Every Telco Team Uses AI. Ask Them What They Pasted Into It.

There is a question I ask early whenever I am in a room with a telecoms team. Who here used an AI tool this week? Every hand goes up, usually with some pride. Then I ask the second question, which is the one that actually matters. Which tool, and what did you paste into it? That is where the room goes quiet. In a telco the answer to that second question carries more weight than in almost any other business. Somebody in customer care pasted a call transcript to get a summary. Somebody in revenue assurance dropped a slice of a churn report into a free chatbot to see whether it could spot the pattern faster. Somebody on the mobile money side was chasing a fraud signal and fed it transaction records. None of them were being careless on purpose. Every one of them was trying to do the job faster, which is exactly what you would want. The problem is what they were feeding, and where it went. ## Why telcos are the sharpest version of this problem I call this BROAI. Bring Your Own AI, a deliberate echo of the Bring Your Own Device wave that had corporate IT losing sleep a decade ago. It is a risk in every industry. In telecommunications it is in a category of its own. Think about what an operator actually holds. Call detail records. Location history, minute by minute, for tens of millions of people. Mobile money transactions. Identity documents from SIM registration. In Nigeria that sits under the Nigerian Communications Commission on one side and the Nigeria Data Protection Act on the other. An operator is holding some of the most sensitive data in the country, and the staff handling it every day have free chatbots open in the next browser tab. The gap is rarely malice. It is usually that nobody ever told them which tools were approved, so they picked their own. ## The three places AI is genuinely earning its keep The conversation in most African boardrooms has shrunk down to chatbots. That popup that says hello, how can I help you today. That is the front door. The value is behind it. **Churn.** Prepaid is the bulk of the African subscriber base, and prepaid customers do not resign, they simply stop recharging. By the time a monthly report shows the drop, the customer is already gone and probably holding a competitor's SIM. Recharge frequency, top-up size, data consumption and call patterns carry the signal weeks earlier. Acting on that signal is worth more than any retention campaign run after the fact. **Network operations.** African operators run thousands of base stations, many of them off-grid or on unreliable grid power, burning diesel and depending on generators that fail. Predicting which sites are about to go down, and why, turns a maintenance team from reactive to planned. Diesel consumption that does not match site output is its own detection problem, and a well understood one. **Fraud.** SIM box fraud, subscription fraud, and the fraud patterns inside mobile money are all pattern recognition problems at a scale no human review team can cover. This is not a new application of machine learning. What is new is how much cheaper it has become to run. Notice that none of these are chatbots. All three are systems that sit inside operations and change what the business can see. ## What good looks like One strategy. Approved tools. Clear ownership of who is accountable when something goes wrong. Company data staying inside systems the company controls, rather than spread across four or five free accounts that nobody is tracking. That is the thinking behind SWITCH, the six phase framework I use with clients. It starts with Scan and Score, because most organisations either overestimate their readiness or chase the wrong use case entirely, and you cannot fix either of those with a tool purchase. The operators that sort this out will pull ahead on cost and on customer experience at the same time. The ones that do not will find out the hard way, probably through a data incident rather than a competitor. So here is the question worth putting to your own team this week. If somebody pasted a call detail record into a free AI tool on Monday, would anybody in this building know by Friday? ## Frequently asked questions **What is the biggest AI risk for a telecoms operator in Africa?** Ungoverned use by well meaning staff. Subscriber data, location history and mobile money records leaving the organisation through free consumer tools that the company does not own or control. **Which AI use case should a telco start with?** Churn prediction, in most cases. The data already exists, the commercial value is easy to measure, and the result shows up within a quarter rather than a year. **Does AI in telecoms mean replacing the call centre?** No. It usually means reducing the volume of contacts that need a human at all, and giving the humans who remain better context when a call does come through. **How does regulation affect AI adoption for Nigerian operators?** The Nigerian Communications Commission governs the sector and the Nigeria Data Protection Act governs the personal data inside it. Both matter before a tool is chosen, not after. **How long before a telco sees real value from AI?** Weeks, if the first use case is chosen properly and the data is already there. A strategy that takes a year to show anything is not a strategy. *Stanislaus Martins advises and trains enterprise teams across Sub Saharan Africa on the practical application of marketing, technology and AI. If you are working on this inside an operator, the [speaking and training page](/speaking) covers formats, or start a conversation on the [work with me page](/work-with-me). Sector regulation for Nigerian operators sits with the [Nigerian Communications Commission](https://www.ncc.gov.ng).*