The productivity debate around enterprise AI has boiled down to one question: How quickly can employees complete current tasks? AnthropicInternal research suggests that framing leaves something out. Company Found 27% of AI-powered work within Anthropic comes from tasks that employees would not have done without AI. This work was not impractical because it lacked value. The time cost made it impractical.
What the data showed
Anthropic Surveyed Engineers and researchers across the organization conducted 53 in-depth interviews and analyzed 200,000 internal CloudCode instances. Employees reported using Claude for 60% of their work, and productivity gains were estimated at approximately 50%, compared to 20% the previous year. Usage increased from 28% of daily work to 60% during the same period.
The output data is more realistic. Across nearly every task category, employees reported spending slightly less time on each task but significantly greater volume of production. Claude Code usage shifted toward more complex work: the average number of consecutive tool calls for a model completed without human intervention nearly doubled from 10 to 21, and the share of tasks that involved implementing new features increased from 14% to 37%.
Engineers described using AI to create interactive dashboards, scale deprioritized projects, fix long-neglected code quality issues, and perform exploratory research that doesn’t justify the time cost manually. One researcher described running several instances of Claude in parallel to test different approaches simultaneously, treating the model less like a faster car and more like a fleet.
OpenAIEnterprise research Found A similar pattern, with 75% of workers surveyed stating that they could complete new tasks that they had not previously been able to perform. EY US AI Pulse Survey Found That 39% of organizations were reinvesting AI-driven productivity gains into R&D, suggesting that the impact of scale extends beyond completing individual tasks.
Where companies struggle
The broader picture of the organization is less consistent. PYMNTS INTELLIGENCE Found 71% of executives at companies with annual revenues of at least $1 billion identified organizational readiness as the primary limit to AI performance. Only 11% said that technology itself was an obstacle. PYMNTS INTELLIGENCE I mentioned 58% of CFOs described talent shortages as a major challenge, and this percentage rose to 71% among service companies.
Advertisement: Scroll to continue
Cost control is a parallel pressure. piments I mentioned Uber’s AI budget has exceeded expectations with increased internal use of Claude Code, with nearly 11% of live updates to its back-end systems now written by AI agents and R&D expenses rising 9% to $3.4 billion in 2025.
Economics of implementation
When AI reduces the cost of analysis, documentation, coding, and research, work that previously fell below the validity threshold moves above it. Deloitte Found Only 34% of organizations are using AI to deeply transform core processes and products, while the remaining two-thirds are achieving efficiency gains without redesigning core processes.
Internal results have limitations. Anthropological engineers have early access to pioneering models, work in a stable field and build the technology themselves. The company acknowledged that the results are not directly generalized to other organizations. PYMNTS INTELLIGENCE Found 34% of CFOs in large companies indicated that productivity is the main reason for adopting artificial intelligence. Anthropic plans to expand the research beyond engineers to understand how AI will impact roles across the organization, with more results expected in 2026.
