On this page
TL;DR:
- We found no study that measures jobs lost to AI in Kenya. What exists are global estimates of which jobs are "exposed", and exposure is different from job loss.
- Those estimates put less risk on poorer economies. The World Bank's 2026 report says 4.5% of jobs in low- and middle-income countries are at risk of automation, against 14.2% in high-income countries.
- Lower risk overall still leaves a sharp problem. Clerical and office jobs are among the most exposed, and in Kenya formal jobs are scarce.
- Data labelling has brought income to Kenyans and has also brought low pay, thin contracts and sudden layoffs. Pay, conditions and responsibility are all disputed, some of it in court.
- For employers, the practical steps are to protect entry-level jobs, redesign tasks with staff and talk honestly about what is unknown.
Who this guide is for
This guide is for managers and leaders in Kenyan organisations who are asked, "Will AI take our jobs?" It sets out what has been measured, what has only been estimated and what nobody knows yet.
This guide makes no single prediction, because the evidence does not support one.
What has actually been measured?
Very little about Kenya specifically. Most figures in the news come from global models that score occupations by their tasks. They were not built from counts of Kenyan workers who lost jobs. Keep that in mind when you read this table.
| Study | What it says | Its limits | How we would use it |
|---|---|---|---|
| ILO and Poland's NASK, Working Paper 140 (May 2025) | One in four workers worldwide are in jobs with some generative AI exposure. Exposure is 11% of employment in low-income countries and 34% in high-income countries. "Transformation of jobs is the most likely impact" | Estimates task exposure. It gives no count of jobs lost and no Kenya figure | A global starting point |
| ILO and World Bank, "Disruption without dividend?" (March 2026) | Covers 135 countries. In developing economies, disruption may arrive before productivity gains. Many workers who could benefit lack reliable internet | A projection. We read the ILO's summary, which gives few numbers | A useful warning for countries like Kenya |
| World Bank, World Development Report 2026 (August 2026) | 4.5% of jobs in low- and middle-income countries at risk from generative AI, against 14.2% in high-income countries. 16.2% and 18.7% could see productivity gains | Grouped by income. The press release does not mention Kenya | The most recent headline numbers |
| IMF staff analysis (January 2024) | AI exposure of about 60% of jobs in advanced economies, 40% in emerging markets and 26% in low-income countries | Older, covers AI in general, and "exposed" includes jobs that may benefit | Explains why IMF figures look higher than the World Bank's |
| World Bank Digital Progress and Trends, as reported by The Star (April 2026) | Job vacancies requiring AI skills in Kenya grew about fourfold from 2021 to 2024 | From "a very low base". No absolute numbers in the news report we read | A sign of new demand, on a small scale |
| Stanford Digital Economy Lab, "Canaries in the Coal Mine" (revised August 2026) | Employment of US workers aged 22 to 25 in AI-exposed jobs is 19% below where it would be had it kept pace with less-exposed peers | US payroll data. The authors call these early, descriptive indicators. Some trends started before generative AI | A warning about entry-level hiring, and untested in Kenya |
| Yale Budget Lab and others, as summarised by Stanford (2026) | Little relationship between AI exposure and changes in overall US employment | US only and looks at the whole economy, so it could miss young workers | The main counterweight to Stanford |
| WageIndicator and Data Labelers Association (2026) | Survey of Kenyan data workers on pay and conditions | 57 responses, members of one association, during layoffs | The closest thing to direct Kenyan data on this work |
Why do the percentages differ so much? They measure different things. The IMF counts any job AI could affect, for better or worse. The World Bank's 4.5% counts jobs at risk of automation. Neither is a forecast of how many people will be out of work on a certain date.
Why might Kenya look different from rich countries?
Because most Kenyans do not work in the kind of job that generative AI affects first.
KNBS's Economic Survey 2026, as reported by The Star, counted 21.6 million people in employment in 2025, excluding small-scale agriculture. Kenya added 822,100 jobs that year, and over 87% of them were informal. Formal employment reached 3.5 million, which is about one job in six.
The ILO and World Bank paper says workers in lower-income economies do fewer non-routine analytical tasks and rely less on computers. Our reading is that a boda boda rider, a market trader or a smallholder is unlikely to be replaced by a chatbot soon. That reading is an inference, and we found no Kenyan study testing it.
