Grab uses AI agents and non-programmer coding to speed product development
Grab is using artificial intelligence to compress its product development cycle, with engineers delegating tasks to AI agents and non-technical staff building prototypes without writing code.
The ride-hailing and delivery super app's Virtual Store Manager, launched in April, exemplifies the approach. The tool analyzes closed-circuit television feeds to monitor store conditions across multiple locations from a single dashboard, eliminating the need for manual video review.
The product's creator, Muhammad Hanif Naufal Eka Wiratama, is not a software engineer. He built the prototype using vibe coding-a method where AI generates applications from written prompts rather than hand-coded instructions.
Prototyping versus production
Grab's chief technology officer Suthen Thomas said vibe coding's primary value lies in speed. "We use it the way a designer uses a high-fidelity mock-up to quickly get something in users' hands and observe how they actually behave, rather than asking them what they want," he said.
The company maintains a clear distinction between rapid prototyping and production code. Thomas compared vibe-coded prototypes to hot-wiring a car: "It gets you moving, but you wouldn't drive it at scale."
Production code requires architecture review, testing, verification, and consideration of how systems interact with back-end infrastructure. "We maintain a strict boundary between casual prototyping and actual production to ensure our engineering standards remain uncompromising," Thomas said.
Agents handling engineering tasks
Grab's broader strategy centers on agentic engineering, where AI agents execute assigned tasks and humans review the completed work rather than collaborating line-by-line with AI assistants.
"The engineer is able to assign higher-level tasks or goals to the AI agent, which then might take minutes or hours to complete," Thomas said. "The human only steps in at the end to check the finished work."
The company calls this approach part of building a "cyborganisation"-borrowing from the science fiction concept of a cyborg-where human and artificial intelligence components work together as a stronger entity.
Measurable productivity gains
Grab conducted a nine-week generative AI upskilling program across the company in 2024. More than 90 percent of engineers now use AI coding tools daily.
Productivity metrics show tangible results. Engineers submit approximately 40 percent more merge requests-code integration requests that indicate faster application completion. Development time has dropped 20 to 30 percent.
Non-engineering teams are also adopting AI to automate tasks and workflows, though Thomas did not specify productivity gains for those groups.
Maintaining standards at scale
Thomas warned against normalizing deviations from engineering standards as AI-assisted work accelerates. The concept, known as normalization of deviance, refers to rule-breaking becoming culturally accepted over time-a concern that emerged from the Challenger space shuttle disaster.
Given Grab's scale operating across Southeast Asia, the risk increases as speed increases. "The faster you can build, the more important the discipline around what you accept," Thomas said.
Grab's advantage includes data from over 20 billion rides and orders plus real-time signals, which inform its ability to sense, predict, and respond to regional operational complexities.
Your membership also unlocks:
