On this page
TL;DR:
- AI onboarding has two jobs: helping HR and managers prepare a better induction, and teaching new staff your AI rules before bad habits form.
- Use AI to draft first-week plans, role guides, FAQs and quizzes from your approved documents. A person still welcomes, coaches and judges the new employee.
- Give every new starter a one-page AI rules card on day one: approved tools, what never goes in, who checks the work and who to ask.
- ID copies, KRA PINs, bank details, medical forms and next-of-kin details should stay out of public AI tools.
- The line manager matters more than any tool. The first-week plan below asks for just under three hours of the manager's time.
Who this guide is for
This guide is for HR officers, line managers and founders in Kenyan organisations who bring in new staff, interns or contractors. You do not need a technical background or a special onboarding system.
If you run training more widely, our guide on how learning and development teams use AI covers design, job aids and evaluation. This post stays on the first days of a new job.
Why AI belongs in onboarding now
New staff are arriving with AI habits, and many workplaces give them no guidance. These are the best figures we could find. Only one names Kenya among the countries it covered.
| Source | Who was asked | Finding |
|---|---|---|
| PwC research, reported by Capital FM (August 2026) | African workers (the report gives no sample size) | 64% already use AI in their roles |
| KnowBe4 survey report (2024; survey dates not stated) | 1,300 smartphone users in 10 countries, including Kenya | 46% said their workplace had no generative AI policy; 8% were banned from using it |
| TalentLMS and BambooHR, reported by UNLEASH (July 2025) | 1,156 recent US hires | 41% said AI was part of their onboarding; 60% got no training on using AI during onboarding |
| Same survey | Gen Z hires | 44% turned to AI for support instead of people |
| Microsoft and LinkedIn (May 2024) | 31,000 people in 31 countries | 78% of AI users brought their own tools to work; 39% of AI users had AI training from their company |
We could not find a reliable study of AI use during onboarding in Kenyan organisations.
The first weeks also matter for ordinary reasons. Gallup wrote in 2021 that only 12% of US employees said their company did a good job of onboarding. It found that employees were 3.4 times as likely to call onboarding successful when their manager took an active part. Both figures are from 2021, and the first is about US employees, so treat them as a guide.
New staff may also gain the most from these tools. A study of customer support agents by Brynjolfsson, Li and Raymond found that an AI assistant raised issues resolved per hour by 14% on average and by 34% for novice and low-skilled workers. That was one type of job in one working paper. It still suggests that new people will lean on AI early, so the rules should reach them early.
The two jobs of AI onboarding
Keep these two jobs separate when you plan, because they have different owners.
| Job 1: AI helps you onboard | Job 2: new staff learn your AI rules | |
|---|---|---|
| Who it serves | HR and line managers | The new employee |
| What it produces | Plans, guides, FAQs, quizzes | Safe daily habits |
| Owner | HR, with each line manager | HR and ICT, with the manager reinforcing |
| Main risk | Wrong information, stale documents, personal data leaks | Private shortcuts, confidential data in personal accounts |
Job 1: what AI can prepare for you
AI is good at turning documents you already have into something a new person can use. Work only from approved, current documents, because AI will repeat an outdated policy with full confidence.
| Item | What AI does | Who checks it |
|---|---|---|
| First-week plan | Drafts a day-by-day plan from the job description | Line manager |
| Role guide | Turns SOPs into "how we do this here" | Process owner |
| Staff handbook FAQ | Turns policies into plain questions and answers | HR |
| Acronym and systems list | Pulls internal terms out of documents | Experienced colleague |
| Scenario quiz | Writes practice situations on your rules | HR and ICT |
| Manager check-in questions | Suggests questions for day 1, week 1 and day 30 | Line manager |
| Kiswahili version of a notice | Produces a draft for staff who prefer it | Fluent colleague |
What must stay human
Some parts of onboarding should stay with people.
