AI Extends Value Creation by Enabling New Types of Work

Enterprise discussions about artificial intelligence have largely focused on how much faster workers can complete existing tasks. However, internal research at Anthropic suggests this view is incomplete.

The company found that 27% of AI-assisted work involved tasks employees wouldn’t have attempted without AI – not because these tasks lacked value, but because the time investment previously made them impractical. This ‘expansion effect’ means AI isn’t just making workers more efficient; it’s enabling entirely new forms of output.

Internal Anthropic Study Reveals Key Findings

Anthropic surveyed engineers and researchers across its organization, conducting in-depth interviews and analyzing extensive usage data from Claude Code. The study showed:

  • Employees now use Claude in 60% of their daily work, up from 28% the prior year
  • Self-reported productivity gains average 50%, double the 20% seen previously
  • Usage has shifted toward more complex tasks – the number of consecutive tool calls without human input increased from 10 to 21
  • The share of tasks involving new feature implementation grew from 14% to 37%

Engineers reported using AI for activities like building interactive dashboards, addressing neglected code quality issues, and conducting exploratory research that previously wouldn’t have justified the time investment.

Broader Implications for Enterprises

These findings align with other research suggesting AI expands beyond simple task automation. OpenAI found 75% of workers could complete new tasks they couldn’t before, while EY’s survey showed 39% of organizations reinvesting productivity gains into R&D.

However, enterprises face challenges in realizing this full potential:

  • Organizational readiness is the primary barrier for 71% of companies with over $1 billion revenue
  • Talent shortages are cited as a leading challenge by 58% of CFOs (rising to 71% in services firms)
  • Cost control remains an issue – Uber’s AI budget has exceeded projections as usage scales

The key takeaway: when AI reduces the cost of analysis, coding, and research, work previously considered uneconomical becomes viable, unlocking new sources of value creation.