GoTo is narrowing its artificial intelligence (AI) investments to applications that can improve customer conversion or reduce the cost of serving customers, ahead of a wider roll-out of AI-driven products and services in 2027, president director and group CEO Hans Patuwo said at the Tech in Asia Conference 2026 in Singapore on Tuesday (15/09).

The Indonesian technology group, which is the parent company of Gojek, has been experimenting with AI since 2024. The company began deciding in 2025 which AI capabilities it should develop itself and where it should rely on external providers.

“We are not trying to solve world hunger,” Patuwo said, explaining GoTo’s focus on smaller and cheaper AI models when they are suitable for a task.

“AI should transform your business, not make you become an entirely different business,” he added.

GoTo Focuses AI Projects on Business Results

After an initial period of experimentation, GoTo has settled on two main measures for its AI projects: whether they improve user conversion or reduce the cost of serving customers.

“There is a period of experimentation where we let 1,000 flowers bloom, but that cannot carry on forever,” Patuwo said.

AI projects at GoTo compete for resources in the same way as other products developed by the company.

Some of the projects are already producing results. AI now handles 35 per cent of GoTo’s customer complaint tickets, and satisfaction scores for these interactions are 20 per cent higher than the norm.

AI bots used for collections in GoTo’s financial services business have also increased collection rates by three percentage points. Another 10 to 15 applications are in the pipeline.

GoTo Uses Smaller AI Models to Control Costs

GoTo uses a mix of frontier models, open-source technology and smaller models that are further trained on Indonesian data, rather than relying entirely on the largest commercially available models.

Patuwo said the company has used models with seven billion and nine billion parameters and is launching a 27 billion-parameter model.

The company’s AI cost calculations are also affected by currency. GoTo earns revenue in Indonesian rupiah, while tokens consumed from many overseas AI providers are paid for in US dollars.

“If you continue to consume in US dollars, that’s not going to be very sustainable,” Patuwo said.

For new applications, GoTo may initially pay to use a frontier model. Once it understands what an application requires, the company can move to smaller models or further train open-source models with Indonesian data, including local language and cultural nuances.

For many applications, smaller models can provide sufficiently fast responses at a much lower cost.

Patuwo cited a production system processing 10 million transactions a day as an example. If every transaction requires calls to an external AI model, the costs can increase quickly.

“We will never be the best, but we could be the best for our particular application,” he said. “We will never be the cheapest, but we could be the cheapest and the lowest latency for some of the services that we need.”

GoTo Takes a Gradual Approach to AI Adoption

GoTo is cautious about handing entire processes to AI. Regulatory work and high-touch interactions, particularly those involving Gojek drivers, are among the areas where the company is less likely to be an early adopter.

AI applications are introduced in stages. GoTo generally stops at around 70 to 80 per cent adoption initially instead of immediately moving to full automation.

The company’s AI expansion comes as its financial performance improves.

GoTo reported a net profit of 252 billion rupiah (US$14.3 million) for the second quarter ended June, its second consecutive profitable quarter, compared with a 375 billion rupiah loss a year earlier.

Net revenue rose 31 per cent to 5.7 trillion rupiah, while adjusted earnings before interest, taxes, depreciation and amortisation more than doubled to 1.01 trillion rupiah.

AI Is Also Changing GoTo’s Hiring Approach

AI is changing how GoTo approaches its roughly 4,000-strong workforce.

Patuwo said GoTo has generally not been backfilling roles left vacant by departing employees for the better part of a year. Managers can still hire, but he has set a higher bar for recruitment.

“(What I tell) them is: ‘I’m not against your hiring. Go ahead, go hire, as long as you can promise the candidate that we are offering them a career, not a job.’”

Patuwo said AI could affect jobs in three ways. Existing roles could be “revolutionised” as workers become more productive, employees could be retrained for jobs requiring greater human interaction, and some roles could eventually be retrenched.

He said careful hiring and natural attrition could reduce the need for the last of these.

GoTo could eventually sell some of the technology it is developing for its own operations.

Patuwo said the group has historically been a “tech-enabled consumer company”, but could increasingly become a “consumer tech company”, with some technology initially developed for its own needs later being “productised externally”.

PHOTO: GOTO

This article was created with AI assistance.

We make every effort to ensure the accuracy of our content, some information may be incorrect or outdated. Please let us know of any corrections at [email protected].