Most companies in Hong Kong already pay for Microsoft 365, so they assume they already have Copilot. Often they have a handful of seats, bought as a trial, spread across people who never asked for them. I have walked into a session designed around Copilot and found that most of the room could not open it. I now count licenses before I count participants, and I would tell any company buying Copilot training to do the same before they sign with me or anyone else.
Why I still start with Copilot
When I wrote about why I teach Copilot instead of ChatGPT, the response was immediate, because every enterprise training buyer faces the same three worries: data security, IT pushback, and staff pasting confidential documents into consumer tools. Copilot sits inside the Microsoft 365 boundary. The data stays where it already lives, IT has already approved the platform, and there is no new login.
For Hong Kong companies dealing with PDPO and cross-border data questions, that matters more than anywhere else. When I ran five sessions for 1,530 banking professionals at a major Hong Kong bank, the first twenty minutes of every session went on data safety using the Traffic Light Protocol. Starting inside the tenant meant we could skip the "is this tool allowed?" conversation and spend the time on actual workflows.
The same rule ran a 6-session Pioneer Program for a Hong Kong food manufacturer's HR team: Copilot only, inside their Microsoft 365 tenant, because their IT lead had already found staff using personal phones for AI. That constraint sounds limiting. It removed a month of arguments.
Two questions for your IT team before you book anyone
- How many of the people who will be in the room have a paid Microsoft 365 Copilot seat today, and which SKU?
- On what date do the rest get one?
If the answer to the second question is "we're looking into it," you are buying a demo, not training.
On price, the number people quote is usually wrong. Microsoft sells Copilot at about US$30 per user per month on enterprise plans, and around US$18 to US$21 for organizations under 300 seats on the standalone business plan. Your IT contact knows which one you have. Ask, because a trainer who prints one flat price gets corrected in the room.
What Copilot training should cover
Bad Copilot training looks like a Microsoft product demo. Someone clicks the Copilot icon in Word, chooses "summarize," and says "isn't that amazing?" The audience nods. Nothing changes on Monday.
Good Copilot training starts with the work your team already does, in the four apps where Copilot earns its seat:
Outlook. Triage and first drafts, not summaries. Summarizing an email you were going to read anyway saves nothing. Drafting the reply to the one you were avoiding does.
Excel. Describe the analysis in plain language, let Copilot write the formula, then check it. Checking is the skill, and it is the part people skip. Bring a real file. A departmental P&L with 36 rows teaches more than any sample dataset, and a messy export with deleted-but-present rows teaches what Copilot does when the data is dirty — it hallucinates until you clean it.
Teams. Meeting recap and action items. This is the fastest win and the one people keep using.
Word. Policy and report drafts from a bullet list. HR teams use it for recruitment policies and onboarding documents.
The three ways companies waste the license
They train everyone at once. A company-wide webinar, no practice, no application to real work. A week later the only people using Copilot are the ones who were using it before.
They treat it as an IT rollout. Licenses deployed, a "tips and tricks" email sent, nobody shown how Copilot fits their Friday report. Start from workflows, not features. "Show me the report you write every week" is a better opening than "here's what Copilot can do."
They measure the wrong thing. Activations, query counts, session numbers. None of those predict whether the Friday report got faster. Measure time saved on named tasks, in hours per week.
What a good program looks like
Session 1 (half day). Safety framework, then Outlook and Teams workflows. Participants leave with three to five things they will use on Monday.
Session 2 (half day, week 2). Excel and Word, tied to each participant's own deliverables. They bring their files. We redesign them together.
Sessions 3 to 6 (60 minutes each, weeks 3 to 6). Weekly check-ins. What did you try, what worked, what didn't. This is where the behavior change happens, through repetition and someone asking.
That is the structure behind the Pioneer Program, and it is why the second department at that food manufacturer was booked before the first one finished.
Funding
Two schemes matter for Hong Kong companies in 2026. NITTP, formerly RTTP, matches course fees 1:1 up to HK$250,000 per enterprise per financial year for VTC-registered courses. The BUD Fund has a HK$7 million cumulative ceiling per company and a HK$150,000 cap per application on the fast-track Easy BUD route, and it funds training only inside a wider upgrading project. TVP closed to new applications in December 2024. The details, checked in August 2026, are in the buyer's guide.
If you have Copilot and nobody uses it
Count the licenses. Pick one department and one weekly task. If the count is right and the task is real, six weeks is enough to change how that department works. If the count is wrong, no training fixes it, and I would rather tell you that before the first session than find out at the second.
