Salesforce Unveils Koa and AIforce at Dreamforce

Salesforce used Dreamforce 2026 to position its platform as the connective layer for enterprise AI, unveiling Koa, its first reasoning model, and AIforce, which links Salesforce data, workflows and permissions to external interfaces including Claude, Slack, ChatGPT and Microsoft Teams. The company expanded Agentforce with tools such as Slackbot and Agentforce Coworker, introduced Claudeforce to bring Claude into Salesforce and Slack workflows, and promoted Slackforce Surfaces, which turns conversations and connected business data into dashboards, reports and other interactive tools. Salesforce expanded partnerships with Anthropic, Google Cloud, Amazon Web Services, Nvidia and Air India as it defended its CRM and SaaS business, arguing that trusted systems of record and enterprise data will remain essential even as AI agents change how employees use software. CIOs and architects face choices between native Salesforce AI and combinations of external models, retrieval systems, vector databases and orchestration, with security, permissions, governance, cost and flexibility central to the decision. AI safety and regulation were major themes, with Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and others urging transparency, stronger safeguards and industry standards after agent-related incidents. The conference also highlighted Certinia’s Veda platform and the potential for role-based agents and shared organizational intelligence to improve professional-services outcomes beyond individual productivity.