Salesforce
Salesforce introduced Koa Decisions on November 6, the next version of HyperClassifier, which routes work inside Agentforce. In Salesforce's test it returned a choice in 26 milliseconds, against 1.4 seconds for a general-purpose LLM. It handles images and audio, with plans for Flows, Apex and an API. I would ask how often the cheap answer is the wrong one.
Talkdesk announced Talkdesk Voyager, a long-horizon orchestration product on its Customer Experience Automation platform. It coordinates AI agents, employees, and enterprise systems to run goal-driven journeys over days, months, or years, with pre-built blueprints for healthcare, retail, and financial services. The company has not given a general availability date. Customers will judge whether those journeys reach the goals Talkdesk sets.
ServiceNow has launched Autonomous CRM Engine, an AI-native tool that maps CRM processes, redesigns workflows and helps deploy ServiceNow CRM in weeks. The company says it can also monitor configurations after deployment. Separately, ReSource Point of Sale says ServiceNow CRM has helped cut average resolution times from 8–10 hours to three hours.
Microsoft added a historical dashboard for the Quality Evaluation Agent in Dynamics 365 Contact Center to its roadmap. It pulls completed evaluations into one view, tracking quality scores and trends across chosen date ranges with up to two years of history. General availability is slated for March 2027. AI scoring every call is useful; proving it's improving is better.
Microsoft has been named a Leader in Gartner's 2026 Magic Quadrant for CRM Customer Engagement Center, making it the only vendor in both that report and the CCaaS Magic Quadrant. Dynamics 365 Customer Service now folds in workforce engagement management and Copilot skills. Two quadrants make a great slide; buyers should still demand proof that the seams have gone.
Salesforce Catalog is a self-updating index of how an org is built, covering objects, fields, flows, Apex, permission sets and Data 360 insights. Agents and assistants, including Claude, ChatGPT and Cursor via an MCP server, query it before answering, running as the signed-in user. It needs a Data 360 edition. Overdue: confident agents guessing field names are expensive.
