Two firms had told us the mainframe replacement was a three-year, one-cutover programme. Anix showed up in week three and said the sequencing was wrong, not the scope. They were right, and they proved it with our own claims data rather than a benchmark.

What I did not expect was that our own engineers would end up running it. There was no handover event — they were in the standups from the first sprint. Nine months after go-live we have not raised a single support ticket with Anix.

They told us our own business case was overstated by about 30% before we had signed anything. That cost them margin and bought them a five-year relationship.

The compliance committee approved an AI deployment on first submission, which has never happened here. The evaluation evidence and the model documentation were simply better than anything we had seen from a vendor.

Partner-level attention was not a sales promise. Amara was in our weekly steering for fourteen months and read every paper before the meeting rather than during it.