Why is lower exposure not the same as low risk?
Because the jobs most exposed are the ones Kenya can least afford to lose. The ILO names clerical and administrative roles among the most exposed. Its UN summary says those jobs are unlikely to disappear entirely, but partial automation could mean fewer responsibilities, flat wages and more insecurity. Worldwide, 4.7% of women's jobs fall in the highest exposure category, against 2.4% of men's, largely because more women work in clerical roles.
The ILO and World Bank paper adds a point that matters for Nairobi. In lower-income countries, automation-prone jobs are often the better jobs: clerical and administrative work that has given women and young people a path to decent employment. Those workers tend to be online already, so disruption can reach them first. Meanwhile, many workers in roles that could gain from AI lack reliable internet.
The UN summary of the ILO's work makes the same point about sub-Saharan Africa. Lower exposure does not mean lower risk where digital access is limited or labour protections are weak.
| Type of work | Exposure signal | What could change | Our reading for Kenya |
|---|---|---|---|
| Clerks, data entry, admin assistants | High in the ILO's work | Routine typing, formatting and scheduling shrink | The group to worry about most |
| Customer service, call centres | Stanford cites customer service among affected US jobs | Basic queries go to bots | Directly affects outsourcing jobs |
| Accountants, analysts, writers | Professional work is exposed | First drafts get faster and juniors do less of them | The training ladder is the concern |
| Teachers, nurses, sales staff | Mixed | Paperwork changes more than the core job | Likely to gain help before losing work |
| Transport, construction, farming | Lower | Little direct change from generative AI | Other technology may matter more |
The last column is our judgement, and it is not a research finding.
Is AI hurting young graduates?
We do not know for Kenya, and the debate in the United States is unsettled. This matters because Kenya's first-job problem is already serious.
The worrying evidence. Stanford's payroll study found the 19% gap for US workers aged 22 to 25, mostly through less hiring and not through more firing. The World Economic Forum says 37% of young workers globally are in jobs with medium to high exposure to AI-driven task change, as reported by The Star. The Forum advises employers to redesign entry-level roles and keep hiring graduates.
The other side. Stanford's own research note summarises work by the Budget Lab at Yale, which finds little relationship between AI exposure and overall US employment. Stanford's authors also say their pattern weakens when they control for education, and that some of the trends began before generative AI arrived.
The Kenyan picture. The Star reports, citing CUE and KNBS data, that 500,000 to 800,000 graduates enter the labour market each year, and employers already cite lack of experience as a barrier. We found no study that tracks whether AI is making this worse here. If junior tasks get automated without any plan, it is reasonable to worry, but that remains a concern and has not been measured.
What is the data-labelling story?
Kenya has become a known location for the human work behind AI: tagging images, rating text, and reviewing harmful content so that systems can learn. It deserves a fair account, because it has brought both income and harm.
What has been good
- It has brought work. The Ministry of ICT said in 2025 that the wider outsourcing sector had created 19,650 jobs, above a target of 10,000, as reported by Capital FM. That figure covers all outsourcing, so it includes much more than AI work. We found no reliable national count of data labellers.
- In one small survey, most earned at least minimum wage levels. In the WageIndicator survey, 94% of women and 72% of men earned at least the relevant minimum wage, although these minimum wage rules do not formally cover this work.
- Graduates can find a way in. Most respondents held a bachelor's degree, in a market where first jobs are hard to get.
- Companies say they provide support. Sama has said it pays four times the local minimum wage and offers unlimited mental health support. Meta has said its contractors must pay above the industry standard and provide on-site support from trained practitioners.
- Workers have organised. The Data Labelers Association, with about 1,100 members as of June 2026, gives workers a collective voice.
