Kennedy Maweu

Nobody else knows how the reconciliation actually works. Somewhere in your company there is one person who holds the month-end close in her head. Call her Grace. If Grace resigns, you inherit a process nobody can run. Most COOs know this and do nothing, because "document the process" is a task nobody has time for. Here is the 45-minute version. Sit Grace down with Claude and this prompt: "You are a process analyst. Interview me step by step about how I perform the month-end close. Ask one question at a time. Probe for exceptions, workarounds, and things I do from memory that are written nowhere. When finished, produce a numbered SOP another accountant could follow without asking me anything, flagging every step that depends on my personal judgment." Forty-five minutes of her answering questions. Out comes the document you have been meaning to write for three years. The flagged judgment steps are your real key-person risk. Fix those first. Save this post.

KES 40 million bought an ERP nobody opens. That is what a manufacturing client paid three years before I met them. Licences, consultants, a go-live party with a cake. Last month I sat in their Nairobi office and watched the month-end close happen. Not one number came out of the ERP. The plant accountant exported to Excel, cleaned it, emailed it to head office, where someone re-typed it into another spreadsheet. The real system of record was WhatsApp. Nobody was being lazy. The ERP was built for the process the consultants designed. The team runs the process reality demands. Those are different companies. The fix was not more training. We built the workflow layer on top of the Excel-and-WhatsApp reality: extraction, checks, and a draft close pack, produced in hours instead of days. The system did not fail. The assumption that people would change for it did. What is your most expensive piece of unused software?

AI readiness will show up in your due diligence report before it shows up in your P&L. PE firms buying into East Africa are already asking a new question in diligence: how dependent is this business on manual processes and irreplaceable individuals? A company where month-end close, credit approvals, and reporting live in three people's heads is a key-person risk line item, and key-person risk gets priced into the multiple. The operators who document and automate core workflows now are not chasing efficiency. They are protecting exit value. Waiting until the sale process starts means fixing two years of process debt in ninety days, under scrutiny, at the worst possible negotiating moment. Agree or disagree?

HR Directors spend hours every quarter reading exit interview notes and producing nothing the COO can act on. Here is a Claude prompt that turns that pile into a board-ready signal in ten minutes: "I am pasting anonymised exit interview notes from the last quarter. Identify the top 5 recurring reasons for departure, ranked by frequency. For each: quote two supporting phrases, name the department pattern if one exists, and state one retention action with an estimated cost. Flag any mention of a specific manager or policy as a separate risk item. Format as a one-page summary I can table at the next executive meeting." Two rules before you run it: remove names and ID numbers first (DPA 2019 applies to employee data, and ODPC is now enforcing), and treat the output as a draft you verify, not a verdict. The gap between HR that reports headcount and HR that predicts attrition is one working system like this. Save this post.

Twelve days. That is how long month-end close was taking at a 400-person firm I sat with in June. Not because the finance team is weak. Because three people hold the whole process in their heads: one knows the reconciliation quirks, one knows which suppliers invoice late, one knows how the ERP actually behaves versus how the manual says it behaves. When one of them travels, the close stretches to fifteen days. The board pack goes out stale. The CFO presents numbers that are already three weeks old by the time anyone acts on them. The fix was not new software. It was extracting what those three people know into documented, AI-assisted workflows that anyone on the team can run. Close is now tracking under six days. Here is the question that started it: if your best finance person resigned today, how many days would your close take next month?

📋 Most HR managers in East Africa spend 3 hours a week writing the same documents from scratch. Job descriptions. Shortlisting notes. Offer letters. Warning letters. Each one started on a blank page, each one taking 45 minutes, that should take 5. Here is how to cut that down today: 1. Open Claude. Start a new conversation. 2. Paste this prompt: "You are an HR assistant for a Kenyan company governed by the Employment Act (Cap 226). I need a formal verbal warning letter for an employee who has been repeatedly late without notice. The tone should be professional and firm. Include: date, employee name placeholder, specific behaviour described, required improvement, consequence of no improvement, and a line confirming the employee received this letter. Keep it under one page." 3. Review the output against your company policy and the Act. 4. Adjust the behaviour description and save it as a template. 5. Use it every time. Reopen the conversation, change the name and behaviour, and export. That is a 45-minute task done in under 4 minutes. The prompt works. The judgment is still yours. Save this post.

