AI’s productivity promise meets the harder question of who benefits

At the IMF-World Bank meetings in Bangkok, speakers argued that AI’s productivity gains remain uncertain, while its effects on workers and the distribution of those gains need attention now. Anthropic’s Jack Clark said infrastructure, skills, wider adoption and risk governance are all needed if AI is to benefit more than a narrow slice of the economy.

Watch Desk analysis

What happened

China Daily reports that IMF Managing Director Kristalina Georgieva said current data does not show widespread job losses, but warned that AI could challenge the historical pattern of new technologies creating more jobs than they destroy. She called on companies to retrain workers and change business models, and said governments should communicate risks clearly and consider helping businesses meet transition costs.

Clark argued that productivity should include better decision-making, not just faster output. He identified four conditions for broader adoption: digital infrastructure and reliable electricity, a skilled workforce, ways to spread AI across industries, and effective risk governance. He also pointed to robotics as a possible route to gains across more kinds of work.

Why it matters

The debate is shifting from what AI might do to what needs to be in place for its benefits to reach workers and businesses. Reliable power, practical skills and wider access are less glamorous than a new model launch, but without them the gains may cluster where the infrastructure and expertise already exist.

There is a tension here worth watching: Georgieva warns that even new jobs could be taken over by AI, while Clark sees the technology improving decisions and extending productivity into physical work. Neither forecast is settled. The concrete question is whether companies redesign work and retrain staff, or simply add tools to old processes and hope the arithmetic behaves.

Our read

This is a useful argument about the conditions around AI, not evidence that productivity has already been transformed. The sharpest test is whether employers and governments can show that adoption improves work and spreads opportunity, rather than merely boosting output in a few well-equipped corners. A training budget and reliable electricity may not make a dazzling keynote, but they are rather good places to start.

What to watch

• Whether companies invest in worker retraining and redesign jobs around AI.
• Whether governments publish measures of AI’s effects on employment and productivity.
• Whether access to infrastructure and AI tools broadens beyond already well-served regions.

Discussion spark: Should governments help businesses pay for AI-related retraining, or should employers bear the full cost of preparing their workers?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.