Synthflow AI

Synthflow is an accredited AI provider in Five9's AI Agent Connect program. Contact centers on Five9 VCC can now bring a Synthflow voice agent into the call flows they already run. They buy Synthflow directly, and Five9 activates AI Agent Connect on their account. A Five9 workflow sends the call to the agent. When the conversation is done, the agent hands the call back with what it learned: what the caller needed and where the call should go next, as call variables Five9 routes and reports on. The routing, the queues and the reporting stay in Five9.

Your Synthflow agents can now run on OpenAI's GPT-6 Luna. OpenAI made Luna to bring GPT-6's advances to a faster model for work at scale. The model is one setting on each agent, so you can move one agent first and the rest when your team is ready.

Synthflow's official plugin for OpenAI's ChatGPT is live. Ask ChatGPT to create a voice agent, refine its prompt, place a test call or review recent calls - and it happens in your Synthflow workspace. You sign in through Synthflow, so there's no API key to copy. The same plugin works in Codex. Your voice agents, one prompt away.

Party Time at CCW Las Vegas! 🪩 We're bringing together CX, AI, and business leaders for a special invite-only gathering alongside industry peers and the team from Synthflow AI, 8x8, Krisp, and more. Expect insightful conversations, networking opportunities, drinks, and an unforgettable atmosphere as we connect with the people shaping the future of customer experience and AI. 🎟️ Want to join us? Spaces are limited, so make sure to reserve your spot today. 👉 Register here: https://luma.com/1tgzjysh Secure your spot, bring your questions, and we -- Matt A., Tim Claudin, Alex Barnes will see you there! #Synthflow #AIAgents #CustomerExperience #VoiceAI #Automation #CXInnovation

Most Voice AI teams are still measuring the wrong thing. We analyzed our Q1 2026 production data across thousands of voice agents, and the pattern was clear: “containment” is not enough. A contained call is not always a completed job. A human handoff is not automatically a failure. An action being triggered does not mean the workflow was resolved. And building agents is not the same as operating them in production. The problem with containment is that it compresses too many different operating states into one number. To understand whether Voice AI is actually working, you need to inspect the full production state: 1. Did the call reach a meaningful end state? 2. What job was the agent supposed to complete? 3. Was the job actually advanced? 4. Did the downstream action complete? 5. Is there an audit trail proving what happened? The teams that win with Voice AI will not be the teams with the cleanest demo metrics. They will be the teams that can prove, in production, that calls moved real work forward. We wrote up the full breakdown here: https://lnkd.in/ea5ZfHDh

Evaluate Voice AI using your real call volume and cost structure. Complete a 5-step ROI assessment in 2 minutes.