Indonesia’s wrong AI race risks leaving women behind

Indonesia’s AI ambitions are large, and President Prabowo Subianto’s government has made no secret of wanting the country to compete in the region’s AI buildout. Later last year, his government unveiled a National AI Roadmap (Opens in new window), naming artificial intelligence a central driver of productivity, innovation and growth on the road to Golden Indonesia 2045. The debate since has been about infrastructure, regulation and investment.

That is not really the country’s hardest challenge. Building AI capacity is, relatively speaking, the more straightforward task.

The harder question is one the roadmap barely addresses: getting far more of its working-age population into productive, quality jobs. This will determine whether Indonesia can build an economy that actually makes use of AI without closing off the lever it will need most to reach high-income status.

Development economists have long described how countries get richer through structural transformation. Workers move out of agriculture, first into industry, then into services such as banking, administration, education and business support.

In Indonesia, this pattern is well documented. Agriculture’s share of employment fell from 48% in 1991 to 27% in 2025, while the services sector’s share rose from 38% to 50% over the same period. Between 2005 and 2025, services alone accounted for roughly 37 million of the 52 million net new jobs Indonesia created, about 70%.

Women captured an even larger share of that expansion, filling about 20 million of the 26 million net new jobs added for women over the same period, according to calculations based on Statistics Indonesia’s Labour Force Survey data (Opens in new window). By 2025, services employed nearly 60% of all female workers.

Generative AI threatens to cut that stage short before Indonesia finishes building it. Older waves of automation displaced physical work – this one goes after desk work instead, drafting a memo, reconciling a ledger, replying to a customer. Stanford’s Digital Economy Lab pushes back on the doom framing (Opens in new window), pointing to firms that used AI to clear away a job’s tedious parts rather than replace the worker, and saw employees resolve more problems while customers grew happier. That holds for jobs built from many tasks (Opens in new window), where AI clears the routine and leaves judgment to people. It holds far less for jobs that were mostly one task, the way typing pools vanished once word processors arrived.

American back offices already show this, with customer service agents and bookkeepers proving most exposed. Indonesia’s entry-level clerical work sits in that category, absorbing the most women into formal employment.

The real risk is too few good jobs, with women having the most to lose.

This is not only an equity story. Indonesia’s female labour force participation rate sits at roughly 53%, well below Vietnam’s near-70% and Thailand’s high-50s. Indonesian women have historically climbed into formal work through exactly the clerical and administrative roles AI now performs fastest. Across ASEAN, women are more than twice as likely as men (Opens in new window) to hold occupations highly exposed to the technology, and a country cannot reach high-income status while leaving so many of its educated workers underemployed.

AI is unlikely to cause the kind of sudden job losses debated in economies with large outsourcing sectors, such as the Philippines. A greater concern is that Indonesia will not create enough good jobs to replace the ones AI reshapes. Entry-level office hiring slows, firms that adopt the technology pull further ahead of those that do not, and upward mobility narrows, especially for women. Yet the policy conversation keeps returning to training more AI engineers and courting more investment, which misses where Indonesia is actually constrained.

The workers are already present. What is missing is diffusion, getting AI tools into the millions of small enterprises and public agencies that have not yet adopted them, and redesigning vocational curricula so graduates learn to work alongside AI tools rather than train for the narrow tasks those tools now do instead.

AI is marketed as an enabler of flexible work, as though flexibility is automatically liberating. For many Indonesian women it is not. Indonesian mothers spend 13.7 hours a day on childcare against 3.8 for fathers, and mothers with eight or more hours of childcare support see their labour force participation rise by 29 percentage points and earnings by 66%, according to a 2024 World Bank survey of 3,000 Indonesian households (Opens in new window). Raising female labour force participation from 53% to 58% would add US$62 billion to the economy, a World Bank general equilibrium model finds (Opens in new window), if childcare spending rises to about 0.5% of GDP.

Flexible or remote work does little to lighten this care burden. Without childcare support, it simply relocates a woman’s paid labour into the same space as her unpaid labour.

Indonesia’s AI strategy should be judged by whether it makes existing workers more productive and widens – rather than narrows – the paths into formal work women have relied on for a generation. That outcome depends on diffusion, education and childcare, not on how much compute Indonesia builds or how many AI companies it launches.