Five uses that stuck
Quotes, meeting minutes, responses to tenders, sorting incoming mail, and writing product sheets.
These are not the most spectacular uses. They are the ones still in place after several months, used every week by teams with no technical background. What they have in common: they save time on a task everyone found tedious, and a human keeps control over the final result.
Context matters. In a Conakry SME, a handful of people often wear several hats, with an internet connection that is sometimes unstable, documents in formal administrative French, and customers who write by email as readily as by WhatsApp. The tools that survive are the ones that adapt to these conditions, not the other way around.
1. Quotes
Preparing a quote often took half a day: finding the prices, rephrasing the customer's request, formatting the document. With an assistant fed with the catalog and a few template quotes, a first version is ready in minutes. The salesperson checks quantities and prices, adjusts, and sends it.
The gain is not just time: quotes go out faster, and in a market where the first to respond often wins the deal, that matters.
2. Meeting minutes
Recording a meeting, transcribing it, pulling out the decisions and the list of actions: this is the use adopted most quickly. Nobody liked writing minutes, and they often arrived too late to be useful. They now go out the same day.
One condition for this to work: proofread before sending. The transcription sometimes mixes up proper names and local terms, and a wrongly recorded decision does more damage than no minutes at all.
3. Responses to tenders
For SMEs that bid on public contracts or on calls from international organizations, the bid file is heavy, repetitive work: company presentation, references, methodology, compliance with requirements. AI helps at two levels: analyzing the specifications to extract the requirements and the documents requested, then drafting a first version of the recurring sections from previous bids.
The part that wins the contract, the technical offer tailored to the need and the price, remains human work. But the team can finally spend most of its time on it, instead of losing it reformatting references.
4. Sorting incoming mail
Emails, website forms, messages: a shared inbox quickly becomes unmanageable. Automatic sorting classifies each message (quote request, complaint, supplier invoice, job application, sales pitch), extracts the essentials and routes it to the right person. Nothing is answered automatically; everything is simply filed and summarized.
It is a discreet use, but it is the one teams say they would no longer want to do without.
5. Writing product sheets
For retailers and distributors who sell online or publish a catalog, writing hundreds of product sheets was a job that kept being postponed. Starting from the technical specifications and a few examples, AI produces consistent descriptions that the team proofreads and corrects in batches. The catalog, half empty for a long time, is finally complete.
Two uses that were dropped
Not everything stuck. Two uses, tested with enthusiasm, were stopped after a few weeks.
The fully autonomous customer chatbot. The idea was appealing: answering customers on WhatsApp at any hour, with no human involvement. In practice, the questions were too varied, the answers sometimes wrong on prices or lead times, and customers wanted to know that a real person was reading them. A single commitment wrongly made by the bot costs more than all the hours saved. The use was transformed: AI now suggests a reply, and a team member approves it before it is sent.
Automatic sales forecasts. On paper, forecasting next month's sales helps you order better. But with a short, incomplete history disrupted by unpredictable events (power cuts, supply shortages, exchange-rate swings), the forecasts were no better than the managers' intuition, and they were less well understood. The project was put on hold while data entry is made more reliable.
What these experiences teach us
- Start with a tedious task, not with a technology. The uses that stick address an identified chore, not a desire to "use AI".
- Keep a human in the loop. Every adopted use relies on a review. Those that aimed for full autonomy failed.
- Start from your own documents. Catalogs, old quotes, past tender files: the value comes from what the company already has.
- Measure simply. Time spent before and after, number of quotes sent, response time. You do not need a sophisticated dashboard to know whether a use is worthwhile.
- Train the teams. A tool people do not know how to use properly is quickly abandoned. Half a day of hands-on training often makes the difference.
- Be careful with the data you hand over. Before sending contracts, customer data or prices to an online service, check where they are stored and what the provider is allowed to do with them.
AI does not transform an SME overnight. It gives it back a few hours a week, on the tasks nobody liked doing. It is modest, and that is precisely why it lasts.
If you run an SME and are wondering where to start, ask your teams: which repetitive task would you like never to do again? The answer is often your first use case.
