AI training for your team, on the work it already does
AI training for a business teaches your people to use assistants such as ChatGPT, Claude and Copilot safely, and to build small tools on your own records. For a smaller manufacturer or distributor, the useful version runs on the team’s real work, in French or English, and the goal is for your team to build something small that works on a job it already does. The same term also describes training AI models on data. This page is about training people.
What AI training means for a business
AI training is the time your team spends learning to do its own work with AI tools. At a plant or a distributor, the estimator, the buyer, the person entering orders and the owner learn what an assistant does well with their files, and where it goes wrong.
Training people and training models
The phrase has two meanings. Model training feeds large amounts of data to a model until it learns a task, and it happens at AI labs and at companies that build models. Training people teaches your team to use models that already exist, on work it already does. This page covers training people.
AI literacy, upskilling and enablement
AI literacy is the baseline: knowing what an assistant is, what it does well, and when its answer needs checking. AI upskilling trains the people you already employ to do more of their current job with AI. AI enablement is the setup around the training, such as accounts, rules about which data goes where, and tools connected to company systems.
What useful AI training for employees looks like
ThriveAI runs AI training for employees on the jobs they already do. At a smaller manufacturer or distributor, corporate AI training with ThriveAI means working sessions on site, on your own documents, with the people who handle them every day.
It starts from your own documents and systems
The sessions start from the paperwork your team handles every week: an emailed purchase order, a supplier quote in PDF, a drawing, an ERP export, or the spreadsheet someone updates by hand. On familiar documents, each person can tell at a glance when the assistant gets something wrong.
Before anyone uploads a company file, check the plan behind each assistant, as the section on where your data goes explains.
The goal is a working result on a real job
The goal of the sessions is for the people who attend to build something that works on a real job. A buyer might build a prompt that reads a supplier quote and lists the lines whose price moved. Someone in order entry might build a tool that turns an emailed purchase order into a draft order.
Each result is kept small, so the person who built it can use it and change it when the job changes. Anything going to a customer or a supplier waits as a draft until someone on your team approves it, the same rule every ThriveAI system follows.
In French or English
Sessions run in French or English, and the examples use your own documents in whichever language your team writes them.
Who should attend
Invite the people who do the work, and their manager. The estimator, the buyer, the person entering orders and the office lead get the most from hands-on sessions, because the examples are their own jobs.
AI for managers and AI training for executives cover a different question: which jobs to hand to AI first, what the rules are for company data, and who approves what the tools draft. Those decisions belong to the owner and managers. A written AI policy records them, so staff know which accounts to use and which files may go where. For what a policy covers, see how to write an AI policy for your company.
What your team learns
The content follows the jobs you choose and draws on the topics below.
Using ChatGPT, Claude and Copilot safely
Your team learns what each assistant does well, how to check an answer before using it, and which accounts are approved for company files. An assistant writes a confident answer whether or not it has the facts, so the training includes checking the output against the source document.
Prompting on real tasks
Prompting is writing the instruction you give an assistant. The habits that matter are plain: give it the document instead of describing it, say what the output should look like, and ask where each figure came from. Your team writes prompts for its own tasks and keeps the ones that work in a shared file.
Connecting company data to an assistant
An assistant that cannot see your records answers from general knowledge. Connecting it to an ERP export, the shared drive or the mailbox lets it answer from your own data and name the record behind each answer. Your team learns what a connection can read, how to limit it to one job, and how to keep it read-only while the tool is tested. For the security side of connecting AI to company systems, see AI security for your business.
Building simple agents and automations
An agent is an assistant that carries out a sequence of steps, such as reading a new email, finding the matching part in a price file and drafting a reply. An automation is a repeated task that runs by itself when something happens. Your team builds small ones on real jobs, and every draft waits for a person to approve it before it goes out.
Knowing where your data goes
The plan behind an assistant decides what happens to what your staff type into it. In the table, the API is the paid connection developers use to build their own tools on a vendor's models. The table uses each vendor's own pages, checked September 25, 2026.
