Anthropic has officially entered the business intelligence game, launching a new suite of reporting tools designed to answer the single question keeping enterprise CTOs awake at night: what are we actually getting for our token bill? The move follows a staggering financial disclosure showing 900% net revenue retention, meaning spending from the same customers in Q3 2025 grew nine times in Q3 2026. While the coding tool Claude Code drove this growth, Anthropic is now pivoting from pure capability to demonstrable ROI.

The Data-Driven Lock-In Strategy

The new tools allow managers to view team workflows over specific date ranges, tracking output volume, cost per task, and breakdown points in the automation pipeline. Crucially, Anthropic has committed to not using this internal analytics data for model training by default, a necessary concession to enterprise privacy concerns. The system also generates recommendations for improving results, effectively acting as a lightweight, automated Forward-Deployed Engineer (FDE). By creating pooled group budgets, Anthropic is giving enterprises the granular control they’ve been demanding as AI costs become a significant line item on the P&L.

Jellyfish Data Confirms the Spend Surge

The timing aligns with new data from Jellyfish, a developer tools company that analyzed 276,000 users. Their report reveals that tools built on Anthropic models, including Claude Code, now account for nearly a third of all proposed code changes in their dataset. More alarmingly for budget holders, spending on these agentic tools has grown 10x since the start of the year, now representing roughly 2% of total engineering costs. Nicholas Arcolano, Chief Scientist at Jellyfish, noted that while 2% might seem small, the 10x growth rate is a shock to most CFOs who didn't plan for such aggressive scaling.

Competition and Switching Costs

Despite the hype around OpenAI’s Astra model, Anthropic’s enterprise moat appears deep. Nikolas Albarran, AI product lead at Jellyfish, observed that while social media chatter suggests Codex is superior, real-world adoption shows Anthropic’s incumbency advantage. "This stuff just worked and worked well, and a lot of people just put a lot of effort into standardizing workflows around it," Albarran stated. The switching cost is no longer just about model quality; it’s about the sunk cost of workflow standardization and the subtle quirks of how different models handle configuration files like agent.md. Once an enterprise has tuned its pipelines to Anthropic’s specific behaviors, the friction to switch becomes prohibitively high.

Key Takeaways

  • Anthropic reported 900% net revenue retention, driven by explosive growth in Claude Code usage.
  • New reporting tools offer granular visibility into token spend and workflow efficiency, promising not to use data for training.
  • Jellyfish data shows Anthropic-based tools now handle nearly a third of code changes for surveyed developers, with costs up 10x YTD.
  • Enterprise switching costs are rising due to workflow standardization and model-specific quirks, not just raw model performance.

The Bottom Line

Anthropic is weaponizing transparency. By giving CFOs the tools to audit their own AI spend, they are turning a cost center into a measurable productivity engine, making it exponentially harder for competitors to steal accounts that have already invested in optimizing their Claude workflows.