A construction firm with 297 staff spent £83k on satellite internet and 5G boosters because ‘connectivity’ was its top helpdesk tag across 200+ tickets. The tickets kept coming. The real problem was a document app that timed out when site managers uploaded hundreds of photos at once.
The fix is to stop counting labels. Two AI agents read what people were trying to do when it failed, group tickets by root cause, and a validation step confirms the fault before anyone spends money.
Figures from the case as reported in aiblBRIEF
Because the team counted the label, not the failure. Site managers wrote ‘connection keeps dropping’, so their tickets were tagged connectivity. A business analyst on site found in ten minutes that the app timed out on large uploads.
With two agents. An Interpreter reads each ticket and records the user’s own diagnosis, the action they were taking and where the workflow broke, with quotes as evidence. A Cluster agent then groups tickets by root cause and assigns an owner, impact and effort.
Reproduce the failure and ask the people affected. The team uploads hundreds of photos on 4G to confirm the timeout, then asks five users whether background upload would solve their problem. No spend is approved until both check out.
If your helpdesk data drives spending decisions, the full case study gives you the method and the prompts:
This case study first appeared in aiblBRIEF. Read the original article: The Support Mirage.
