On this page
Product managers use AI as a fast writing and sorting partner. It groups user research notes into themes, drafts PRDs and release notes, prepares prioritisation discussions and turns meeting notes into action lists. The product decisions stay with the PM: what to build, for whom and why, and every claim about users must trace back to real evidence.
This guide is for product managers, product owners and founders doing product work at Kenyan fintechs, startups, banks' digital teams and software companies. If you deliver projects to a deadline and budget, our page on AI training for project managers fits you better.
Each sample prompt below follows the four Cs we teach: Clarity (the task), Context (your situation), Constraints (the rules) and Customisation (the format and reader). Remove names, phone numbers and account details before you paste anything.
Six tasks where AI helps product managers today
1. Sorting user research into themes
After ten interviews, you have pages of notes. AI can sort them into themes in minutes.
Group these user interview notes into themes. Context: we interviewed 10 small shop owners in Nairobi about how they record M-PESA sales; names are removed. Constraints: use only the notes; under each theme, give how many interviews mention it and two short quotes; list anything that contradicts our idea that shop owners want a daily report. Customisation: a table, then five bullet points for the team.
Check: open the notes and find each quote yourself. Count the mentions again. AI tools can merge two different complaints into one neat theme, or drop the awkward one. For more on this, read Can AI Help With Qualitative Data Analysis?
2. Drafting a PRD
A product requirements document is mostly structured writing, which suits AI well.
Draft a one-page PRD for the feature below. Context: [problem, target users, evidence, launch goal]. Constraints: sections for Problem, Users, Goals, Out of scope, Requirements, Acceptance criteria, Open questions; do not invent metrics, dates or user numbers; where something is missing, add it to Open questions. Customisation: plain English that an engineer and a support agent can both follow.
Check: the Open questions list is often the most useful part. Make sure every requirement can be tested, and that nothing appears in scope that nobody agreed.
3. Finding edge cases and writing user stories
List the edge cases for this payment confirmation flow. Context: [describe the flow]. Constraints: consider only cases that could happen with our users, such as a phone going offline mid-payment, a delayed confirmation message or a shared phone; group them by likelihood; turn the top five into user stories with acceptance criteria.
Check: go through the list with an engineer and a support agent. They know which cases already happen every week.
4. Writing release notes
Write release notes for version 2.4 from the merged changes below. Constraints: describe only what shipped; no promises about future features; one version for customers in simple English and simple Kiswahili, under 80 words each, and one for our support team with known issues. Changes: [paste the change list].
Check: compare every line with what actually went live. A release note that promises something missing creates support tickets.
5. Preparing prioritisation discussions
AI cannot know your reach, revenue or effort numbers. It can organise the numbers you give it and argue against your favourite.
Score these eight backlog items using Reach, Impact, Confidence and Effort, with the estimates I give for each. Constraints: show your working in a table; do not change my estimates; then give the strongest argument against the top-ranked item and the strongest argument for the lowest-ranked one.
Check: treat the ranking as a starting point for the team conversation. The decision, and the explanation to stakeholders, belong to you.
6. Turning meetings into decisions and actions
Meeting tools now do much of this inside the call, depending on your licence. As of November 2026, Microsoft says Copilot in Teams can summarise key points and suggest action items, and that it will not work in meetings hosted outside the participant's organisation. Google says its "Take notes for me" feature in Meet needs an eligible Google Workspace edition or Google AI plan, and supports one language at a time. That matters for meetings that move between English and Kiswahili.
If you write notes yourself, paste them in with this:
Turn these notes into a list of decisions, actions with an owner and date, and open questions. Constraints: do not invent owners or dates; mark anything unclear with [CHECK].
Check: tell people when a meeting is being transcribed or summarised, and read the summary before you send it. Summaries can miss a disagreement or record a "maybe" as a decision.
What AI should not do for a product manager
- Invent evidence. A quote, a percentage or a "users want" line with no source behind it does not belong in a PRD.
- Make the call. Prioritisation involves strategy, politics and trade-offs a tool cannot see.
- See raw user data in a public tool. Phone numbers, transaction records, ID numbers and support tickets with names stay out.
- See confidential plans in tools your company has not approved. Unreleased roadmaps, pricing and partner terms included.
- Sign off regulated features. Payments, lending, insurance and health features need your compliance and legal teams.
The Kenyan context
User research produces personal data. Kenya's Data Protection Act, 2019 is regulated by the Office of the Data Protection Commissioner (ODPC), whose mandate includes regulating how personal data is processed and protecting individuals' privacy. In practice, ask for consent before you record an interview, store recordings in approved places and mask transcripts before any AI tool sees them. Our guide to AI and Kenya's Data Protection Act explains the duties in more depth.
Your tools also need your context. A generic answer will not think about shared phones, patchy data or mobile money confirmations unless you tell it those things matter to your users.
How FuKazee trains product and startup teams
Our AI training for startups page lists customer discovery among the first workflows we teach: preparing interview guides, clustering notes into themes and looking for evidence that contradicts your favourite idea. FuKazee has trained the HackHouse Africa community and works from HackHouse in Westlands. We cover ChatGPT, Claude, Gemini and Microsoft Copilot, starting with free versions, and we promote none of them. We assume nobody in the room knows AI yet. Since 2022 we have trained more than 3,000 people, and we are rated 4.9 from 38 Google reviews, as shown on our reviews page on 14 November 2026. Individual PMs can book one-to-one training.
Questions people ask
Will AI replace product managers?
AI speeds up the writing around product work. It cannot talk to your users for you, weigh trade-offs between teams or own a decision. In our view, those parts of the job matter more as drafts get easier.
Which AI tool is best for product managers?
Start with the tool your company has approved. The free versions of ChatGPT, Gemini, Claude and Copilot all handle summaries and drafts. Clear prompts matter more than the brand.
Can I paste user interview transcripts into ChatGPT?
Only after removing names, phone numbers and anything that identifies a person, and only if your consent forms and company rules allow it. When in doubt, ask your data protection lead first.
Can AI write my whole PRD?
It can write a solid first draft from your notes. You still decide the scope, check every requirement with engineering and remove anything the evidence does not support.
Do I need to code to use AI as a PM?
No. Everything in this guide uses plain-language prompts. Knowing enough to read a short technical summary helps when you review AI-assisted specs with engineers.
Sources
All sources last checked 14 November 2026.
- Microsoft Support, Catch up on meetings with Microsoft Copilot in Teams: summaries, action items, meetings hosted outside the organisation.
- Google Meet Help, "Take notes for me" in Google Meet: eligible plans, one language at a time.
- Office of the Data Protection Commissioner: mandate under the Data Protection Act, 2019.
- FuKazee, AI training for startups in Kenya: customer discovery workflow, HackHouse, tools covered.
- FuKazee, Reviews and letters: 4.9 rating from 38 Google reviews.
Where FuKazee fits in
FuKazee trains product, startup and corporate teams to use AI on the work in front of them, with the free version of each tool first and clear rules on user data.
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.





