Anthropic published a new report on June 25, 2026 showing Claude Code and Cowork usage surging across enterprise environments. The headline finding: AI agents are increasingly handling multi-step workflows that previously required constant human oversight.
What the Report Says
Anthropic argues that agentic AI changes the basic unit of knowledge work — from short, self-contained interactions to delegated, long-horizon tasks. Agents can operate independently for minutes or hours, orchestrating tool calls, interacting with environments, and iterating toward a solution without a human at every step.
Claude Code Surges in the Enterprise
Within sampled enterprise customers, Claude Code has become the default environment for a rapidly growing share of software work:
- Long-horizon tasks — more than 70% of sampled requests now correspond to work that would take a person more than an hour
- Parallel agents — heavy users run many agent tasks at once across a single day
- Deeper usage — output-token volume per active engineer has multiplied several times over the past year
Cowork and Multi-Step Workflows
Cowork mode lets teams and agents share a workspace, with agents picking up steps across planning, coding, review, and verification. Anthropic credits the surge to agents finally becoming reliable enough for workflows that previously required a human at every stage — the difference between a toy and a tool that sticks.
The Rise of the Non-Developer
The fastest-growing group is not developers. Since adoption broadened, non-developer usage in enterprise accounts is climbing faster than developer usage — recruiters, lawyers, and analysts now lean on agents for automation, data transformation, debugging, and structured analysis. Over one-fourth of agent work in business functions is now engineering or coding.
What This Means
The report’s economic point is that agents expand what a worker can do, lowering the cost of crossing task boundaries. For employers redesigning workflows and for workers deciding which skills to build, the direction is clear: long-horizon, agent-delegated work is becoming the default — in enterprises and beyond.