ChatGPT or Copilot for business: a rollout plan that changes how work gets done
A practical six-step plan for rolling out ChatGPT Business or Microsoft 365 Copilot so that named workflows change.
BY STEFAN MARITZ··7 MIN READ
TL;DR
- Rolling out ChatGPT or Microsoft Copilot to a team is an adoption project with a licence attached. Buying seats is the easy part; changing how specific workflows run is the work.
- Choose the tool by where your work already lives. Teams running on Microsoft 365 will often find Copilot fits more naturally, while mixed-tool teams often prefer ChatGPT Business, a self-serve workspace plan for teams of two to 200.
- Start with three or four named workflows per team, a short usage policy, and a small group of early adopters who coach their colleagues.
- Train people on their own work, using their own documents and tasks. Generic prompt training produces enthusiasm in the room and little change at the desk.
- Measure changed workflows and time saved on named tasks. Active-user counts only tell you the tool was opened.
The pattern we keep seeing with ChatGPT and Copilot rollouts is familiar. Usage jumps in the first fortnight, settles at a handful of enthusiasts, and the renewal conversation a year later is mostly about whether to cut seats.
This post is a practical rollout plan for either tool. It covers why rollouts stall, how to choose between ChatGPT Business and Microsoft 365 Copilot, a six-step plan, and what to measure. I have written it for the operations, HR, or marketing lead who has been handed the licences and asked to make them worth it.
Why do most ChatGPT and Copilot rollouts stall?
Because the rollout is planned as a technology deployment and the hard part is behavioural. Licences get assigned, a launch email goes out, there is a training session, and then people return to deadlines with a new tab open and no reason to change how they work.
Three gaps come up again and again. Nobody named which tasks should change. Training covered the tool and skipped the work. And the measure of success was licences issued, which tells you almost nothing. Whether a person's actual workflow changed is the measure of AI adoption that counts.
ChatGPT Business or Microsoft 365 Copilot: how to choose
Choose by where the work lives and what your administrators can support. Both are serious business tools, and the wrong one for your environment will be used less, whatever its features.
ChatGPT Business is a self-serve workspace plan with centralised billing, admin controls, and Standard or Premium seats. OpenAI positions it for teams of two to 200 employees, and business content is not used to train its models by default. Microsoft 365 Copilot sits inside the Microsoft 365 apps, and Microsoft publishes a full Copilot adoption guide organised into four phases: plan, implement, adopt, and manage.
| Question | Points towards ChatGPT Business | Points towards Microsoft 365 Copilot |
|---|---|---|
| Where do documents, email, and meetings live? | Mixed tools, or Google Workspace | Microsoft 365 throughout |
| Who will administer it? | A small team wanting self-serve setup | An IT team already managing Microsoft 365 |
| What will people do most? | Drafting, analysis, and research across varied sources | Work inside Outlook, Teams, Word, and Excel files |
| How many seats? | Up to about 200 | Any size, managed through existing licensing |
Pricing changes often, so check each vendor's current page before you commit. Whichever you choose, the rollout plan below applies almost unchanged.
How to roll out ChatGPT or Copilot in six steps
- Name three or four workflows per team. Specific ones, such as "first draft of the weekly client report" or "summarise supplier contracts before review". If you cannot name the workflow, the licence is not ready to be issued.
- Publish a one-page usage policy before launch. Which data may go in, which outputs need review, and where to ask questions. A short AI policy people will follow also brings personal-account use back into the open, which is the usual source of shadow AI.
- Start with early adopters, and give them a role. Pick one or two people per team who are curious and respected. Microsoft's own adoption planning checklist starts with selecting early adopters for the same reason. Their job is to work out what good looks like on the named workflows and then show colleagues.
- Train on people's own work. Run sessions where people bring a real task and finish it with the tool in the room. The best AI training for employees uses their documents and their deadlines.
- Rewrite the process on paper. Decide what each workflow now looks like, including who reviews AI output and which step it replaces. If the old step still exists, people will keep doing it.
- Review at 30, 60, and 90 days. Check the named workflows, collect the prompts and examples that worked, and move licences from people who have not used them to people waiting for them.
Think of it as commissioning a new machine on a line. Installing it takes a day. Getting the operators to run it at full output, with the old manual workaround retired, takes the rest of the quarter.
What should you measure?
Measure the work, and use tool statistics only as a smoke alarm. Usage reporting, such as the Copilot Dashboard and usage reports Microsoft provides, is useful for spotting who has never logged in. It cannot tell you whether anything improved.
| Measures that mislead on their own | Measures that show change |
|---|---|
| Licences assigned | Time taken on each named workflow, before and after |
| Weekly active users | Share of each named workflow now done with the tool |
| Training attendance | Rework or error rates on AI-assisted output |
| Prompts sent | Whether people say they would object to losing it |
The same logic applies to measuring AI training effectiveness, and most of it carries straight over to a rollout.
Frequently asked questions
How do you roll out ChatGPT or Copilot to a team?
Name three or four specific workflows per team, publish a short usage policy, start with early adopters, train people on their own work, update the process on paper, and review at 30, 60, and 90 days. Treat it as an adoption project rather than a software deployment.
Should we choose ChatGPT Business or Microsoft 365 Copilot?
Choose based on where your work lives and who will administer it. Organisations running entirely on Microsoft 365 often find Copilot fits more naturally, while smaller or mixed-tool teams often prefer ChatGPT Business.
Is company data used to train ChatGPT Business models?
OpenAI states that content in ChatGPT Business is not used to train its models by default. Check the current terms on OpenAI's site and set your own rules on which data may be used.
How long does a Copilot or ChatGPT rollout take?
Licences can be live in a day. In our view, changing how a team works on its named workflows takes about a quarter of deliberate effort, with reviews along the way.
How do you measure the success of a ChatGPT or Copilot rollout?
Measure time and quality on specific workflows before and after, and the share of those workflows now done with the tool. Active-user counts help spot non-users but do not show whether work improved.
If you want your team to leave a session with their own workflows already changed, that is what our AI workshops are built around.
WORK WITH PRAXES
Ready to move from reading about AI to getting your organisation ready for it? Start with the diagnostic.
Take the AI Readiness Diagnostic