Anthropic shipped Claude Sonnet 5 at the end of June 2026, positioning it as its most agentic mid-tier model. The practical shift: mid-priced models can now carry multi-step workflows like intake, quoting and reporting that used to need the flagship tier, which changes the economics of automation for smaller firms.
The interesting part of the Sonnet 5 release was not the benchmark chart, it was the pricing tier. Agent-grade reliability used to be a flagship feature, priced accordingly. When a mid-tier model can plan a task, call your tools, recover from errors and finish the job, automation stops being an enterprise luxury.
For a 20-person company, the arithmetic moved. A workflow that costs a few thousand rupees a month in model usage and saves forty staff hours pays for itself in the first week.
Where to point it first
The best first agents are boring: enquiry triage that reads, classifies and routes messages, quote drafting from your actual price list, weekly report assembly from your existing dashboards.
Keep a human on the send button for anything customer-facing. The goal is drafts that are right nine times out of ten, reviewed in seconds, not autonomous replies you discover a week later.