[2/2] How AI Is Transforming Work at Anthropic: Insights for Engineers (Category AI)
The second half of this review looks at what managers can learn from Anthropic’s report.
ROI and productivity The report describes a substantial economic effect. An internal survey found roughly 50% higher productivity per employee, supported by activity metrics: daily code changes, measured as PRs per engineer, rose 67% after Claude’s introduction. The team was doing considerably more in the same time. Around 27% of tasks supported by Claude had not previously been done because resources were lacking; these improvements and experiments now enhance the product and open new possibilities.
AI augments rather than replaces people Despite these gains, people remain essential. Most engineers use Claude daily, but can fully automate only up to 20% of their work. The rest needs human task definition, oversight and corrections. AI accelerates routine parts and supplies drafts; experts finish the work. Claude is an accelerator, not an autonomous employee.
Changes in team dynamics Developers now ask Claude before asking colleagues. This reduces small interruptions for experienced staff and lets them concentrate on harder problems. But it can weaken collaboration and mentoring: newcomers rely on AI instead of asking senior colleagues. Without deliberate attention, knowledge transfer may suffer. Managers should consider formalising mentoring and finding new ways to develop junior colleagues as AI takes over part of the teaching.
The risk of skill loss AI broadens engineers’ capabilities, but infrequently used fundamentals can become rusty. Some staff acknowledge less practice in writing code themselves and working through algorithmic details, though it is not yet a major obstacle. Some deliberately solve tasks without Claude to stay in practice.
Workforce planning and training The engineer’s role is changing. Many effectively manage AI agents, directing several instances at once. Work moves to a higher level of abstraction, with less manual execution and more oversight and coordination. As one team lead joked, the job is now to be responsible for “5 or 100 copies of Claude” rather than one developer. Systems design and guiding AI may become more central, making clear task definition and answer verification especially valuable skills.
Uncertainty and adaptation Leaders need to prepare for several scenarios. Even engineers struggle to say how their roles will change in 3–5 years. Many have mixed feelings: things look good today, but AI could take over more tasks tomorrow. Some enthusiasts expect better safeguards, learning built into tools and people growing alongside the technology. The shared requirement is flexibility.
How Anthropic is preparing The company is discussing working rules for AI, ways to encourage collaboration and knowledge sharing, and support for professional growth. It is also considering new career paths and internal reskilling programmes as model capabilities expand. Anthropic is extending its research beyond engineers and helping external partners adapt educational programmes for an AI-shaped future.
#Engineering #Software #Processes #Productivity #Economics