What is contested
| Issue | One account | Another account | Status |
|---|---|---|---|
| Pay on ChatGPT safety work (2021 to 2022) | TIME reported workers took home about US$1.32 to US$2 an hour, while OpenAI paid Sama US$12.50 an hour | Sama said workers earned US$1.46 to US$3.74 after tax, and that US$12.50 covered all costs including infrastructure | Disputed. TIME interviewed four workers |
| Current pay and security | WageIndicator found average monthly pay of Ksh 21,850 for women and Ksh 26,107 for men, with only about 2 in 10 men and no women earning a living tariff | Sample of 57, members of one association, surveyed during layoffs. Pay was not adjusted for hours | Treat as a signal only |
| Workload and mental health | AP reports that about 200 former moderators sued, alleging low pay, poor mental health support and long hours, and seek US$1.6 billion | Sama and Meta point to the support described above. Meta says it did not directly hire or fire these workers, which they dispute | In court. Rulings due on 12 February 2026 were adjourned. On 27 April 2026 Capital FM reported a ruling in favour of 187 former moderators, including an order that a Meta director appear for cross-examination. It reported no final finding on liability, and we found no later outcome as of 10 November 2026 |
| The 2026 layoffs | WeeTracker reported that a Kenyan workers' group alleged the job losses followed staff speaking out about sensitive footage from smart glasses | WeeTracker also reported that Meta told the BBC Sama's work did not meet its standards. Sama defended its quality and said it was never told of any failure | Disputed |
The clearest recent fact is the scale of the disruption. On 16 April 2026, AP reported that Sama had issued a layoff notice to 1,108 staff after Meta ended a major engagement at its Nairobi office. A single client's decision affected more than a thousand jobs, and any work tied to a few overseas buyers carries that risk.
WageIndicator's survey suggests how little cushion there is. Only four in ten respondents had a contract with their platform or outsourcing firm. Over eight in ten worried about job security, and six in ten found the work mentally exhausting. Remember the limits of that sample.
What does Kenyan law say about these workers?
The law has not settled this. Strathmore University's CIPIT published a report in 2025 arguing that many data annotators are classed as independent contractors, which keeps them out of protections such as minimum wage, paid leave and social security. It also records that the Employment and Labour Relations Court ruled in 2023 that Meta and its contractors could be sued in Kenya.
Be careful with one name. Two different Bills were called the Business Laws (Amendment) Bill, 2024:
- National Assembly Bill No. 49 of 2024 became law. Mzalendo records presidential assent on 11 December 2024.
- Senate Bill No. 51 of 2024 is the one that would change employment law for outsourcing and remote workers, and it is still a Bill. The National Assembly's Bills Tracker, as at 15 October 2026, shows the House passed it at Third Reading on 19 August 2026 and referred it back to the Senate for reconsideration on 20 August 2026. It records no presidential assent, and we found no later report as of 10 November 2026.
According to Mzalendo's copy of the Bill as published, Senate Bill No. 51 would widen "employee" to include people working through outsourcing arrangements, and require outsourcing firms to provide tools and answer for claims. Because it went back to the Senate, the final wording may differ. Views differ too. Labour leaders told The Standard in July 2026 that some clauses favour foreign investors and do too little for workers hired online. Lawyers at Spencer West, writing in November 2024 about a Bill of this name with the same employment clause (their note gives no Bill number), warned that the wider definition of "employer" could expose managers to personal liability. In June 2026 a High Court judge dismissed a challenge by Oversight Lab and 35 other petitioners as premature, because a Bill is not yet law.
How should you read the next scary headline?
Ask these five questions before you share it or base a decision on it.
- Does it measure exposure, or jobs actually lost?
- Is it about Kenya, or is Kenya just mentioned in a story about a global figure?
- Does "at risk" mean the job goes, or that tasks inside it change?
- Who published it, and do they sell AI, training or consulting? That includes us.
- What period does the data cover, and which AI tools existed then?
A report may be useful even with weaknesses. You just need to know what it can and cannot tell you.
What can a Kenyan employer do now?
You can act on the parts that are clear, even while the research is unsettled.
Audit tasks before roles. Ask each team which tasks are repetitive, which need judgement and which need trust. Jobs are bundles of tasks, and most bundles will change in part.
Protect the entry-level ladder. If AI writes the first draft, juniors can check, correct and explain it, with a senior reviewing their reasoning. Keep hiring some graduates. A company with no juniors today will have no experienced staff in ten years.
Talk honestly. Say what you know and what you do not. Promising that nobody will be affected is as risky as saying nothing. Our guide on what a CEO should ask before an AI rollout covers this in more depth.