The welcome. A generated message can sound warm, and a new person can usually tell when nobody actually wrote to them. Use AI for the schedule and write the greeting yourself. Have the manager meet the new person on the first morning.
Coaching. A chat tool can explain a process. It cannot tell someone how your team really works or who to ask when things go wrong.
Probation judgements. Under section 42(2) of the Employment Act (2022 revised edition), probation may not exceed six months and may be extended by up to six more only with the employee's agreement. Section 42(1) of that edition says the hearing rule in section 41 does not apply when a probationary contract ends. In 2021, in the Monica Munira Kibuchi case against Mount Kenya University, the Employment and Labour Relations Court declared section 42(1) unconstitutional. A law firm's summary at the time said the declaration would take effect only once Parliament changed the law. We could not open the judgment itself or confirm how later courts have applied it, so check the current position with an advocate and give a fair hearing either way. Section 35 of the Data Protection Act also gives people a right not to face a decision based solely on automated processing that significantly affects them, subject to exceptions in the Act. AI may help a manager organise their own notes. The manager makes and owns the judgement. This is our plain reading and not legal advice.
New employees' data: what stays out
Onboarding collects some of the most sensitive data an employer holds, so draw the line clearly.
Keep these out of public AI tools:
- National ID or passport copies
- KRA PIN, NSSF and health insurance numbers
- Bank and M-PESA details
- Medical forms and disability information
- Next-of-kin and children's details
- Salary and contract terms
- References and background checks
Under the Data Protection Act, sensitive personal data includes health status, biometric data, and family details such as the names of a person's children, parents or spouse. Section 29 says that, as far as practicable, you must tell people what you are collecting and why before you collect it. Our guide to the Data Protection Act and AI explains the rest.
Take care with welcome posts. AI makes it easy to produce a "meet our new colleague" graphic. In Complaint 1618 of 2024, decided in January 2025, the Data Commissioner ordered a solar company to pay a sales agent KES 500,000 for using his photo in marketing flyers without explicit consent, as described in an EY alert. Get written consent before you put a new employee's face or story on social media.
Job 2: teach the AI rules on day one
New staff need four answers on day one: which tools are approved, what never goes in, who checks the work, and who to ask. In our judgement, thirty minutes is enough to start.
Put it on one page. Here is a card you can copy and adapt once your organisation has a written AI usage policy.
Our AI rules, one page
1. Tools you may use for work: [list approved tools and how to sign in]. Use your work account and never a personal one for work material.
2. Never paste into a public AI tool: customer or staff personal data, payroll, contracts, passwords, unpublished financial results, or anything marked confidential. If you would not pin it on the noticeboard, keep it out.
3. You own what you send. Read and check every AI output. Check names, numbers, dates, references and quotes against the source.
4. AI does not decide about people. It does not decide on hiring, discipline, pay, credit or anything else that significantly affects a person.
5. Be open. Tell your manager when AI did a large part of a piece of work. Nobody will be penalised for honest use within these rules.
6. Bring nothing from your last job. Do not upload or paste a previous employer's documents, client lists or prompts that contain their information.
7. Made a mistake? Say so at once. Tell [named contact] the same day. Under section 43 of the Data Protection Act, a reportable breach must reach the Data Commissioner within 72 hours, so early warning matters.
8. Not sure? Ask [name, phone or email] before you paste.
Check understanding with scenarios
A signed form tells you little about what someone understood. Ask the new person to sort realistic situations, then talk through the answers.
| Situation | Answer | Why |
|---|---|---|
| Asking a tool to improve the tone of an email with no names in it | Fine | No personal or confidential data |
| Pasting a customer complaint with the customer's phone number | Not in a public tool | Remove the details or use an approved tool |
| Uploading last year's payroll to get a summary | No | Salary and personal data |
| Using a personal account because the work tool is slow | No | Tell your manager about the problem instead |
| Sending a client report with figures AI wrote, unchecked | No | You own what you send |
| Asking AI to explain a public regulation in plain English | Fine, then check | Read the original before relying on it |
| Using a template you brought from your last employer | Ask first | It may be their confidential information |
The first-week plan
This is our own suggested plan for a desk-based role, and the times are our judgement. It asks for just under three hours of the manager's time across the week. Adjust it for your setting.
