Training employees on ChatGPT Enterprise works when you teach it in stages, inside the apps people already use, with managers deciding which custom GPTs their teams share. One launch webinar gives people access. It does not change habits.
This guide is for IT directors and CIOs who need employees to use ChatGPT Enterprise well and safely. It explains how to train employees on ChatGPT Enterprise in four stages, how to bring personal ChatGPT use into the company workspace, and how to position it next to Microsoft 365 Copilot.
The short version
- Start with why the Enterprise workspace is different from personal ChatGPT. That difference is the reason employees should switch, so teach it first.
- Train in four stages: foundations, prompting in the flow of work, custom GPTs for team workflows, then projects and company knowledge.
- Turn on ChatGPT in Word, Excel and PowerPoint early. Using AI where the work already happens is what makes the habit stick.
- Build custom GPTs with managers, not just with enthusiasts. A GPT the team has signed off can be shared and reused, and that's where the efficiency comes from.
- If you run more than one AI tool, decide who gets what and why before you train anyone. Clear roles make for clear messaging.
How ChatGPT Enterprise differs from personal ChatGPT
ChatGPT Enterprise is a company workspace. IT controls identity, data and features. Personal ChatGPT is controlled by each individual. Most employees have already tried ChatGPT at home, so the first training module should explain exactly what changes when they sign in through the company.
OpenAI states that it does not train its models on business data by default. Enterprise workspaces also add encryption, retention controls, SCIM provisioning, role based access and user analytics (OpenAI business data privacy). Company knowledge answers questions from connected sources and only returns what the employee already has permission to see (OpenAI Help Center).
| Area | Personal ChatGPT | ChatGPT Enterprise |
|---|---|---|
| Who owns the account | The employee, often on a personal email | The organization, through SSO |
| Use of data for model training | Set by each user's own settings | Not used for training by default |
| User provisioning | None | SCIM provisioning and deprovisioning |
| Admin control | None | Admin console, role based access, user analytics |
| Data retention and residency | Individual settings | Configurable retention; data residency for eligible customers |
| Sharing | Individual chats | Custom GPTs and projects shared in the workspace |
| Company data | Manual uploads | Company knowledge that respects source permissions |
In training, show this rather than present it. Sign in through SSO, open a shared custom GPT, and ask company knowledge about a familiar policy. Employees switch when they see the Enterprise version does more for them.
A four stage ChatGPT Enterprise training roadmap
The best ChatGPT Enterprise training moves people from occasional prompting to repeatable team workflows. OpenAI's own data shows how big that gap is. In its State of Enterprise AI report, the most active 5 percent of workers sent 6 times more messages than the median employee. Use of custom GPTs and projects grew 19 times in the year to date. Depth, not access, separates the leaders from everyone else.
We map each stage to the PROSCI ADKAR model. That way training builds Awareness and Desire first, then Knowledge and Ability, and ends with Reinforcement.
| Stage | Suggested timing | What employees learn | ADKAR focus | Format that works |
|---|---|---|---|---|
| 1. Foundations | Weeks 1 to 2 | What the workspace is, data rules, checking outputs for errors | Awareness, Desire | Short Show Me session led by a leader |
| 2. Prompting in the flow of work | Weeks 2 to 6 | Prompt structure, ChatGPT in Word, Excel and PowerPoint, reusable prompts | Knowledge | Prompt a Thon with real tasks |
| 3. Custom GPTs for team workflows | Weeks 6 to 10 | Building, testing and sharing GPTs for repeated team tasks | Ability | Let Me Try workshops with managers in the room |
| 4. Projects, company knowledge and deep research | Weeks 10 onward | Longer work in projects, connected sources, research tasks | Ability, Reinforcement | Champion office hours and self serve learning |
Keep Stage 1 short: clear rules and a reason to care. Stage 2 takes the most time, because prompting skill grows only through practice on real work. Stages 3 and 4 are where the payback starts, because one good GPT or project serves a whole team.
Bring ChatGPT into Word, Excel and PowerPoint early
The fastest way to build a daily habit is to put ChatGPT where people already work. OpenAI now offers one ChatGPT extension that works across Word, Excel and PowerPoint. It opens in a sidebar where employees can ask about the open document, revise selected text or draft from notes (OpenAI Help Center).
It removes the copy and paste step between a browser tab and the document. In the rollouts we support, tools used in the flow of work stick, and tools that need a separate window fade.
Two settings decide whether employees can use it, so check both before Stage 2 begins. Workspace admins can turn the extension on or off in the ChatGPT admin console. Microsoft 365 admins must also allow it before it appears for employees. OpenAI says it is enabled by default from October 1, 2026. Your organization may already be exposing it, so decide your position now rather than discovering it in a support ticket.
