On this page
TL;DR:
- This guide is for owners and directors. Generative AI is useful today for first drafts, summaries, spreadsheet help and sorting information, as long as a person checks the result.
- It cannot be left alone with legal, tax or financial decisions or with promises to customers, and it will not fix a weak business.
- The best Kenyan evidence is a warning. In a trial with 640 Kenyan entrepreneurs, the AI assistant made no clear difference on average. Weaker businesses did nearly 10% worse, and stronger ones may have gained over 15%.
- Your business is ready to start if you have a repetitive task, someone to own it, and someone able to check the work. Use the 10-question test below.
- Start with one task, free tools and a one-page rule. Measure for 30 days before you spend money.
Who this guide is for
This guide is for the person who answers for the business: an owner, a managing director or a board member of a company with a handful of staff or a few hundred. You do not need to understand the technology.
By the end, you should be able to decide whether to act now, what to try first, and what to keep away from AI.
If you run a very small business and mainly want a list of free tools, go to our guide to free AI tools for small businesses in Kenya.
What is generative AI, in plain words?
Generative AI is software that produces new text, images, audio or code when you give it an instruction. ChatGPT, Claude, Gemini and Microsoft Copilot are well-known examples.
It works by predicting what is likely to come next, based on the material it learned from. The US National Institute of Standards and Technology (NIST) notes that this design can produce confident but false content. So think of it as a fast junior assistant who has read a great deal, knows nothing about your business until you explain it, and sometimes makes things up without warning.
Your staff are probably using it already. DataReportal's Digital 2026 Mid-Year Global Update (April 2026) found that 97.5% of people surveyed in Kenya had used at least one AI tool in the past month, the highest of the 54 countries in a survey of people aged 16 and over. That includes AI built into everyday products such as Microsoft Office and Canva. Tech-ish, reporting on DataReportal's Digital 2027 report (6 November 2026), says 18.5% of Kenyans online aged 16 and over pay for an AI subscription each month, against 9.5% worldwide. These are surveys of people online, so they do not describe every Kenyan. Other measures are much lower: Microsoft's AI Diffusion report, also as reported by Tech-ish, put Kenya at 8.9% of people aged 15 to 64 in the second quarter of 2026.
The practical question for a director is whether that use happens with your guidance or without it.
What can it do for a business today?
It can save time on work that is repetitive, mostly text-based and easy for someone to check. Here is how that looks across a typical business.
| Area | What it does well | Who must check | Risk if nobody checks |
|---|---|---|---|
| Marketing | First drafts of posts, product descriptions, adverts in English and Kiswahili | Whoever owns the brand | Medium: wrong claims, odd Kiswahili |
| Customer care | Draft replies to common questions, sorting complaints by type | A trained staff member, before sending | High: wrong promises |
| Sales | Proposal outlines, follow-up messages, meeting preparation | The salesperson | Medium |
| Finance | Explaining Excel formulas, drafting variance notes | Your accountant | High: wrong figures |
| HR | Job adverts, staff notices, draft policies | HR lead or lawyer | High if it touches individuals |
| Operations | Writing procedures from rough notes, creating checklists | The person who does the job | Low to medium |
| Management | Summarising long reports, arguing against your plan | You | Medium |
These ratings are our own judgement. Adjust them for your business.
What does the research say about the gains?
The gains are real on suitable tasks, and they are uneven.
- A study of 5,179 customer support agents, published as a working paper by the US National Bureau of Economic Research, found that an AI assistant raised issues resolved per hour by about 14% on average. Novice workers improved by 34%. Experienced workers saw little change.
- In an experiment with 758 Boston Consulting Group consultants, those using AI on suitable tasks finished 12.2% more of them and worked 25.1% faster. On a task picked to be beyond the tool's abilities, they were 19 percentage points less likely to get the right answer.
Both studies date from 2023, and neither looked at Kenyan firms. Newer tools may do better. Treat these as signs of the pattern, and measure your own results before believing any figure, including one from a training company.
What can it not do?
It cannot take responsibility, and it cannot know what you have not told it. In practice, that leaves several things you should not expect.
- It cannot make your decisions. It can argue both sides. You carry the consequences.
- It cannot be your lawyer or tax adviser. Kenyan courts have already met this problem. Justice Isaac Lenaola of the Supreme Court has warned advocates after a case involving AI-generated fake authorities, according to Strathmore University's CIPIT. A draft Judiciary policy, reported in May 2026, would make advocates verify every authority they cite.
