ServiceNow’s October 21 announcement makes its long-running employee agent generally available across EmployeeWorks Foundation, Advanced and Prime, without an additional charge. Mac is supported; Windows is expected within weeks. AI Passport and its NVIDIA enforcement integrations remain a future release.
Kognitos’s October 21 announcement introduces 61 finance and banking agent roles backed by a 515-procedure library. Availability is preview for existing customers. Its catalog labels all six Customer Service Representative procedures “to build.”
On October 22, the Agentic AI Foundation announced Agent Skills from Anthropic, Open Responses from OpenAI, OpenEnv from Hugging Face and OpenSandbox from Alibaba as hosted projects. The additions bring its total to ten.
The October 22 launch adds Google Cloud IAM access and billing for Claude Platform capabilities, including long-running agents. Anthropic operates this offering; the existing Google-managed option remains separate.
Anthropic announced Claude Duet on October 21: a proactive assistant that keeps working after you close the app. Access starts through a US Pro and Max waitlist; background work shares your Claude usage limits.
Cisco’s October 22 Webex Contact Center update adds nl-BE and en-SG to General Availability for AI Assistant. Nine further locales enter Controlled Availability. Feature, channel and configuration limits still apply.
Microsoft’s October 21 announcement puts Contact Center Intelligence in public preview globally except Austria. It adds AI to existing contact centers through connectors; real-time quality evaluation and human-agent assist remain on the roadmap.
Ada’s October 22 release adds median and slowest-ten-percent API response times to its usage report, plus the api_tool_performance MCP metric. The recorded duration measures API waiting, rather than the whole customer response.
Ada’s October 22 release adds a queue banner for customers waiting after a Salesforce Messaging handoff. It displays estimated wait time or queue position, depending on the signal configured for that Messaging Channel.
Decagon’s October 22 announcement introduces experiment blocks that lock the variants under test while leaving other agent content editable. Results count conversations exposed to the tested step, rather than all traffic.
An October 20 vendor account describes IncidentWatch and IncidentSweep in its own support operation. Engineers confirm incidents and approve recipient batches; the account does not establish a public launch of these workflows.
Ada’s October 19 release separates AI-agent and human-agent satisfaction metrics from the HANDOFF conversation filter. Assistants that used that filter to select whose score they wanted need a different metric type; unchanged requests can now mean something different.