𝗧𝗵𝗲 𝗯𝗶𝗴𝗴𝗲𝘀𝘁 𝗔𝗜 𝗿𝗶𝘀𝗸 𝗶𝗻 𝘆𝗼𝘂𝗿 𝗼𝗿𝗴𝗮𝗻𝗶𝘀𝗮𝘁𝗶𝗼𝗻 𝗶𝘀 𝗻𝗼𝘁 𝗮𝗱𝗼𝗽𝘁𝗶𝗼𝗻; 𝗶𝘁 𝗶𝘀 𝗶𝗻𝗮𝗰𝘁𝗶𝗼𝗻. Every governance review, every policy committee, every IT approval queue has a cost that nobody is calculating. It does not appear on a risk register. It does not show up in a board report. It accumulates quietly in manual hours, slow decisions, and work that your competitors complete in minutes while your team is still on a shared spreadsheet. The organizations I see winning with AI in East Africa right now are not the most sophisticated. They are the most decisive. They picked a process, started, and measured. That is the entire strategy. The risk conversation in African boardrooms is always about what could go wrong if you adopt AI. Nobody is presenting what is already going wrong because you have not. That calculation is overdue. Agree or disagree?

📋 𝗠𝗼𝘀𝘁 𝗹𝗲𝗮𝗱𝗲𝗿𝘀 𝗮𝘀𝘀𝘂𝗺𝗲 𝘁𝗵𝗲𝗶𝗿 𝘁𝗲𝗮𝗺𝘀 𝗿𝗲𝘀𝗶𝘀𝘁 𝗔𝗜 𝗯𝗲𝗰𝗮𝘂𝘀𝗲 𝗼𝗳 𝗯𝘂𝗱𝗴𝗲𝘁, 𝘀𝗸𝗶𝗹𝗹𝘀, 𝗼𝗿 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲. The real reasons are quieter and harder to address from a slide deck. In every organisation I have worked with across East Africa, 𝗿𝗲𝘀𝗶𝘀𝘁𝗮𝗻𝗰𝗲 𝘁𝗿𝗮𝗰𝗲𝘀 𝗯𝗮𝗰𝗸 𝘁𝗼 𝘁𝗵𝗿𝗲𝗲 𝘂𝗻𝘀𝗽𝗼𝗸𝗲𝗻 𝗳𝗲𝗮𝗿𝘀: 1. "This will make me redundant." Nobody says it in a meeting. Everyone is thinking it. 2. "I will look incompetent if I use it wrong." Senior professionals especially. The fear of failure in public is stronger than the appeal of saving time. 3. "Leadership is not using it either." If the COO still sends manually typed status updates, the signal to the team is that AI is optional. None of these is addressed by a training programme or a policy document. 𝗧𝗵𝗲𝘆 𝗮𝗿𝗲 𝗮𝗱𝗱𝗿𝗲𝘀𝘀𝗲𝗱 𝗯𝘆 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻, 𝗺𝗼𝗱𝗲𝗹𝗹𝗶𝗻𝗴, 𝗮𝗻𝗱 𝘀𝗽𝗲𝗰𝗶𝗳𝗶𝗰𝗶𝘁𝘆. Here is the prompt that helps you start: "I am a [role] at a [type of organization] in [country]. My team is resistant to using AI tools. Write an internal message I can send to my direct reports that: addresses job security concerns honestly, shows one specific task AI will handle, and explains what that frees them to focus on instead. Tone: direct and reassuring. Length: under 200 words." Use it before your next team meeting. 𝗦𝗮𝘃𝗲 𝘁𝗵𝗶𝘀 𝗽𝗼𝘀𝘁.

🏢 A COO at a Nairobi firm told me her team had been waiting four months for IT to approve a Claude account. Four months. To access a browser-based tool that requires no installation, no integration, and no infrastructure change. The same tool her team was already using on personal devices at home. IT had raised a data security query. The query went into a queue. The queue had no deadline. The operations team kept doing manually what Claude could have handled in minutes. I asked her what would happen if she just bought the team subscriptions on a company card and started. She said she had not thought of that. That conversation happens more than it should. AI tools like Claude do not sit inside your IT infrastructure. They sit in a browser. The risk assessment that applies to an ERP implementation does not apply to a writing and reasoning tool that your team accesses with a work email. IT has a role in enterprise AI deployment. That role is not approving whether your HR manager can draft a job description faster. Governance and gatekeeping are not the same thing. One protects the organization. The other stalls it. Who in your organization is currently waiting for permission that may never arrive?

