On this page
TL;DR:
- A good on-site programme has small groups, working logins, a tested internet connection, real work to practise on, and a follow-up date already in the diary.
- Most failures happen before the trainer arrives. Sort out accounts, passwords, firewalls and devices at least two weeks ahead.
- Test the Wi-Fi from the training room, with everyone connected, and keep a mobile data backup.
- Staff need their phones in the room. Many work logins ask for a code sent to the phone.
- Book the follow-up before the training starts. Without it, new habits are hard to keep.
Who this guide is for
This guide is for the person who has been told, "Please organise the AI training." You may be in HR, administration or operations, or be an executive assistant. You do not need to know anything about AI.
It works for a programme at your own office, a branch, a factory, a hotel or a retreat venue, anywhere in Kenya. Use the checklist near the end as your working document.
What does a good on-site programme look like?
A good programme has staff practising on their own kind of work, on tools they can still use the next morning. Everything else in this guide supports that.
| Part | What good looks like |
|---|---|
| Group | Up to about 20 people, sorted by the work they do |
| Room | Tables, power at every seat, a screen everyone can read |
| Internet | Tested at the same time of day, with a backup |
| Devices | A laptop for each person, or at worst one between two |
| Tool access | Work logins tested a week before |
| Content | Built from real tasks, cleaned of personal data |
| Structure | Shared foundation first, then practice by department |
| Follow-up | Dates set, a prompt library, a named internal contact |
When is on-site the right choice?
Choose on-site when hands-on help matters, when staff work in the same place, or when people need protected time away from their desks.
| Your situation | Likely best format |
|---|---|
| New to AI, needs someone at their shoulder | On-site |
| Shift workers, one location | On-site, split into teams |
| Many branches, confident users | Live online, or hybrid |
| Leadership team needs privacy and focus | On-site or a retreat venue |
| Very tight budget, small team | Live online |
| Weak internet at your site | Fix it first, or move to a venue with better internet |
FuKazee's base is in Westlands, Nairobi. Anywhere else in Kenya we work on-site by arrangement, or live online. If you are weighing up who should deliver the training at all, see in-house versus external AI training.
How should you split the groups?
Keep groups to about 20 people, and sort them by the work they do. A trainer can only reach so many screens, and the people who get stuck are often the ones who stay quiet.
Three decisions come up every time.
Should managers train with their teams? For a first session, usually separate. Staff ask fewer basic questions in front of their boss, and managers need a different discussion: how to review AI-assisted work and what to expect from their teams.
Should departments be mixed? Mix them for the shared foundation if you must. Separate them for the practice sessions. Finance staff need finance examples.
What about staff on shifts? Plan around operations. In October 2026 we trained two teams at A.O. Bayusuf and Sons, a transport and haulage company in Mombasa, on-site, over three days each. Starting early and splitting staff into teams lets an operation keep running while people train.
Also ask about ability. A short survey tells you who has never used AI and who uses it every day. Pair them up, because the confident ones become your helpers.
What should the training room have?
Choose a room where people can work with both hands, see the screen and plug in. A boardroom with one long table often looks right and works badly, because the trainer cannot reach people in the middle.
| Item | What to check |
|---|---|
| Layout | Small tables or a U-shape, with space to walk behind every chair |
| Power | A socket for every laptop. Bring extension cables and tape them down |
| Screen | Readable from the back row. Sit there and check it yourself |
| Connection to the screen | The right cable or adapter for the trainer's laptop |
| Sound | Quiet, no generator or road noise, no echo |
| Air | Working air conditioning or windows that open. A hot room after lunch ends learning |
| Light | Blinds, so that the screen can be read at 3 pm |
| Writing | Flip chart and markers, for when the screen fails |
| Food | Tea and lunch near the room, ready on time |
| Privacy | A door that closes, so staff can speak freely |
Plan for power cuts. Ask the facilities team whether the room is on backup power. Ask everyone to arrive with a fully charged laptop and phone. If the power goes, you can continue on batteries for a while, and the flip chart will carry a discussion.
Keep people away from their desks. Training at the office has one big weakness: colleagues will come looking for your trainees. Ask managers to cover the work. Put an out-of-office message on for the day. If that fails, a room on another floor or an outside venue is worth the cost. For leadership teams who want to get away, Naivasha is one option.
How do you make sure the internet will work?
Test it from the training room, at the same time of day as the session, with as many devices connected as you can gather. A speed test run alone at 7 am tells you very little about 20 people at 11 am.
Kenya's connectivity figures explain why you need a backup. By the end of March 2026, the Communications Authority counted about 2.66 million fixed internet subscriptions, of which about 1.47 million were on fibre. It counted 62.6 million mobile data subscriptions, and 84.4% of those were on mobile broadband. For many sites, a mobile connection is the most realistic second option.