| When | Activity | Led by | Time | What exists at the end |
|---|---|---|---|---|
| Before day 1 | Create work accounts and tool access; draft the first-week plan from the job description; name a buddy | HR, ICT and manager | 45 minutes | Login details and a plan |
| Day 1, morning | Personal welcome; introductions; what the role is for | Manager | 45 minutes | A clear first goal |
| Day 1, afternoon | AI rules card and scenario check | HR or ICT | 30 minutes | Signed card and questions answered |
| Day 2 | Sit with the buddy; see real work; note down unfamiliar terms | Buddy | Half a day | A personal glossary |
| Day 2 | Use an approved tool to explain a non-confidential SOP | New employee | 20 minutes | A first safe use of AI |
| Day 3 | Hands-on: the four Cs on one real, low-risk task | Buddy or AI champion | 45 minutes | One prompt the person wrote |
| Day 3 | Check that output together and find the errors | Buddy | 20 minutes | The habit of checking |
| Day 4 | Complete a small piece of real work with AI, then review it with the manager | Manager | 30 minutes | Feedback on real work |
| Day 5 | Week-one conversation: what is clear, what is not, what they need | Manager | 45 minutes | A plan for weeks 2 to 4 |
The four Cs are how we teach prompting: Clarity (what exactly you want), Context (what the tool cannot guess), Constraints (limits and what to avoid) and Customisation (who will read it and where it will be used).
Days 8 to 90
- Week 2: Add role-specific AI uses from the team's prompt library.
- Day 30: The manager reviews one AI-assisted piece of work and asks what the employee has stopped checking.
- Day 60: The employee teaches the team one thing that worked, or one error they caught.
- Day 90: Before any probation review, ask the employee what confused them about the rules, then fix the card.
One induction session is only a beginning. BCG's 2025 survey of more than 10,600 people found regular AI use was sharply higher among those with at least five hours of training and access to in-person training and coaching. Our guide on how learning and development teams use AI covers follow-up. Managers set the tone here, so our piece on getting middle managers to use AI is worth reading too.
Adapting it for different new starters
Different people need different versions of the same rules.
| New starter | What changes |
|---|---|
| Shift or field staff without laptops | Print the card. Use a phone-based tool only if approved. Plan offline copies where the connection is weak |
| Interns and attachment students | Same rules and closer supervision. Give them no access to personal data they do not need |
| Contractors and consultants | Put the AI rules and confidentiality in the contract, and check who owns the work |
| New managers | Add how to review AI-assisted work and how to respond when someone admits a mistake |
| Confident AI users | Focus on data, your approved tools, and leaving old employers' material behind |
| People who fear AI | Start with one small task and a kind buddy. Nobody becomes an expert in a week |
Six prompts you can copy
Use approved documents and remove names and confidential details first.
1. First-week plan from a job description
Below is the job description for a new [role] with names removed. Draft a five-day onboarding plan as a table with columns: Day, Activity, Led by (job title only), Time needed, What exists at the end. Context: we are a [sector] organisation in Kenya of about [number] staff, and the manager has about three hours to give this week. Constraints: include a 30-minute AI rules session on day one and one supervised use of AI by day three. Do not invent systems or colleagues that are not mentioned; write [CHECK] instead. Customisation: a busy manager will print this.
