contact center AI

Dialpad and Ada expect to close on or about November 12, and both say contracts, pricing and integrations stay as they are. Those promises carry no end date, so Ada customers on third-party contact centers should get them in writing.

Five9 is buying Level AI for $325M in cash and will keep selling its QA, agent assist and VoC apps to contact centers on other platforms. Level AI customers should get pricing, data-use and integration terms in writing before the Q4 close.

Salesforce says Cresta will keep supporting the contact center platforms its customers use, but names none and sets no end date. Customers have until a close expected between February and April 2027 to get terms in writing.

Deepgram is acquiring Daily, the maker of Pipecat. The BSD-2 framework stays open and gets a promised foundation, but Daily's WebRTC, telephony and Pipecat Cloud contracts change owner.

Azure Neural TTS lists $15 per million characters against $38 to $80 for Cartesia and ElevenLabs, and keeps no real-time text by default. The vendors' latency claims measure different things, so test p95 from your own region before you sign.

Deutsche Telekom's €2.5B AI savings target for 2030 is gross and excludes token costs, which it caps at a low double-digit share of savings. Its per-function structure is worth copying; its mixed baselines are not.

Microsoft's MAI-Transcribe-2-Streaming claims a 2.5% streaming WER at $0.54 an audio hour, but it ships as a public preview with no SLA and no word-level timestamps, at 2.7 times Grok's streaming price.

A $0.05 voice AI minute bills $0.09 to $0.14 once transcription, the model, the voice and the phone line are added. On a 40,000-minute inbound workload, ElevenLabs Agents was the cheapest managed platform and Bland AI the most expensive.

Inworld bought Ultravox and gave its own voices a free TTS-2 upgrade. Agents on ElevenLabs, Cartesia and other third-party voices got one line: nothing changes today.

Inception's Mercury Voice cuts the voice-agent LLM line to about $0.009 a minute, but its 320ms latency is model-only and its benchmark win arrives without numbers. Bake it off on your own calls before you swap.