Do these five things.
- Find out which network staff will use. Guest Wi-Fi and the staff network often have different rules.
- Ask IT to open what the tools need. A company firewall can block AI tools even when the speed is fine. Microsoft says Copilot can fail when a network blocks secure WebSocket connections, when a security device inspects and interferes with traffic, or when a proxy cuts connections too quickly. It asks organisations to allow all of
*.cloud.microsoft. - Run the test. Microsoft's Copilot requirements page links to a connectivity test for the Copilot app. Ask IT to run it from the training room network.
- Have a backup. A dedicated mobile router on a different network from the main line, with enough data bought in advance. Check signal strength inside the room, since thick walls and basements can block it.
- Reduce the load. Typing prompts uses little data. Uploading large files and generating images use much more. Share practice files before the day, and ask staff to close video and cloud-sync apps.
Have the trainer bring an offline version too: printed prompts, screenshots of the steps and a paper exercise. If the internet dies for an hour, people can improve prompts on paper and compare them with a partner.
What devices do people need?
One laptop per person is the goal. Phones work for a taste of AI, but it is hard to do real work on a phone with Excel or a long report.
Send a short device survey two weeks ahead. Ask what device each person will bring, what operating system it runs, and whether it is a company or personal machine.
| Situation | What to do |
|---|---|
| Company laptops, locked down | Ask IT to confirm tools are reachable. Staff may not be able to install anything |
| Older computers | Use a current browser. Microsoft lists supported browsers for web use |
| Some staff have no laptop | Pair them, no more than two per screen, or borrow from another department |
| Staff with phones only | Fine for chat practice, weak for documents |
| Personal laptops | Check your policy allows company material on them |
If you plan to use Microsoft tools, know these points as of November 2026. The Microsoft Copilot desktop app needs Windows 11 version 22H2 or later, or macOS 13 or later. On phones it needs iOS 16 or Android 10 or later. For Copilot to work in the web versions of Word, Excel and PowerPoint, third-party cookies must be enabled, and some company browser policies block them. Check these before the day.
How do you sort out tool access?
Decide which tools people will use, and test a real login, at least one week before. Lost passwords can swallow the first hour of a training day.
| What your organisation has | Sensible plan |
|---|---|
| Eligible Microsoft 365 business or enterprise plan | Microsoft Copilot Chat with work accounts. Needs a Microsoft Entra work or school account |
| Microsoft Copilot paid licences for some staff | Teach the shared free features first, then show what the licence adds |
| Google Workspace | Ask your administrator whether Gemini is switched on. Work accounts require users to be 18 or over |
| No business AI tools | Free personal accounts for learning, using only public or made-up material |
As of November 2026, Microsoft says Copilot Chat without the paid licence covers web-based answers and files you upload, and that Copilot Chat in Outlook can reach some of your emails, meetings, chats and files. Full access to that work data needs the paid licence. For more detail see our Microsoft Copilot training guide. We never promote a single tool, so match the training to what your staff can really use after we leave.
Check which account people are using. Microsoft notes that work and personal sign-ins are kept separate, and that enterprise protections depend on using the work account. Look for the "Work" label and the green shield. Other tools have similar differences. Google says that Gemini chats in editions without enterprise-grade protections may be reviewed by people. Anthropic lets users of its Free, Pro and Max plans choose whether their data is used to improve Claude. Read the current terms of whichever tool you use.
The phone rule. Many work accounts ask for a code on the phone at sign-in. If phones are collected at the door, or left at a desk, nobody can log in. Ask everyone to bring their phone and to know their password. Have someone from IT in the room for the first 30 minutes.
How do you prepare practice material safely?
Collect five to ten real tasks from each group, then remove anything personal or confidential. Practising on real work is what makes training stick, but real files can carry real risk.
Kenya's Data Protection Act is in force, and it applies to personal data pasted into AI tools. Before any document is used for practice:
- Replace names, phone numbers, ID numbers, KRA PINs and account numbers.
- Remove anything about health, salaries or disciplinary matters.
- Change or remove customer names and contract values.
- Have one named person approve the final set.
Agree the rules for that room in advance. If your organisation does not have an AI usage policy yet, write a one-page version, and make it the first thing the trainer covers.
Photos and recordings. Ask before you take any. Tell staff who will see the material and where. Allow anyone to say no, without having to explain.
What should the timetable look like?
Build from shared foundation to department practice, and change the format every 30 to 45 minutes. This sample is a starting point, so adjust it to your shifts and prayer times.