- It cannot be trusted with numbers on its own. Use it to explain how to do a calculation, and let the spreadsheet do the calculating.
- It cannot speak for your company without supervision. In Moffatt v Air Canada (2024), a Canadian tribunal held the airline responsible when its website chatbot gave a customer wrong information about fares. That is a foreign decision, and a Kenyan court may see things differently. The commercial lesson still travels.
- It cannot be relied on for flawless work under your name. In 2025 Deloitte agreed to partly refund the Australian government over a report with a fabricated quote from a judge and references to research that did not exist. Deloitte did not say whether AI caused those errors. The corrected report disclosed that AI had been used.
- It does not know your stock, your customers or your margins unless you give it that data, and giving it that data brings its own duties.
- It will not repair a weak business. A cash flow problem or a poor product stays the same after AI has written a better report about it.
What does the Kenyan evidence show?
The most relevant study was done here, and it should cool any excitement about AI as a shortcut.
A research team that included Harvard and Berkeley academics ran a five-month trial in 2023 with 640 Kenyan small business owners, in trades such as fast food, poultry and cyber shops. Half got a GPT-4 business mentor they could use through WhatsApp. The others received links to business training guides. The results were published in the journal Management Science.
| Group | Effect of the AI mentor |
|---|---|
| All businesses, on average | No clear effect on revenue or profit |
| Businesses already doing well | May have gained over 15% |
| Businesses already struggling | Did nearly 10% worse |
The paper says the gap does not appear to come from the questions owners asked or the advice they got. It points to the advice they chose to act on. Stronger owners were more likely to pick suggestions tailored to their own business. Weaker owners more often took up general ideas, such as cutting prices or advertising, that can hurt performance.
We draw two lessons from this. Both are our reading of the study.
- AI rewards judgement you already have. The owner who knows which advice fits their situation gets more from it.
- Be careful with general advice. Ask AI for options, then test each one against your own costs and customers before you act.
Note the limits. The study used a 2023 model and small businesses, so a larger firm or a newer tool could get different results.
What rules apply to a Kenyan business?
Kenya had no dedicated AI law as of 10 November 2026, and the Data Protection Act already applies. As a director, you need only the short version.
| Item | Status | What it means for you |
|---|---|---|
| Data Protection Act, 2019 | In force | Section 63 allows penalties of up to KES 5 million, or 1% of annual turnover for an undertaking, whichever is lower |
| Section 35 of the Act | In force | People have a right not to face significant decisions made solely by automated processing, with some exceptions |
| Section 48 of the Act | In force | Sets conditions for sending personal data out of Kenya, which matters because many AI tools are run from abroad |
| ODPC draft Guidance Note on AI | Draft | Reported to treat generative AI platforms as high risk and to expect impact assessments |
| Artificial Intelligence Bill, 2026 | Bill only | A Senate committee rejected it in October 2026. We found no report of a full Senate vote |
| CBK draft Guidance Note on AI in banking | Consultation closed 7 November 2026 | Relevant to banks. Watch for the final text |
This is general information, and you should take legal advice on your own position. Our guide to the Data Protection Act and AI goes further.
Is your business ready?
Answer these ten questions with a plain yes or no. The scoring is our own rule of thumb.
| # | Question |
|---|---|
| 1 | Can you name one task that happens every week and is mostly writing, reading or sorting? |
| 2 | Could someone in the business tell a good result from a bad one? |
| 3 | Is there one person with time to own a small trial? |
| 4 | Do you know, roughly, which AI tools staff already use? |
| 5 | Have you written down what must never go into a public AI tool? |
| 6 | Do you know what personal data your business holds and where it sits? |
| 7 | Are your key documents and records organised enough to find quickly? |
| 8 | Do you have reliable internet and devices where the work happens? |
| 9 | Can you say how long the chosen task takes today? |
| 10 | Will you, as the leader, use it yourself and accept honest reports of mistakes? |
- 8 to 10 yes: Start the 30-day plan this week.
- 5 to 7 yes: Start, and fix the gaps during the first two weeks.
- Fewer than 5 yes: Get the basics in place first. Questions 3, 5 and 6 matter most.
For a fuller check with a score across usage, skills, data and leadership, try our free AI readiness self-assessment.
What should you do first?
Do one small thing well. A company-wide launch is usually the wrong first move.
