What changed in AI-support tooling this week. Every item was checked against its source before it entered this list.
Anthropic- launchedClaude Fable 5.1 and Mythos 5.1(Sept 1): a 1M-token context window by default, always-on adaptive thinking, and prompt cache reads cut to $0.25 per million tokens.
Salesforce- theWinter '27 release(Aug 31) pushes agents from answering questions toward running full workflows - resolving service cases, booking appointments, underwriting risk - plus an Agentic Commerce Search feature the company reports lifted conversion 13%.
Intercom- a new pair of webhooks fires when translated replies are saved, delivering the text in each language alongside the original (Sept 2 changelog entry); the next day, background activation let more complex multi-step Series deploy without errors (Sept 3).
Decagon-Campaign Composerlets teams configure channels, hours, retry logic, and suppression rules for outbound campaigns without engineering help (Sept 2); a few days earlier its agents gainedlive web search through Perplexity's Search API, letting them answer time-sensitive questions with cited sources instead of escalating (Aug 28).
KT- the South Korean carrier won a project torebuild Woori Bank's AI chatbot and consultation-bot infrastructure, linking chat, consultation-bot, and AI-banker channels through an "Agent Connector" that carries conversation context across them (Aug 31).