Day 1: shared foundation
| Time | Session | Format |
|---|---|---|
| 8:30 | Logins, tea, phones ready | IT support in the room |
| 9:00 | What AI does well, and where it fails | Demonstration, examples from your sector |
| 10:00 | The rules for our organisation: what never goes in | Discussion |
| 10:30 | Break | |
| 11:00 | The four Cs: Clarity, Context, Constraints, Customisation | Everyone writes and improves a prompt |
| 12:30 | Lunch | |
| 1:30 | Catching mistakes: find the errors in an AI draft | Pairs |
| 2:30 | Microsoft 365 and Copilot in Word, Excel and PowerPoint, or your own tools | Hands-on |
| 3:45 | Each person picks one task to try tomorrow | Individual |
| 4:15 | Close and feedback | Form |
Day 2: your department's real work
| Time | Session | Format |
|---|---|---|
| 8:30 | What happened when you tried it | Group sharing |
| 9:00 | Department task clinic, part one | Small groups, trainer moves between them |
| 11:00 | Spreadsheets and data tasks | Hands-on |
| 1:30 | Department task clinic, part two | Each person finishes a real output |
| 3:00 | Build the team's prompt library | Group work |
| 4:00 | Commitments, follow-up dates | Each person states one habit |
Day 2 should feel different from day 1: less talk from the front, more work at the table. Ask any provider how the structure changes from day to day, as well as the content.
If you add a third day, use it for deeper practice on current work, or a separate session for managers on reviewing AI-assisted work.
Be honest about what this achieves. Nobody becomes an expert in two or three days. People should leave safe, able to do a few useful things well, and clear on how to keep learning.
What if your site is outside Nairobi?
Plan an extra day for travel and setup, and confirm in writing what the quotation includes. Ask whether travel, accommodation and materials are part of the total, so nothing arrives as a surprise.
Ask the local venue for photos of the actual room, since a venue's website may show its best one. Test the connection before you commit. We have location pages for Nairobi, Mombasa and other towns if you want to see what that looks like in practice.
What should follow-up include?
Book it before the training begins. Without a date, it tends not to happen, and new habits are hard to keep. Our article on why corporate AI training doesn't stick explains this in more detail.
| When | What | Owner |
|---|---|---|
| Same week | Share prompts, slides and recordings in one shared folder | Organiser |
| Week 2 | Short online clinic for questions | Trainer |
| Week 4 | Managers ask each team: what are you using, and what is blocking you? | Managers |
| Week 6 | Internal champions share the three most useful prompts | Champions |
| Week 12 | Check what has changed against your starting point | HR, sponsor |
Name one or two champions per department, and pick curious, patient people rather than simply the most technical. Give them an hour a week for the first month.
The organiser's timeline
| When | Task |
|---|---|
| 6 weeks before | Agree the goal, who attends, dates and the budget |
| 5 weeks before | Book the room. Check power, internet and layout in person |
| 4 weeks before | Send a survey on skills, tasks and devices |
| 3 weeks before | Collect real tasks. Confirm which tools will be used |
| 2 weeks before | IT tests network and access. Clean practice material |
| 1 week before | Test logins for every participant. Send joining instructions |
| 2 days before | Confirm catering, cables, backup internet and printouts |
| Day before | Set up the room and do a full test on the screen and Wi-Fi |
| Day of training | Arrive an hour early. IT present for the first 30 minutes |
| After | Share materials and run the follow-up table |
Three prompts organisers can copy
These use the four Cs we teach. They help you do the admin faster, and they work in any of the main AI tools.
Prompt 1: joining instructions
Write joining instructions for staff attending a two-day AI training at our office. Context: many staff have never used AI and some feel nervous. Constraints: under 200 words; list the date, room and start time as [placeholders]; tell them to bring a fully charged laptop, their charger, their phone and their work password; tell them to clear their diaries; say no experience is needed. Customisation: it will be sent by email and read on phones, so make it friendly, in simple English, with short lines.
Prompt 2: make a real document safe to practise on
Below is a work document with names and details already replaced by me. Check it for anything that might still identify a real person or customer, such as job titles that only one person holds, locations, dates of specific events, or amounts. List each item and suggest a neutral replacement. Do not rewrite the document. Constraints: be cautious and flag anything doubtful. Document: [paste text with personal data already removed]
Remove the obvious personal data yourself first. Use an approved tool, and have a colleague check the result.
