Short, sourced answers to the questions managers ask before they train a team to use AI at work: cost, time, coverage, data safety, pilots and measurement.
AI training costs per employee are the subscription or course price plus the paid time each person spends learning and the time a manager spends reviewing their work, and for a small team the time is often the larger share.
Read the guideA useful first round of AI training for employees takes a few hours of learning spread over two to six weeks, because short sessions with practice on real tasks stick better than a single long workshop, and skills keep developing after that.
Read the guideWhen few people use Copilot or ChatGPT after rollout, find out why before buying more training, then give each role one approved task it does every week, show people how to do it well and check the work together.
Read the guideUsing ChatGPT with company data can be acceptable when the organisation has chosen a plan whose data terms it has read and approved, and staff are trained on what may and may not be entered, but pasting confidential or personal data into a personal account is not safe practice.
Read the guideMeasure whether AI training worked by comparing a work sample before and after against the same stated standard, checking whether people now use AI on real tasks, and only then estimating business results you can actually observe.
Read the guideAI training for employees should cover how the approved tools work and fail, what information may be used, how to brief a tool well, how to check output before relying on it, and practice on tasks from each person's own role.
Read the guideA small company should run an AI training pilot with three to eight people who share a repeatable task, for about four weeks, with one approved tool, a written rule on data, a baseline work sample and an agreed decision at the end.
Read the guideEvery answer opens with a one-sentence reply to the question in its heading, so you can stop reading there if that is all you need. The sections that follow restate the answer, give an at-a-glance list, then work through the detail with a fictional example where one helps.
Each page lists the public sources we opened and read on the date shown, such as NIST, the European Commission, the UK NCSC and vendor help centres. We have no proprietary survey data, so you will not find invented percentages here. Where a claim needs a number we could not verify, we describe it without one.
Tosheo.ai is a planned self-serve AI training product for teams of 5 to 200 people, and it is still being built. Each answer has a section on where Tosheo.ai fits and a section on where it does not, because a free vendor resource, a policy document or a conversation with a specialist is sometimes the better next step.
Product details on these pages describe planned launch scope, not a service you can inspect today. Legal and regulatory points are general information, not legal advice.
Explore the planned role-specific lessons, practical tasks and manager view.
