Beyond one-size-fits-all: Why Grab embraces multiple AI coding assistants

Software engineering is undergoing a transformation thanks to AI-powered coding assistants.

These tools leverage artificial intelligence and machine learning to assist engineers in writing, reviewing, and optimising code. They can provide intelligent code suggestions, automate repetitive tasks, detect errors, and even generate documentation, allowing teams to collaborate, innovate, and deliver code with more speed and efficiency.

AI-powered coding tools are a fairly new phenomenon. Their long-term impact on the business outcomes of tech companies, or the end-user experience, still needs to be observed. However, it’s a shift within software engineering teams that’s worth paying attention to.

In this article, we explain why we’ve taken on a multi-tool approach, and which tools we’ve found helpful as we navigate this transformation.

Grab’s multi-AI approach

Most companies start their AI coding journey with a single assistant. It’s a straightforward approach—pick a tool, integrate it, and scale. But at Grab, we took a different path.

We recognised early that engineering challenges are not monolithic. Different problems demand different strengths—much like how a craftsperson relies on a variety of tools, not just a single wrench. Instead of limiting our engineers to one AI assistant, we expanded the playing field and decided to explore multiple AI tools to see how they fit in our engineers’ day to day work. This strategy also allows us to pivot quickly, making sure we have access to the best available tools in a fast-evolving landscape.

Today, tools like ChatGPT Enterprise, GitHub Copilot, Sourcegraph Cody, and Cursor work in tandem to enhance our engineering workflow.

And adoption is growing fast—more than half of our engineers use an AI coding assistant regularly, allowing them to improve their productivity.

How AI coding assistants are changing the game

Engineering tasks vary widely, and the needs of our engineers shift accordingly:

Many tasks benefit from context-aware suggestions. Coding assistants can, for instance, auto-complete code intelligently by taking into account the existing codebase and programming environment, or internal documentation, thus improving speed and efficiency of the coding process.

Some tasks need quick code generation. For example, engineers might want to build a quick prototype to test something, or they want to get through repetitive coding tasks quickly.

Others require deep code understanding. Imagine having to assess a large codebase during a security audit, or transferring ownership of a complex software product to a new team lead. 

Coding assistants can also help in these scenarios, for instance, by auto-generating standard code, or creating mock data to run tests with, or helping query large codebases for specific functions and relationships.

What’s even more exciting is that these tools open up new possibilities for engineers at all skill levels. While AI still requires human oversight, coding assistants are lowering the barrier for coding, making it easier even for non-engineers or junior engineers to focus on understanding the problems and designing solutions.

A tool for every task

Rather than force-fitting a single AI into every scenario, we enable multiple AI coding assistants to complement each other. This ensures engineers can choose the best fit for their task at hand.

Based on real feedback from our engineers, here’s how AI tools contribute to our workflows.