Retrain before you restructure. Look first at who could move into new work. If redundancies do become necessary, section 40 of the Employment Act sets out requirements including notice, selection criteria and at least 15 days' severance pay for each completed year of service. Take legal advice early.
Look out for clerical staff, and for women in those roles. The research consistently points to this group. Offer training to these staff first and not last.
If you buy outsourced data work, ask questions. Who employs these people, on what contract, and what mental health support do they get? Those are fair questions for any supplier.
Do not read the skills data too simply. Our guide to why AI skills matter explains what the term covers, and what Kenyan employers screen for looks at the hiring side. Human skills still count, which is the subject of our piece on soft skills and AI literacy.
What to tell staff who ask you directly
Staff will ask. Something like this is honest and can be adapted to your own facts:
"Nobody knows exactly how AI will change jobs in Kenya. The research suggests that tasks change more often than whole jobs disappear, and that office and admin work is affected first. Here is what we are doing: [your plan]. Here is what we have not decided: [your open questions]. We will tell you when that changes."
Two prompts for managers
Remove names and personal details before you paste anything.
1. Break a job into tasks
Below is a job description for a [role] at a Kenyan [type of organisation]. Context: we want to understand how AI might change this role. Constraints: list the main tasks in a table and, for each one, say whether AI could draft it, assist with it or is unlikely to help, and what a person must still do. Do not estimate how many jobs will be lost. Say "unclear" where you are unsure. Customisation: the reader is the line manager and will discuss this with the person in the role. Job description: [paste]
2. Check a report before you believe it
I will paste a news story about AI and jobs. Context: I lead a team in Kenya and want to know if I should act on it. Constraints: tell me what was measured, in which country, over what period, who published it and whether it counts jobs lost or only exposure. List what the article does not tell me. Quote only from the text I give you. Customisation: use short, plain sentences. Article: [paste]
What we still do not know
As of November 2026, we found no reliable answer to these questions.
- How many jobs in Kenya have been lost or created because of AI so far.
- How many Kenyans do data labelling for a living.
- Whether Kenyan graduates are being hired less because of AI.
- How quickly AI use at home turns into changed work. The Star, citing the Digital 2026 Mid-Year update, says 97.5% of Kenyan internet users aged 16 and over use AI platforms monthly, but that excludes people without internet and says little about depth of use.
- Whether AI will reduce demand for data labelling itself.
- What the courts will finally decide in the moderators' cases.
Questions people ask
Will AI take jobs in Kenya?
Some tasks will change, and some jobs may go, most likely first in clerical and routine office work. Global research suggests that change within jobs is more likely than mass job loss. We found no study that has counted AI job losses in Kenya.
Which jobs in Kenya are most at risk from AI?
The ILO names clerical and administrative roles among the most exposed groups. Routine outsourcing work, such as basic customer queries and transcription, also faces pressure. The list is an estimate based on tasks and global data.
How many jobs will AI create in Kenya?
We found no reliable forecast. The World Bank reports AI-skills vacancies in Kenya grew about fourfold from 2021 to 2024, from a very low base. The Ministry of ICT said in 2025 that outsourcing had created 19,650 jobs, which is a wider category than AI.
How much are data labellers paid in Kenya?
A 2026 WageIndicator survey of 57 Kenyan data workers found average monthly pay from their main platform or firm of about Ksh 21,850 for women and Ksh 26,107 for men. Pay was not adjusted for hours and the sample is small. Companies and workers disagree on other reported figures.
Is data labelling good work?
It depends on the employer and the task. It has offered income to graduates who struggled to find jobs. Workers have also reported low pay, weak contracts and mental strain, especially on harmful content, and some of those claims are before the courts.
Is the Business Laws (Amendment) Bill law?
There were two Bills with that name. National Assembly Bill No. 49 of 2024 was signed in December 2024. Senate Bill No. 51 of 2024, which would extend employment law to outsourcing and remote workers, is still a Bill on the latest record we found: the National Assembly passed it on 19 August 2026 and sent it back to the Senate, with no assent shown as at 15 October 2026.
Does Kenya have a law on AI and jobs?
