Anthropic Found AI Makes Impractical Work Worth Doing

Why It Matters

The findings prove AI can unlock previously impractical work, expanding the value pool for enterprises, but scaling those gains requires addressing talent and readiness constraints.

Key Takeaways

• 27% of AI‑assisted tasks were previously unattempted by Anthropic staff
• Claude usage rose to 60% of work, delivering ~50% productivity boost
• Consecutive tool calls per task doubled, indicating deeper model autonomy
• New feature work grew from 14% to 37% of AI‑driven tasks
• CFOs cite productivity as top AI adoption driver, yet talent gaps persist

Pulse Analysis

Anthropic’s internal data provides a rare glimpse into how generative AI reshapes daily engineering work. By embedding Claude Code into 60% of tasks, the company recorded a 50% average productivity gain and a dramatic shift toward more sophisticated activities—consecutive tool calls per task rose from roughly ten to twenty‑one, and new feature implementation grew from 14% to 37% of AI‑driven work. This expansion effect means employees are tackling projects that were previously deemed too time‑consuming, from interactive dashboards to long‑neglected code refactoring, effectively turning low‑value chores into high‑impact deliverables.

The broader enterprise landscape mirrors Anthropic’s experience but highlights systemic bottlenecks. Surveys from OpenAI and EY confirm that three‑quarters of workers can now attempt tasks they once avoided, yet 71% of CEOs at billion‑dollar firms cite organizational readiness as the primary AI hurdle, and talent shortages affect 58% of CFOs. Cost pressures are already visible; Uber’s AI budget surged as AI‑written code now accounts for 11% of live backend updates, contributing to a 9% rise in R&D spending to $3.4 billion in 2025. These figures underscore that while AI can boost output, firms must invest in skill development and change management to realize its full potential.

Strategically, the data suggests a two‑track approach for companies seeking sustainable AI advantage. First, capture immediate efficiency gains by deploying models in high‑volume, low‑complexity tasks, then progressively expand into areas that were previously uneconomical, as Anthropic’s engineers have demonstrated. Second, address the talent and readiness gaps that Deloitte finds limiting deep transformation—only 34% of enterprises are using AI to fundamentally redesign core processes. By aligning AI adoption with talent pipelines, governance frameworks, and clear ROI metrics, firms can move beyond incremental gains toward a reimagined operating model that leverages AI as a core engine of innovation.