In a product update published on 27 October 2026, Apifonica says it has spent two quarters rebuilding its voice platform around AI agents instead of call scenarios. The core call-handling logic is unchanged. One client moved to the new version a few weeks before the post and is running live traffic on it; the rest are to follow one at a time. The company does not name that client.
The main addition is memory. Where a client switches it on, the platform keeps each contact’s call history, call outcomes, preferred language and open support requests, so an agent can recognise someone ringing back about the same fault. Apifonica says it used to build a custom CRM integration for this in three cases out of five. Contacts can also be grouped into segments and called automatically.
On-premise delivery was first reported in a post on 21 September. Development finished in mid-August, Apifonica said then, and a standard package with an 11-section readiness checklist brings a typical installation to between three and eight weeks. That post described the first external customer as a national mobile and broadband operator in Uzbekistan and set a target of automating about 30 per cent of its traffic. The October update says servers are already in place at a national telecom operator and delivery is due by the end of the year. Neither post names the operator, and neither says whether they refer to the same one.