Custom GPTs: build them with managers, then share them
Custom GPTs are solutions every employee can use, so treat them as team assets, not personal experiments. The GPTs that create real efficiency are the ones a team agrees on and shares. A clever GPT built by one enthusiast tends to stay with that one person.
Involve managers in deciding which GPTs to build. Managers know which tasks repeat, which outputs need a consistent format and which shortcuts their team will trust. A GPT that a manager has signed off becomes the sanctioned way to do that task, and colleagues are far more likely to use it.
Here is the process we teach for a team GPT:
- The manager names one repeated task that costs the team time every week.
- A champion on the team builds a first version, with clear instructions and approved reference files.
- Three to five team members test it on real, permitted work for one week and note where it fails.
- The manager signs off and names an owner who keeps the GPT current.
- The owner shares it in the workspace and adds it to the team's GPT library with a one line description.
- Each quarter, the owner reviews usage and updates or retires the GPT.
This turns scattered experiments into a library people can find, trust and reuse, and gives managers a visible role in adoption.
Move shadow use into the Enterprise workspace
Many employees already use AI tools their company has not approved. A PagerDuty survey of 1,250 office professionals at large organizations found that 66 percent have used unauthorized AI tools at work. Training is your best chance to bring that use into a governed space.
Make the approved path the easier one. Announce the Enterprise workspace with a clear message: here is why it's better for you, here is what data is allowed, and here is how to bring over the prompts you already use. Avoid opening with penalties. Employees who used personal ChatGPT are often your most motivated learners, and many of them make good champions.
When ChatGPT Enterprise and Copilot both exist: decide who gets what
It's not unusual for an organization to run more than one AI tool. What matters is that IT leaders decide who gets which tool and why, before training starts. Without that decision, every training session turns into an argument about which tool to open, and messaging falls apart.
Our view is that most organizations already invested in Microsoft 365 get the most from Microsoft 365 Copilot. It works across the Microsoft 365 apps and is grounded in mail, meetings and files. Every extra tool adds spend to manage, security settings to maintain and another product to train people on. A second tool should earn its place with a clear use case.
| Group | Primary tool | Reason to write down |
|---|---|---|
| Most Microsoft 365 knowledge workers | Microsoft 365 Copilot | Grounded in their mail, meetings and files inside existing permissions |
| Teams with a defined ChatGPT use case | ChatGPT Enterprise | Specific custom GPTs, research or connected sources Copilot does not cover for them |
| Developers and specialist roles | The approved tool for that work | A clear use case such as code, design or analysis |
| Everyone | One AI acceptable use policy | One set of data rules across every tool |
Once this is written down, each group learns one primary tool in depth and everyone hears the same reasons. For the full set of options, see our guides on choosing the right AI for the work and Amazon Quick vs Microsoft Copilot.
Measure ChatGPT Enterprise adoption and keep managers reinforcing it
Measure depth, not just logins. Use the user analytics in the ChatGPT Enterprise admin console, plus a short manager check in at 30, 60 and 90 days. Track four things:
- weekly active users as a share of licences
- employees using at least one shared custom GPT
- number of team GPTs signed off and in active use
- tasks managers can name that now take less time
Managers are the reinforcement layer. When a manager asks in a team meeting which GPT saved someone time this week, usage climbs. When nobody asks, old habits come back. For a broader benchmark view, our post on what good Copilot adoption looks like uses the same measures.
The rule to remember: govern the account, teach the habit, let managers keep it.
Licences give people access. Training gives them the skill. Managers give them permission to change how the work gets done.
What IT leaders should do next
- 1
Write down which groups get ChatGPT Enterprise, which get Copilot, and the reason for each.
- 2
Publish one AI acceptable use policy that covers every tool you license.
- 3
Decide your position on the ChatGPT extension for Word, Excel and PowerPoint in both the ChatGPT and Microsoft 365 admin settings.
- 4
Recruit one champion per team, starting with the people who already used ChatGPT on their own.
- 5
Run Stage 1 and Stage 2 with real tasks from each department, not generic demos.
- 6
Ask every manager to name one task for a team custom GPT, and run the sign off process.
- 7
Set up 30, 60 and 90 day check ins using admin analytics and manager feedback.
How Change Champions helps
Change Champions has helped more than 500 organizations and 60,000 users adopt new technology, using PROSCI and ADKAR to focus on the people side. We help IT leaders turn AI licences into everyday habits, whether the tool is ChatGPT Enterprise, Microsoft 365 Copilot or both.
- AI enablement: a staged training roadmap, champion network and adoption measurement.
- Prompt a Thon: hands on sessions where teams build prompts on their own work.
- Change enablement: manager toolkits and communications that make adoption stick.
Ready to turn rollouts into real adoption?
Whether you're driving Microsoft 365 and AI adoption, reducing human risk, or leading a change program, a 30 minute discovery call is the fastest way to see if we're a fit.