Your support team keeps saying, “We don't offer that.” Read those tickets.
Matías Bonvin· Updated
“We don't offer that service.”
A perfectly correct answer can close a support ticket and leave a customer looking for someone else to do the work.
Suppose you distribute equipment to business customers. Your support team answers questions about specifications and delivery. Customers also ask whether someone can check which items they need across several sites, prepare a replacement schedule and arrange the orders.
The team points them to the catalogue. The request gets tagged “advice” and closed.
A better chatbot could send the catalogue faster. Before building it, read what the customer was trying to buy.
Separate a missing answer from missing work
Unresolved tickets contain different problems. Some customers cannot find information you already provide. Others received a faulty product or an incorrect invoice. Those cases need a working service, not a new paid offer.
A different group may be asking for work outside your current scope. In the distributor example, selecting equipment across several sites could require an inventory, technical checks and a purchasing plan. Sending specifications answers only part of that request.
Review a manageable sample of recent tickets. Include resolved cases too: “closed” can describe your team's status without describing the customer's outcome. Preserve the original wording and the final response.
For each case, write down what the customer wanted to accomplish, what prevented it and whether the existing offer was supposed to cover it. If the answer is already included in the contract, fix the delivery. Charging again would hide a service failure inside a new product name.
Count customers, not messages
A long thread can make one unusual request look like a market. Group messages by customer and underlying need before counting them. Keep reopenings attached to the original case.
Then look for customers in similar situations. Are they trying to solve the same problem at the same moment? A distributor preparing a new site and one replacing worn equipment may use the same words but need different services.
AI can help group the language and surface candidate patterns. Give the reviewer a link to each original case, not just a polished summary. Have someone who understands the service check a sample from each group and inspect the cases the model couldn't place. Otherwise, the neat categories may erase the differences that determine whether the service is deliverable.
Use only the records your team is authorised to process, and remove unnecessary personal information before analysis. A commercial hypothesis rarely needs the entire customer file.
Find out what they do after your reply
Return to a few customers whose requests fit the same pattern. Ask what happened after the support exchange. Did someone inside their business do the work? Did they hire another provider? Did the job remain unfinished?
Those answers reveal effort and alternatives. “That would be useful” reveals much less.
For the equipment example, the next step could be a bounded offer to review one site's requirements. Describe what the buyer would receive, what information they must supply and what the work excludes. Check internally that you can deliver it before discussing a commitment.
Then ask for a purchase decision on that specific scope. A paid trial or an accepted proposal is stronger evidence than a request for another presentation. It still doesn't prove a repeatable market.
Check whether the service survives delivery
Before automating it, deliver a small, agreed version and record the actual work. Include time spent collecting missing information, answering follow-up questions and correcting recommendations. Compare that effort with the revenue and responsibility the business takes on.
Set a reason to stop before enthusiasm takes over. You might reject the idea if customers only want it when it is free, if each case requires unrelated expertise, or if it duplicates work already promised in the base service.
A viable service might later use AI to prepare the inventory or assemble supporting documents. The person accountable for the recommendation still needs a way to check it. Automating the first draft doesn't remove the obligations attached to what you sell.
Solve IT can help connect support records with the information needed to investigate a recurring request and build the agreed workflow once the opportunity is defined. The useful starting point is the conversation your current process keeps closing.
Bring the tickets where your team had to say, “We don't offer that.”