2. Handbook section into an FAQ
Below is a section of our staff handbook. Turn it into 10 questions a new employee would really ask, each with an answer of no more than three sentences. Constraints: use only what the text says, give the section number for each answer, and write "Ask HR" where the text is silent. Do not give legal advice. Customisation: plain English for someone in their first week. Text: [paste]
3. Scenario quiz on your AI rules
Below are our AI rules. Write eight short workplace scenarios for new staff in [department], each with an answer of Fine, Not in a public tool, or Ask first, and a one-sentence reason. Constraints: use invented, realistic Kenyan workplace situations with no real names. Base each answer only on the rules provided, and write [CHECK] where the rules do not settle it. Customisation: suitable to talk through with a manager in 15 minutes. Rules: [paste]
4. Glossary from internal documents
From the text below, list every acronym, system name and piece of internal language a newcomer might not understand. Give a table with Term, Where it appears, and Meaning. Constraints: if the text does not explain a term, write "Ask a colleague". Do not guess. Customisation: for someone in their first week. Text: [paste]
5. For the new employee: explain a process
I am new at work and learning this process. Below is a non-confidential SOP. Explain it in plain words, then list five questions I should ask my buddy to check I have understood it. Constraints: use only what the SOP says and tell me where it is unclear. Do not fill the gaps with general advice. Customisation: short sentences. SOP: [paste]
6. Manager check-in questions
Suggest questions a line manager can ask a new [role] on day one, on day five and at day 30. Context: the aim is to find out what is unclear and whether they feel safe asking for help, including about AI use at work. Constraints: maximum five questions at each point, open questions only, and no questions about health, family or personal life. Customisation: natural spoken English for a Kenyan workplace.
Should you build an onboarding chatbot?
Probably later. A chatbot that answers staff questions from your documents helps only when those documents are current and someone owns them.
Before building one, check these:
- Your handbook and SOPs are up to date, with one version of each.
- The tool is approved, and you know where the data goes.
- It answers from your documents and shows which one it used.
- It says "I do not know" and points to a person.
- A named owner tests it after every policy change.
- It is kept away from payroll, medical and disciplinary records.
For many smaller organisations, a good FAQ document and a buddy will do the same job with much less effort.
Which tools to use, and what they do with your data
Show new staff the free version of whatever you approve, so that they can practise. Facts here are as of November 2026, and these products change often.
- Microsoft Copilot Chat. With a work account, Microsoft says enterprise data protection applies and that prompts and responses are not used to train its foundation models. It can use files you upload. Full access to work email and files needs a paid licence.
- ChatGPT. A guide from the legal software company Spellbook, updated on 5 November 2026, says Free, Plus and Pro chats are trained on unless you opt out, and that Business and Enterprise plans are excluded by default. We could not open OpenAI's own help pages when we checked, so confirm this there before relying on it.
- Gemini. Google says some chats are read by human reviewers, that reviewed chats are kept for up to three years, and that you should not enter confidential information. Work and school accounts may have different terms.
- Claude. Anthropic offers a Free plan with limited usage. On Free, Pro and Max plans, a privacy setting controls whether chats train its models. Commercial plans are excluded.
The practical lesson for new staff: personal accounts and work accounts can be treated differently, so work belongs in the work account.
What goes wrong, and how to avoid it
| What goes wrong | How it shows up | What to do |
|---|---|---|
| AI answers from an old policy | A new person follows a rule you changed last year | One current version of each document, with an owner and a review date |
| A warning and nothing more | A slide of prohibitions, and then silence | Give one safe task to try by day three |
| Automated welcome | A new person gets messages and meets nobody | Manager meets them on the first morning |
| Personal documents in a public tool | HR pastes contracts to save time | Use templates with invented details |
| Quiet use of old tools | The new hire keeps a personal account | Explain why, make the approved tool easy to reach and check in at day 30 |
| Fear of admitting a mistake | A leak is discovered weeks later | Tell people early that honest reporting is expected and protected |
| Rules that go stale | The card names a tool you dropped | Review the card every six months and after each new tool |
How to measure it
Keep this simple and honest. We found no independent figure for time saved on onboarding in Kenyan organisations.