Choose the first task
Score candidates from 1 to 3 on each line, then pick the highest total.
| Test | 1 | 2 | 3 |
|---|---|---|---|
| How often does it happen? | Monthly | Weekly | Daily |
| How safe is the data? | Personal or confidential | Can be cleaned | Public or harmless |
| How easy is it to check? | Needs an expert | Takes some effort | Quick to verify |
| What if it goes wrong? | Serious harm | Embarrassing | Easily fixed |
Good first tasks tend to be internal: turning notes into procedures, drafting staff notices, or summarising public reports. Keep customer-facing automation for later.
Follow a 30-day plan
| When | What to do | Result |
|---|---|---|
| Week 1 | Pick the task and owner. Time the task as it is done today. Write a one-page rule covering approved tools, what must stay out and who signs off | A baseline and a rule |
| Week 2 | Owner and two colleagues try the task on free tools. Save the prompts that work | A small prompt collection |
| Week 3 | Do the work with AI and a named checker. Log each error | An error log |
| Week 4 | Compare the time, including checking. Ask the checker about quality. Decide to stop, continue or add a second task | A decision based on your own numbers |
Count the checking time. A draft that takes two minutes to make and forty minutes to repair has saved nothing. Our guide to evaluating ROI on AI training has a fuller method.
What will it cost to start?
You can start for nothing, and you should. Each assistant below has an option at no extra cost, though Copilot Chat needs an eligible Microsoft 365 plan. The larger cost is your staff's time. Prices are as of November 2026, taken from each company's own page. They change often.
| Tool (A to Z) | Free option | Paid examples |
|---|---|---|
| ChatGPT | Free plan with limits | OpenAI's page showed no amounts when we opened it. Business starts at 2 users |
| Claude | Free plan with limits | Pro: $17 a month billed yearly, or $20 monthly. Team: from $20 a seat a month billed yearly |
| Gemini | Free plan with a Google Account | Google AI Plus: KES 900 a month. Google AI Pro: KES 3,700 a month |
| Microsoft Copilot Chat | No additional cost with eligible Microsoft 365 plans | Microsoft 365 Copilot Business add-on: $21 a user a month on a yearly plan, with a promotional $18 for eligible customers until 31 December 2026 |
Dollar prices may differ in Kenya, and taxes may be added. Before paying for a business plan, ask what happens to your data on that plan and on the free one. The answers differ by product.
For planning money across a whole organisation, see how to budget for AI training.
When is staff training worth paying for?
Training is worth paying for once you know which work you want to improve and at least a few people have tried it. Before that point, you risk paying to answer questions nobody in the business has asked yet.
It tends to help when staff use AI unevenly, when mistakes or data slips are starting to appear, or when you have paid licences nobody uses.
It will not help if nobody has time to practise, if there are no rules, or if the real problem is an unclear process. Fix those first.
When you are ready, compare options on our pages about generative AI training in Kenya and AI training for small businesses. If you lead a larger firm, read what a CEO should ask before an AI rollout.
Three prompts a director can try this week
Use them in any assistant. Do not include customer names, staff details, bank details or confidential figures. Each prompt follows the four Cs we teach: Clarity, Context, Constraints and Customisation.
Find your first task
I run a [type of business] in [town], Kenya, with [number] staff. Here is a rough list of what my team does in a normal week: [list, with no names or figures]. Identify the three tasks most suited to help from a generative AI tool. Choose only tasks that are repetitive, low risk and easy for a person to check. For each, say why it fits, what could go wrong and who should check the output. Keep it to one page in plain English.
Test a decision before you make it
I am considering [decision, without confidential figures]. Act as a cautious board member who knows the Kenyan market. Give me the five strongest reasons not to do this and the three facts I should confirm first. Tell me where you are unsure. Do not invent statistics. If you have no reliable figure, say so.
Write your one-page AI rule
Draft a one-page AI use rule for a [type of business] in Kenya with [number] staff. Cover which tools are allowed, what must never be pasted into a public AI tool, who checks work before it goes out, and who to tell if something goes wrong. Write for staff with no technical knowledge. Use short sentences. Mark anything my lawyer should review with [LAWYER].
What goes wrong?
| Mistake | Why it hurts | Better approach |
|---|---|---|
| Buying licences first | Money goes out before anyone knows the use | Trial on free tools for a month |
| Starting with a customer chatbot | Errors reach customers directly | Start internally |
| Announcing job cuts on day one | Staff hide problems and stop helping | Be clear about your intentions and honest about what you do not yet know |
| Banning it outright | Staff may carry on using it on their phones, unseen | Give rules and approved tools |
| Believing vendor figures | Their case is a different business from yours | Measure your own task |
| Acting on general AI advice without testing it | In the Kenyan trial, weaker businesses more often took up general ideas, and they did worse | Begin with narrow, specific tasks, and test any advice against your own numbers |
Questions people ask
Is generative AI worth it for a small business in Kenya?
