We release on a quarterly cadence and we have been explicit internally about preferring reliability over feature count. That is easy to say and awkward to publish, because it means some quarters look thinner than a roadmap slide would like. Here is the quarter.
Quality control got the most work
Quality managers can now define their own categories, attributes, weights and scorecard templates, rather than working inside a fixed structure. Alongside it, agents can formally dispute a completed evaluation. Those two shipped together on purpose: configurable scoring without a dispute route is a governance problem waiting to happen, and the dispute mechanism is what makes automated scoring legitimate with the people being scored.
Knowledge ingestion
Knowledge sources can now be built from a page and the links beneath it, turned into an article usable by both agents and AI, and refreshed on a schedule or on demand. The point is not the crawler. The point is that a policy change on your own website reaches the AI without anyone remembering to update a second place.
Identity and access
Single sign-on now spans the RoboDesk applications, and organisations can authenticate against their own identity provider — Azure AD, Okta or Google Workspace. This is the least visible item on the list and the one that comes up in every enterprise security review.
Interaction experience
A cleaner interaction list, a unified top bar, a social interaction view, and a clearer AI summary display in both the interaction view and the side panel. Individually small; together they are most of what an agent looks at for eight hours.
What did not ship
Collaborative interaction handling — several agents on one conversation with private whisper notes — is still in development and did not make the quarter. Forecasting and capacity planning in workforce management remain limited-availability rather than general. We would rather say that here than let a roadmap imply otherwise.