else. That two hours is your training plan.
If you want a second opinion on whether your chosen task is a good candidate for automation, book a free 30-minute process review at smarterworker.co.uk. We will tell you what is realistic, what is not, and whether it is worth building further.