Kenya has no AI-specific law in force as of November 2026. The Artificial Intelligence Bill, 2026 is still a Bill, and in October 2026 a Senate committee recommended not proceeding with it. The Employment Act already applies to any redundancy. See our explainer on the AI Bill.
Sources
- ILO, Generative AI and Jobs: A Refined Global Index of Occupational Exposure, Working Paper 140 (20 May 2025), last checked 10 November 2026
- UN News, summary of the ILO and NASK study (20 May 2025), last checked 10 November 2026
- ILO, "New ILO-World Bank paper highlights uneven global impact of generative AI on jobs" (27 March 2026), last checked 10 November 2026
- World Bank, press release on World Development Report 2026 (4 August 2026), last checked 10 November 2026
- IMF, "AI Will Transform the Global Economy. Let's Make Sure It Benefits Humanity" (14 January 2024), last checked 10 November 2026
- The Star, "Kenya among countries with high job vacancies requiring AI skills" (27 April 2026), last checked 10 November 2026
- The Star, report on the KNBS Economic Survey 2026 (30 April 2026), last checked 10 November 2026
- The Star, report on the World Economic Forum's entry-level work paper (2 July 2026), last checked 10 November 2026
- The Star, "Countries with the highest use of AI tools" (28 April 2026), last checked 10 November 2026
- Brynjolfsson, Chandar and Chen, Canaries in the Coal Mine (revised August 2026), last checked 10 November 2026
- Stanford Digital Economy Lab, AI Economic Indicators research note (June 2026), last checked 10 November 2026
- WageIndicator Foundation, Work and Wages of Data Labelers in Kenya (2026), last checked 10 November 2026
- WageIndicator, "Life After Meta" (5 June 2026), last checked 10 November 2026
- TIME, "OpenAI Used Kenyan Workers on Less Than $2 Per Hour" (18 January 2023), last checked 10 November 2026
- AP via WTOP, "Former Meta contractor Sama to lay off more than 1,000 workers in Kenya" (16 April 2026), last checked 10 November 2026
- WeeTracker, report on the Sama layoffs and smart glasses (30 April 2026), last checked 10 November 2026
- Business and Human Rights Resource Centre, Kenyan court postpones rulings in cases against Meta (12 February 2026), last checked 10 November 2026
- Capital FM via AllAfrica, "Court Backs Facebook Moderators in Case Against Meta, Sama" (27 April 2026), last checked 10 November 2026
- Capital FM via AllAfrica, "Remote Work 2.0: Can Kenya Become Africa's Digital Services Hub?" (8 September 2026), last checked 10 November 2026
- CIPIT, Strathmore University, AI and the African Gig Economy (2025), last checked 10 November 2026
- Mzalendo, The Business Laws (Amendment) Bill, 2024, National Assembly Bill No. 49, last checked 10 November 2026
- Mzalendo, The Business Laws (Amendment) Bill, 2024, Senate Bill No. 51, last checked 10 November 2026
- National Assembly, Bills Tracker, status as at 15 October 2026, last checked 10 November 2026
- The Star, "Court declines to halt controversial Business Laws Bill" (10 June 2026), last checked 10 November 2026
- The Standard, "How a new bill could deny gig workers employee protections" (8 July 2026), last checked 10 November 2026
- Spencer West, Legal Alert on the Business Laws (Amendment) Bill, 2024 (26 November 2024), last checked 10 November 2026
- Employment Act, No. 11 of 2007 (Revised Edition 2022), hosted copy, section 40, last checked 10 November 2026
- The Star, "Senate committee shoots down Nyamu's Bill on AI" (15 October 2026), last checked 10 November 2026
Where FuKazee fits in
FuKazee trains people in Kenyan organisations to use AI in the work in front of them. Part of our job is helping teams see which tasks will change and which skills still matter. We train in AI, and also in Microsoft 365, Google Workspace, leadership and communication. Since 2022 we have trained more than 3,000 people.
We have an interest in this subject, because companies that worry about AI may hire us, so weigh our views with that in mind. We assume nobody knows AI, we avoid jargon and we say plainly that nobody becomes an expert in a few days. We cannot tell you what will happen to any job. We can help your people get ready for the changes that are already visible.
See our page on AI training in Kenya, or read about AI training for HR teams.
To talk about preparing your team, 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.