| Plan | Used to train the vendor's models | Where data is kept and processed |
|---|---|---|
| ChatGPT for individuals | OpenAI may use your content to train its models, and you can opt out. | Only eligible API customers and new Enterprise or Edu customers can choose the country where their stored data is kept. |
| ChatGPT Business, Enterprise and the API | Not by default. For most API features, OpenAI may keep inputs and outputs for up to 30 days. You can ask for zero data retention, an agreement that the vendor does not keep your requests or answers after it replies, on eligible API features if you have a qualifying use case, and OpenAI must approve it first. | Eligible API customers and new Enterprise or Edu customers can have their stored data kept in Canada. For ChatGPT, keeping the processing itself in one region, which OpenAI calls inference residency, is offered only in Europe, the United States and the United Arab Emirates. |
| Claude Free, Pro and Max | Anthropic trains on your data when the model improvement setting is on. | With that setting on, Anthropic may keep your chats and coding sessions for up to 5 years, in de-identified form. For Claude Code with the setting off, it keeps data for 30 days. |
| Claude Team, Enterprise and the API | Not by default. | On the API, you can have requests processed in the United States only, or in any region where Anthropic has capacity. Those are the only two choices, and stored API data is kept in the United States. |
| Microsoft Copilot with a work account | Prompts, responses and the emails, chats and documents Copilot reads in your Microsoft 365 account are not used to train the general models behind Copilot. | Outside the EU, queries may be processed in the US, the EU or other regions. In an April 2026 update, Microsoft said it expects to offer local inferencing for Copilot in Canada in 2027. Inferencing is the step where the model reads your request and writes the answer. |
A staff member who pastes a customer drawing into a personal ChatGPT or Claude account is working under that account's consumer rules. When staff use AI tools the company has not approved, the practice is often called shadow AI.
None of the plans in the table lets you choose Canada as the place where the model reads your request and writes the answer. OpenAI's API documentation lists Canada for stored data only.
The platform ThriveAI builds on is designed to keep your company's data on its own server in Canada. You choose the model that reads it: one that runs on that server, or a hosted model (one that a vendor such as Anthropic or OpenAI runs on its own computers) under a written zero data retention agreement. A hosted model may process requests outside Canada.
Anthropic keeps prompts and answers from its Covered Models, which include Claude Fable 5 and 5.1, for at least 30 days. Zero data retention is not available for them, apart from a limited-time arrangement that lets some eligible organizations use it with Fable 5 and 5.1. The contract for your system names the model and its service tier, so you know which rule applies.
For assistants that keep company data under your control, with zero data retention and a model you choose, see ChatGPT alternatives for business.
Start with one job
Tell Derik which job takes the most hours in your week and which assistants your team uses today. He will tell you what your team could build on it.
Start a conversationChatGPT, Claude and Copilot training
Training starts from the assistant your company already pays for, or the one your staff already use on their own accounts.
ChatGPT training
ChatGPT training covers the business plans, where OpenAI says it does not train on your data by default, and the habits that make each answer checkable. Staff who use a personal account learn which work belongs on the company plan.
Claude training
Claude AI training covers the Claude apps for everyday writing and analysis, and Claude Code for people who want to build small tools. Fahd Alhattab, founder of Unicorn Labs, learned to build in Claude Code alongside Derik while they joined his company's onboarding tools into one process. If you are weighing a build done entirely in-house, see ThriveAI compared with building it yourself with Claude Code.
Copilot training
Microsoft Copilot training fits companies that already run on Microsoft 365. Microsoft says Copilot draws on the emails, chats and documents a person has permission to access, and surfaces only data that person can at least view. That puts folder permissions inside the training, because a shared folder that every employee can open is also open to each employee's Copilot.
AI workshops: how ThriveAI runs training
ThriveAI is an AI engineering company in Ottawa. It builds private AI systems for manufacturers and distributors in Ontario and Quebec, on their own data, and works hands on with your team. An AI workshop is one working session, where people bring a real task and build on it. ThriveAI’s AI training runs as working sessions on site, on your documents.
Derik leads every project
Derik Lawlis, ThriveAI’s founder, leads every ThriveAI project and stays close to the build. The sessions are hands on: your people work on their own tasks, so they learn how to change what they build.
Where sessions run and how they are planned
Sessions run on site at your plant or office, in French or English. Each engagement is planned and quoted on its own, around the jobs you choose and the people who take part.
What to gather before you start
- Pick one or two jobs that take the most hours, and gather a handful of real examples of each.
- List the assistants your team already uses, and the plan each one is on.
- If a job depends on the ERP, ask whoever runs it whether it can export the records that job needs.
Training and a stage one prototype
Stage one of a ThriveAI project is a working prototype for the one job that costs you the most, at a fixed price, in weeks. It runs on your own systems, and if you stop there, you keep it. The people who will use the tool shape it while it gets built, so it fits how your team already works. Every project is planned and quoted on its own, and no rates are published.
Build while you learn
Start small. Pick one job, build the smallest tool that helps with it, and put it in front of the people who do that job. Their corrections become the specification for the next version.
He helped me map the flow visually on a whiteboard, then teaching me alongside him in Claude Code, we integrated all those tools into one process.
Fahd Alhattab, Founder, Unicorn Labs.
Before you recruit a specialist, train the people you have on one real job and see how far they get. For the cost of a hire, see ThriveAI compared with hiring an in-house AI engineer.
