For small African firms, AI's first draft was only the start of the work
Interviews with 27 workers in Kenya and Nigeria found that client-ready output depended on verification, local knowledge and repair. The fieldwork was conducted in 2023, so this is evidence about early workplace practice—not current model performance.
The short version
- What happened
- Microsoft Research Africa researchers conducted 21 English-language interviews with 27 regular AI users from 21 small businesses: 18 participants in Kenya and nine in Nigeria.
- Who it affects
- A small-business worker may gain a quick draft yet remain personally responsible for whether it is accurate, locally credible and safe to send. That checking and repair work should appear in procurement and workload decisions.
- What remains unproven
- The study is a preprint and has not been peer reviewed.
Keep in mind
The study is a preprint and has not been peer reviewed.
The 27 participants were recruited through professional networks and were already frequent AI users; they are not representative of either country or African small businesses generally.
English-language interviews may under-observe practices conducted in local languages or code-mixed conversation.
Fieldwork occurred in 2023, and the study measured reported practice rather than current model performance, causal productivity or time saved.
Source analysis · 1 linked source. Primary analysed source date: 22 October 2026.
Keep following the evidence
Join The Impact Brief founding list for selected AI reports, their effect on people and the evidence limits. Read a sample first.
Read the full report · 3 min The evidence, methods, implications and limitations
The difficult step came after generation
Participants often used AI to get past a blank page, polish a sensitive email or assemble a first version of client material. Before circulating it, they rewrote for voice, checked factual claims, reconstructed missing sources and decided what should not be delegated. The researchers call this narrower boundary task 'release work': making one artefact defensible to an external audience.[1]
Local context created repair work
Accounts described failures involving local institutions and markets, code-mixed language and accents, and credible representation of African people and places. Workers responded by adding local sources, tightening prompts, comparing several tools or abandoning an output when repair cost exceeded its value. Agreement between tools was treated as a clue, not independent verification.[1]
The distinction that matters
From draft to accountable output
The interviews show why generation time is only one part of the workflow.
Draft
Generate a starting point
Faster options or wording
Repair
Localise voice, place, language and representation
Context fit requires human work
Release
Verify claims, rebuild traceability and set boundaries
A defensible client-facing artefact
Useful evidence, but not a productivity estimate
The study deliberately recruited people already using generative AI at least three times a week through professional networks. It captures how early adopters described their work, not how common these practices are or whether AI saved more time than checking consumed. For employers and vendors, the testable implication is to measure the full path from prompt to approved deliverable—including local adaptation, source checks and sign-off.[1]
Watch the background, then check the evidence
Illustrative photo · not a video frame
UNESCO · 11 min 27 sec · Background explainer
UNESCO Talks: Interview with 2024 Nobel laureate Professor Daron Acemoglu
YouTube loads only when you choose; playback starts when you press play.
Economist Daron Acemoglu discusses how choices about AI affect workers, inequality and society. Pair the interview with our comparison of task exposure, hiring and skills evidence.
Evidence check: An expert interview from 2024, offering a perspective on policy and work rather than a measured forecast of job losses.
Published 2024-10-14. Source checked 14 October 2026. Check captions on the publisher’s video page.
Read the AI and jobs guidePhoto: Dylan Gillis / Unsplash · Licence
Continue the story
Read next
Continue exploring verification: compare the finding and its evidence limits.
When AI work looks faster, who counts the mistakes? A preprint tests the missing ledger
New analysis today of an unreviewed 23 October preprint: reanalysis shows that one productivity dataset supports different slowdown estimates, while 40 paired AI-review trials found that an inflated reliability claim cut checking and let wrong figures through. This does not measure human worker productivity.
Read the next reportExplore the full source and evidence record
What this means for people
- A small-business worker may gain a quick draft yet remain personally responsible for whether it is accurate, locally credible and safe to send. That checking and repair work should appear in procurement and workload decisions.
Global context
The sample is deliberately situated in Kenya and Nigeria rather than presented as 'the Global South'. Its broader value is the practical question it exposes: who bears the labour of making AI output locally credible and accountable before it reaches a customer?
What the evidence does not yet show
- The study is a preprint and has not been peer reviewed.
- The 27 participants were recruited through professional networks and were already frequent AI users; they are not representative of either country or African small businesses generally.
- English-language interviews may under-observe practices conducted in local languages or code-mixed conversation.
- Fieldwork occurred in 2023, and the study measured reported practice rather than current model performance, causal productivity or time saved.
What to watch next
- Replications with local-language interviews
- Direct workflow observation and measured checking time
- Comparisons across sectors and countries
- Whether source-grounding and locale controls reduce repair work without weakening accountability
Living evidence record
Impact record IAI-18GJ6UL
Evidence stage
Studied
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Not yet
Record status
Monitoring
Last checked
23 October 2026
Source trail
1 direct source across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Single-source reporting disclosure
This record analyses one direct source. It can establish what arXiv published or reported, but it is not independent corroboration of every performance claim or predicted outcome. The confidence label will change only when broader evidence is added.
Evidence trail
Sources used for this report
Links checked 23 October 2026
Research question, topics and regions
How workers in Kenyan and Nigerian small businesses make generative-AI output accountable and locally credible before client-facing use
Topics: Small business · Workplace AI · Localisation · Verification · Accountability
Regions: Kenya · Nigeria · Africa
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Reader commentary
What does this mean where you are?
Tell us what you have seen at work, in your community or in everyday life. A question is welcome too. No account is needed; we review notes before publishing them.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.