It can be, for specific repetitive tasks, and the free versions make it cheap to find out. The Kenyan trial of 640 entrepreneurs found no clear average gain from an AI mentor, so do not expect automatic results. Try one task and measure it.
Will generative AI replace my staff?
For most small and mid-sized businesses, the near-term effect is more likely to be on tasks than on whole jobs, in our view. AI cannot take responsibility for a result, so someone must still check it. Be open with your team about your plans, because fear leads people to hide what they are doing.
Can I put my customer data into ChatGPT or similar tools?
Be very careful about putting it into a public tool. Kenya's Data Protection Act covers personal data, and section 48 sets conditions for moving it abroad. Use cleaned data or a tool your business has assessed, and take advice if the data is sensitive.
Do I need a data scientist to start?
No. Drafting, summarising and spreadsheet help need no technical staff. You will want specialist help if you connect AI to your own systems or customer records.
How long before we see results?
You can tell within about 30 days whether one task is improving, if you timed it beforehand. Wider change takes longer because habits change slowly.
Should we use one tool for the whole company?
Often yes, if you already pay for a suite. Businesses on Microsoft 365 can look at Copilot Chat, and those on Google Workspace can look at Gemini. Test on your own work before committing. In our view no single tool is best at everything.
Is there a law against using AI in my business in Kenya?
We found none as of 10 November 2026. The Data Protection Act applies whenever personal data is used, and sector regulators may add their own rules.
Sources
- Otis and others, "The Uneven Impact of Generative AI on Entrepreneurial Performance: Evidence from a Field Experiment in Kenya", Management Science, open copy on eScholarship, last checked 10 November 2026
- Brynjolfsson, Li and Raymond, "Generative AI at Work", NBER Working Paper 31161, last checked 10 November 2026
- Dell'Acqua and others, "Navigating the Jagged Technological Frontier", working paper 24-013 (2023), last checked 10 November 2026
- NIST, Generative AI Profile (NIST AI 600-1, July 2024), last checked 10 November 2026
- DataReportal, Digital 2026 Mid-Year Global Update Report, last checked 10 November 2026
- Tech-ish, report on DataReportal's Digital 2027 and Microsoft's AI Diffusion report (6 November 2026), last checked 10 November 2026
- The Data Protection Act, No. 24 of 2019, Gazette copy on the ODPC website, sections 35, 48 and 63. This is the 2019 text. We found no report that these sections have changed, and Vellum reported a draft amendment Bill in May 2025; last checked 10 November 2026
- PPC Land, report on the ODPC's draft Guidance Note on AI, a secondary source because the ODPC document did not open for us; last checked 10 November 2026
- Techweez, "Senate committee rejects AI Bill, sending it back to the drawing board" (16 October 2026), last checked 10 November 2026
- Digital Policy Alert, CBK consultation on draft AI guidance for banks, last checked 10 November 2026
- Strathmore CIPIT, "AI as the side kick and not superhero: navigating AI in the legal field in Kenya" (5 February 2026), last checked 10 November 2026
- Nairobi Law Monthly, "Judiciary moves to regulate AI in courts" (28 May 2026), last checked 10 November 2026
- DWW, "BC Tribunal finds Air Canada liable for inaccurate advice given by website chatbot", a law firm summary of Moffatt v Air Canada, 2024 BCCRT 149, not the decision itself; last checked 10 November 2026
- AP report via WJXT, "Deloitte to partially refund Australian government for report with apparent AI-generated errors" (7 October 2025), last checked 10 November 2026
- OpenAI, ChatGPT plans and pricing, last checked 10 November 2026
- Anthropic, Claude pricing, last checked 10 November 2026
- Google, Gemini subscriptions, last checked 10 November 2026
- Microsoft, Microsoft 365 Copilot pricing, last checked 10 November 2026
Where FuKazee fits in
FuKazee trains teams in Kenyan organisations to use AI on the work in front of them. We assume nobody in the room knows AI yet, we avoid jargon, and we say plainly that nobody becomes an expert in a few days. We show the free version of every feature and never promote one tool. Since 2022 we have trained more than 3,000 people.
You can follow every step in this guide without us. If you get to the point of wanting help with choosing the first task, setting the rules or training your people, we would be glad to talk.
To discuss what makes sense for your business, 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.
