Voice AI companies often need to run workflows in realtime to get a result back to a customer that’s waiting on the phone.

We’ve gotten 20-step scripts down to 35 seconds p99 latency end-to-end, with a p50 of 28 seconds. We used a combination of caching + parallelization + site-specific network optimizations to achieve this.

Talk to us on Slack for running workflows in realtime — most optimizations are site-specific and require an understanding of Simplex’s internal infrastructure.

There’s unfortunately no one-size fits all guide, but we can almost definitely speed up your workflows. So talk to us!