In a plant, good first jobs sit where work arrives as email and leaves as a typed entry, such as quoting, purchasing and order entry. For what AI drafts at six desks, see AI for manufacturing, module by module.
AI training funding in Ontario and Quebec
The programs below cover employer training in general, and none of them is specific to AI. Each one limits who may deliver the training, so read that rule before you pick a trainer. Every figure below comes from the government's own page, checked September 25, 2026. This page is general information and is not legal or tax advice, so confirm your eligibility with the program before you plan around it.
Scale AI, Canada's AI cluster, runs a Custom Artificial Intelligence Training Program that gives organizations grants to train their teams on custom AI programs. Its page says it is not currently accepting new applications.
Ontario Job Grant
The Ontario Job Grant (OJG) helps employers train their employees. It covers eligible training costs up to $10,000 per trainee. Employers with 100 or more employees pay half of the training costs, and employers with fewer than 100 pay at least one-sixth of each trainee's costs.
Ontario accepts applications year-round for training 25 or fewer participants. The employer needs a physical location in Ontario, in business there for at least one year, and the training must be completed within 52 weeks of its start.
Check the provider rule first. The provider must keep a physical presence in Ontario and have been “in the business of providing the applicable training for at least two years.” It must also be one of the listed types: a publicly assisted college, university or Indigenous Institute, a District School Board, a career college or vocational program provider that is registered under the Ontario Career Colleges Act, 2005 or exempt from registration under it, a union-based training centre, a sector council or an industry association. The employer is responsible for finding a provider that meets these rules.
The grant does not fund business consulting services, membership or subscription fees, or training for business owners, including people who control the corporation. ThriveAI does not claim that its sessions qualify for the OJG. If the grant matters most, ask an eligible provider, such as a local college, whether it runs AI courses for employers.
Quebec: Programme de formation de la main-d'œuvre
In Quebec, the Programme de formation de la main-d'œuvre helps employers pay for staff training. Every employer located in Quebec can apply, except public bodies and political organizations. The program lists training on new equipment or new technology as an example, and counts trainer fees and training needs analysis among eligible expenses. To apply, contact a business advisor at your regional Services Québec office, whose analysis sets the amount of aid and the eligible expenses.
The program's guide, updated December 18, 2025, says aid generally runs up to 50 percent of eligible expenses. It rises to 75 percent for training made necessary by new equipment or new technology meant to raise productivity. Ordinary office tools such as Word and Excel do not count as new technology, and the Ministry's contribution generally does not exceed $100,000 per agreement.
The employer chooses the trainer, and the Ministry must approve that choice. The guide does not mention AI, so ask the advisor at the first meeting whether training on AI assistants counts as new technology.
Quebec's 1 percent law for payrolls over $2 million
A Quebec employer whose total payroll for the year is over $2 million must spend at least 1 percent of it on eligible training, according to Revenu Québec. An employer that falls short pays the difference into the Workforce Skills Development and Recognition Fund.
The Commission des partenaires du marché du travail (CPMT) lists what counts. Training can come from a recognized educational institution or professional order, from an accredited trainer or training body, or under a company training plan, which covers in-house training and trainers who are not accredited. For on-the-job training under such a plan, eligible costs include the salaries of the people who give and receive it. The CPMT asks employers to keep supporting records for every expense they count.
Federal funding
The Ontario Job Grant is funded in part by the Government of Canada, through the Canada-Ontario Labour Market Development Agreement.
An AI training course or training for your team
An AI training course or an AI for business course suits one person learning the basics at their own pace. Google's AI Essentials is a self-paced course of under five hours, which Google sells as a monthly subscription after a seven-day free trial in Canada (checked September 25, 2026). Google also offers no-cost workshops, including Make AI Work for You for business owners. OpenAI Academy courses are free with a ChatGPT account, and AI Foundations takes 60 to 75 minutes. Claude Academy offers free courses and tutorials from Anthropic's education team.
For teams, Amii, the Alberta Machine Intelligence Institute, offers ML Foundations 1 and 2, AI Ethics 1 and 2, and custom programs. Amii describes ML Foundations as building a shared understanding of AI terms and concepts.
Statistics Canada reported on June 11, 2026 that 19.2 percent of businesses used AI to produce goods or deliver services in the preceding 12 months, with wholesale trade among the lowest at 7.9 percent. Among businesses using AI, 32.0 percent reported AI-related training for existing employees. In manufacturing, 12.0 percent of businesses named a lack of skilled workers as a barrier to using AI.
Team training on your own work fits when you want a specific job done differently next month. It also settles what a course leaves to each learner: which accounts are approved, which files may go where, and who approves what the tools draft.