AI agents hit 50% in code, 4% in sales. Here is why that matters.
Anthropic published deployment data this week from nearly 1 million production AI agent tool calls. Software engineering owns 49.7% of activity. Sales and CRM sits at 4.3%. Finance at 4.0%. Legal at 0.9%.
The wrong take: AI works for developers, failed everywhere else.
The right take: Sales automation is next in line, and the pattern is clear.
Why code came first
Developers had the tools, the API access, and the autonomy to deploy agents without procurement cycles. They also had clear, repeatable tasks: write tests, review code, generate documentation. Sales tasks are messier: qualifying leads, handling objections, navigating CRM workflows. That complexity delayed deployment, but it did not kill the use case.
Anthropic's data shows 57% of technical leaders have already deployed AI agents, with 80% reporting ROI. Sales and marketing teams expect 46% of those gains in 2026. The infrastructure is being built. The sales applications are coming.
What this means for ANZ sales teams
AI SDR tools are already live: outbound sequencing, lead scoring, pipeline updates. Companies like Salesforce and Clari are shipping agent features into revenue workflows. The 4.3% figure is not a ceiling. It is early adoption data.
For sales leaders: the question is not whether AI agents will automate parts of your workflow. It is which parts, and how fast. Outbound prospecting is the obvious first target. Pipeline management and forecasting are next. CROs need to be mapping this now, not in 2027.
For individual contributors: if you are an SDR doing purely manual outreach, your role is changing. The reps who will thrive are the ones who learn to work alongside automation, focusing on the parts machines cannot handle: complex deals, relationship building, strategic accounts.
The numbers that matter
Anthropic raised $18 billion and hit a $40 billion valuation in 2024. Claude API usage is growing fast, particularly in enterprise. Internal deployments show 800+ AI agents achieving 89% adoption. That is not a pilot. That is production scale.
Sales automation lagged code automation by design. Now the infrastructure exists, and the deployment data shows enterprise buyers are ready. The 4.3% is not the end state. It is the starting line.