Prompt 3: build the run sheet
Create a minute-by-minute run sheet for a training day. Context: [number] participants, [location], start [time], end [time], with the sessions listed below. Constraints: a table with columns for time, activity, who is responsible, and what could go wrong; include a backup plan for loss of internet and loss of power; do not add any session I have not listed. Customisation: I will print it on one page and carry it on the day. Sessions: [list]
What goes wrong, and how to avoid it
| What goes wrong | How to avoid it |
|---|---|
| The first hour disappears on password resets | Test every login a week before; IT in the room |
| AI sites are blocked on the office network | Ask IT to test the tools from that room two weeks before |
| Wi-Fi collapses with everyone connected | Test under load, and keep a mobile backup on another network |
| Staff keep leaving for "urgent" work | Managers cover their work; use a room away from desks |
| Examples are about other industries | Collect real tasks three weeks before |
| Someone pastes a real customer file | Clean practice material, and give the data rules first |
| A senior person dominates the room | Run managers separately for the first session |
| People learn on a tool they cannot access later | Train on what they will really have next week |
| Good energy, then nothing | Put follow-up dates in diaries before training starts |
The one-page organiser's checklist
People
- [ ] Goal of the training agreed in writing
- [ ] Groups of up to about 20, sorted by work
- [ ] Skills and device survey completed
- [ ] Managers have agreed to cover the work
Room
- [ ] Layout lets the trainer reach every seat
- [ ] Power at every seat; extension cables ready
- [ ] Screen checked from the back row
- [ ] Backup power confirmed; flip chart in the room
Internet
- [ ] Tested in the room, with many devices, at the same time of day
- [ ] IT has confirmed AI tools are not blocked
- [ ] Mobile backup on a second network, with data loaded
Tools and data
- [ ] Tools chosen; every login tested
- [ ] Phones in the room for sign-in codes
- [ ] Practice material cleaned and approved
- [ ] Data rules ready for the first session
After
- [ ] Follow-up dates in diaries
- [ ] Champions named
- [ ] Shared folder set up
- [ ] Starting point recorded, so you can check what changed
Questions people ask
How long should on-site AI training be?
It depends on the goal. One day can build awareness and safe use. Two or three days allow real practice on department work. Our work at A.O. Bayusuf ran for three days per team. Shorter and spaced is often easier for operations than one long block.
How many people can be trained at once?
We plan for groups of up to about 20. Larger groups can listen to a talk, but hands-on help becomes thin. If you have 60 people, plan for three groups.
Can staff use their phones instead of laptops?
For a first look at AI chat tools, yes. For work in Word, Excel or long documents, a laptop is much better. Pair people up if you are short of machines.
Do we need to buy AI licences before the training?
Often no. If you have an eligible Microsoft 365 plan, Copilot Chat may already be available with work accounts. Free versions of other tools are enough to learn on, as long as staff use only public or made-up material.
What if our internet is poor?
Fix what you can, carry a mobile backup, and share files ahead of time. If the connection simply cannot cope, move the session to a venue that has been tested, or use an approach based on smaller groups taking turns.
Can you train in Kiswahili?
Ask your provider directly, and test the quality of the tools too. AI output in Kiswahili varies, so a fluent speaker should always check anything that will be used outside the room.
Should we use real company documents?
Yes, after removing personal and confidential details. Real work makes training useful. Have one named person approve the practice set.
Sources
All links last checked 9 November 2026.
- Communications Authority of Kenya, Sector Statistics Report, Q3 FY 2025/26 (1 January to 31 March 2026)
- Microsoft Learn, Microsoft Copilot requirements
- Microsoft Learn, Manage Copilot Chat
- Microsoft Learn, Copilot Chat privacy and protections
- Microsoft Support, Copilot Chat with and without a Microsoft Copilot licence
- Google, Use Gemini Apps with a work or school Google Account
- Anthropic, Updates to consumer terms and privacy policy
- Data Protection Act, 2019 (official text on the ODPC website)
Where FuKazee fits in
FuKazee is a Nairobi AI training company. Since 2022 we have trained more than 3,000 people. We can run training at your premises, across branches, online or as a hybrid. Outside Nairobi, on-site delivery is by arrangement.
We assume nobody in the room knows AI yet. We avoid jargon, show the free version of each feature, and build exercises from your own kind of work. Participants keep reusable prompt libraries and recordings, and we check in with you afterwards. See how it works on our AI training in Kenya page.
To plan a programme, tell us your location, how many people need training, what they do every day and when you would like to run it. Email lead@fukazee.com.
Want guided support instead?
Plan AI training around your team's real work. Answer a few quick questions about your sector, team size and goals, and we will recommend the right programme and format. Sessions are built on your workflows and include a usage policy.
Rodney has run FuKazee's training since 2022 and still teaches inside every programme line: corporate sessions, the postgraduate cohort and the KCSE examiner webinars. First Class Honours in Journalism and Mass Communication, University of Nairobi.