- Ask each new starter at day 30: "Do you know which AI tools you may use, and who to ask?"
- Track whether scenario answers improve.
- Record how long HR takes to prepare an induction, including checking.
- Note which questions people keep asking, and add them to the FAQ.
- Count how many data mistakes are reported early. A rise in early reporting can be a good sign.
Questions people ask
Should AI be part of new employee onboarding?
Yes, if your staff will meet AI at work, as many now do. Teach the basic rules during induction and let each department add role-specific skills later. Keep the human parts of onboarding in human hands.
Can HR use ChatGPT to write onboarding plans?
Yes, if you use a job description with names removed and a manager checks the result. Keep contracts, ID copies and salary details out of public tools. Check your organisation's approved tools first.
How long should AI rules training take in induction?
Thirty minutes on day one is a realistic start, followed by a supervised task by day three. More practice should follow. BCG found regular use was much higher among people with at least five hours of training.
What if we do not have an AI policy yet?
Write a short one first, or use the card above as a temporary version approved by management. A new hire with no guidance will make up their own rules.
Can AI judge whether someone passes probation?
The manager must make that judgement. Section 35 of the Data Protection Act protects people from significant decisions based solely on automated processing, and a Kenyan court has stressed the right to be heard during probation. Get legal advice for any difficult case.
Do interns and casual staff need the same rules?
Yes, the core rules apply to everyone who touches your information. Adjust the format. A printed card and a short conversation may work better than a slide deck.
Should we allow new hires to use their own AI tools?
Only tools your organisation has approved. If a new hire shows you a better tool, send it through your approval route and keep it from spreading informally.
Sources
All sources last checked 10 November 2026. Section 42 comes from the Employment Act, Revised Edition 2022. The Kibuchi case comes from a law firm article and a press report, since we could not open the judgment. Confirm the position on Kenya Law before relying on it for a legal decision.
- Capital FM via AllAfrica, "From AI Pilots to Profits" (25 August 2026), reporting PwC research
- KnowBe4, Generative AI in the Middle East and Africa Survey 2024
- UNLEASH, report on the TalentLMS and BambooHR onboarding survey (25 July 2025)
- Microsoft, 2024 Work Trend Index announcement (8 May 2024)
- Gallup, onboarding new employees in the remote work era (12 August 2021)
- Brynjolfsson, Li and Raymond, "Generative AI at Work", NBER Working Paper 31161
- BCG, "AI at Work: Momentum Builds, but Gaps Remain" (26 June 2025)
- The Data Protection Act, No. 24 of 2019 (ODPC)
- EY Tax News, ODPC ruling on employees' images, Complaint 1618 of 2024 (30 October 2025)
- The Employment Act, No. 11 of 2007, Revised Edition 2022
- The Standard, "It's unconstitutional to fire staff on probation at will, court rules" (3 August 2021)
- ENS via Legal 500, "Terminating probationary contracts without a hearing declared unconstitutional" (11 August 2021)
- Microsoft Learn, Copilot Chat privacy and protections
- Microsoft Support, Copilot Chat with and without a licence
- Spellbook, "Is ChatGPT Private?" (updated 5 November 2026)
- Google, Gemini Apps Privacy Hub
- Anthropic, updates to consumer terms (28 August 2025)
- Anthropic, Claude pricing
Where FuKazee fits in
FuKazee trains teams to use AI on the work in front of them, and has trained more than 3,000 people since 2022. We assume nobody in the room knows AI yet, we avoid jargon, and we show the free version of every feature. We never promote one tool, and Microsoft 365 and Copilot are always part of the programme.
For onboarding, we can work with your HR team on the rules card, the scenarios and the first-week plan, using your policies and invented data. Training will not fix an induction that has no manager involved, so we start by asking who will actually be in the room with the new person.
See our AI training for HR teams in Kenya, our HR guide, or our wider AI training in Kenya.
To discuss AI onboarding for your new staff, email us at 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.
