Changelog
Every change, in plain language.
No version numbers without explanation. Every release says what moved, who it affects and whether you need to do anything about it.
Voice Layer goes multi-region
Feature
Voice agents now run in EU and APAC regions. Pick a residency per workspace and round trips stay under 200ms without routing audio across borders.
Existing workspaces stay where they are until you move them, and moving is a single setting with no re-training. Latency inside a region held under 200ms across our load tests, including the handoff to a person.
Rehearse on real calls
Improvement
Composer can replay an agent against last month of real conversations before you put it in front of anyone. Diffs are shown turn by turn.
Rehearsals run against a copy of the transcript, never the live queue, so nothing you test can reach a customer. Every difference is attributed to the policy or prompt that caused it.
Root-cause clustering in Signal
Feature
Signal groups conversations by underlying cause rather than tag, so ten tickets about one broken checkout finally read as one problem.
Clusters are rebuilt nightly and ranked by the volume they would remove rather than how often the words appear. You can push any cluster straight into Composer as a policy.
Customer-managed encryption keys
Security
Bring your own KMS key for transcripts at rest. Rotation and revocation stay on your side, and revoking a key locks the workspace within a minute.
Keys live in your own KMS and we never hold a copy. Revoking one locks the workspace within a minute and leaves an entry in the audit log that we cannot edit.
Sermo 2.0
Feature
A rebuilt voice pipeline with barge-in, interruption handling and warm handoff. Sub-second responses on live calls, and the agent keeps caller context when a person takes over.
This is the release the last eighteen months were building toward. Voice, chat, Composer and Signal now share one transcript store, so an improvement in one shows up everywhere the same day.
Faster handoffs
Improvement
Handoff to a person now carries the full transcript, the policy the agent applied and its confidence, so the first thing your team reads is not hello again.
The receiving agent sees what was tried, which policy applied and how confident the agent was before it stepped back. No more asking the customer to start again.
Transcript sync retries
Fix
Fixed a case where a helpdesk rate limit could silently drop a batch of transcripts. Sync now retries with backoff and reports any gap on this status page.
Retries now use exponential backoff with a ceiling, and any gap that cannot be closed is reported on the status page rather than failing quietly in the background.
Zendesk and Intercom write-back
Feature
Resolutions write back into the ticket that started them, with the macro, tags and internal note your team already expects.
Write-back respects the macros and tags your team already uses, so reporting built on those fields keeps working exactly as it did before Sermo was involved.
Get started
Ready to hear what your support could sound like?
We will build you a working agent on your own transcripts — and show you the numbers before you commit to anything.