𝗔𝗳𝗿𝗶𝗰𝗮 𝗱𝗼𝗲𝘀𝗻'𝘁 𝗵𝗮𝘃𝗲 𝗮𝗻 𝗔𝗜 𝘀𝗸𝗶𝗹𝗹𝘀 𝗴𝗮𝗽. 𝗜𝘁 𝗵𝗮𝘀 𝗮 𝗽𝗲𝗿𝗺𝗶𝘀𝘀𝗶𝗼𝗻 𝗴𝗮𝗽. I have trained professionals at 𝗞𝗲𝗻𝗚𝗲𝗻, 𝗞𝗲𝗻𝘆𝗮 𝗔𝗶𝗿𝗽𝗼𝗿𝘁𝘀 𝗔𝘂𝘁𝗵𝗼𝗿𝗶𝘁𝘆, and the 𝗡𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗔𝗜 𝗖𝗼𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝗶𝗻 𝗔𝗿𝘂𝘀𝗵𝗮. Capable, credentialed, operationally sharp people. Every room had someone who wanted to use AI in their work that afternoon. Not one was 𝘄𝗮𝗶𝘁𝗶𝗻𝗴 𝘁𝗼 𝗹𝗲𝗮𝗿𝗻 𝗵𝗼𝘄. They were waiting for 𝘁𝗵𝗲𝗶𝗿 𝗼𝗿𝗴𝗮𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝘁𝗼 𝘀𝗮𝘆 𝘆𝗲𝘀. 𝗡𝗼 𝗮𝗰𝗰𝗼𝘂𝗻𝘁 𝗽𝗿𝗼𝘃𝗶𝘀𝗶𝗼𝗻𝗲𝗱. 𝗡𝗼 𝗽𝗼𝗹𝗶𝗰𝘆 𝘁𝗼 𝗽𝗼𝗶𝗻𝘁 𝘁𝗼. 𝗡𝗼 𝗺𝗮𝗻𝗮𝗴𝗲𝗿 𝘄𝗵𝗼 𝗵𝗮𝗱 𝗳𝗼𝗿𝗺𝗮𝗹𝗹𝘆 𝗮𝗽𝗽𝗿𝗼𝘃𝗲𝗱 𝘁𝗵𝗲 𝘁𝗼𝗼𝗹. So they sat on the capability and kept doing the work the slow way. The skills gap narrative is convenient for training vendors and conference organizers. 𝗜𝘁 𝗽𝘂𝘁𝘀 𝘁𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺 𝗶𝗻𝘀𝗶𝗱𝗲 𝘁𝗵𝗲 𝗶𝗻𝗱𝗶𝘃𝗶𝗱𝘂𝗮𝗹, 𝘄𝗵𝗶𝗰𝗵 𝗺𝗲𝗮𝗻𝘀 𝘁𝗵𝗲 𝗼𝗿𝗴𝗮𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗱𝗼𝗲𝘀 𝗻𝗼𝘁 𝗵𝗮𝘃𝗲 𝘁𝗼 𝗰𝗵𝗮𝗻𝗴𝗲 𝗮𝗻𝘆𝘁𝗵𝗶𝗻𝗴. 𝗧𝗵𝗲 𝗿𝗲𝗮𝗹 𝗽𝗿𝗼𝗯𝗹𝗲𝗺 𝗶𝘀 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗮𝗹. 𝗦𝗼𝗺𝗲𝗼𝗻𝗲 𝗮𝘁 𝘁𝗵𝗲 𝘁𝗼𝗽 𝗵𝗮𝘀 𝗻𝗼𝘁 𝗱𝗲𝗰𝗶𝗱𝗲𝗱 𝘁𝗵𝗮𝘁 𝗔𝗜 𝘂𝘀𝗲 𝗶𝘀 𝗲𝘅𝗽𝗲𝗰𝘁𝗲𝗱, 𝗻𝗼𝘁 𝗼𝗽𝘁𝗶𝗼𝗻𝗮𝗹. Until that decision is made and communicated, every capable person below it will wait. You do not close a permission gap with a training program. 𝗬𝗼𝘂 𝗰𝗹𝗼𝘀𝗲 𝗶𝘁 𝘄𝗶𝘁𝗵 𝗮 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻. Has your organization made that decision yet? Agree or disagree?