How AI is Transforming Analytics at Grab
AI accelerates analytics processes and simplifies decision-making at Grab.
1The Role of Analytics at Grab
At Grab, analytics sits at the heart of nearly every important decision. The company's goal is to democratize intelligence, enabling everyone to quickly access reliable answers. Over the past two years, model capabilities have crossed a threshold that makes this shift possible. Now, agents can complete tasks that used to take a week in just minutes.
2Speeding Up Analytical Processes
At Grab, analytical processes are no longer limited by how fast an individual can write code or perform deep analysis. Instead, they are limited by how quickly the right question can be framed, the right answer judged, and the right decision influenced. This shift has transformed analysts' roles from producers to advisors and strategists.
3AI-Assisted Analytics
Grab has created its own analytics ladder inspired by Dan Shapiro's five-level model for AI coding. This model defines where human judgment comes into play and how much the agent should take over in an analytics loop. Human judgment plays a critical role at every level.
4Levels of Autonomy in Analytics
At Grab, analytical processes operate at different levels of autonomy, from L2 to L5. At L2, agents draft SQL queries, while at L5, they autonomously detect anomalies and uncover opportunities. Human judgment and accountability are maintained at every level.
5Five Core Capabilities in Analytics
Five core capabilities drive the climb up the analytics ladder at Grab: execution, knowledge, control, review, and learning. These capabilities enable agents to focus on more complex and high-stakes tasks.
6Spartan: Autonomous Analytics Workflow
Spartan is a fully autonomous analytics workflow used at Grab. Integrated into platforms like Slack, Spartan is used by non-analysts as well. From ad sales teams to country teams, everyone uses this tool for daily tasks.
7Analytics with Natural Language Processing
Spartan uses natural language processing technology to understand and analyze users' questions. For example, when a commercial manager asks why revenue dropped in the Philippines, the system automatically analyzes and answers the question.
8Frameworks and Skills in Analytics
Behind Spartan are over 50 skills and more than 120 analysis frameworks. These frameworks determine how agents analyze questions and which metrics to use, making analyses more consistent and reliable.
9Knowledge Base in Analytics
Analytical processes at Grab rely on a vast knowledge base. This base determines what data agents should search for, how to query it, and how to think about it. The knowledge base makes analysts' knowledge shareable.
10Learning Loop in Analytics
At Grab, agents learn from their mistakes and improve themselves. When an agent makes the same mistake twice, it is added to context documents, golden datasets, and evaluation criteria. This prevents similar mistakes from recurring in the future.
11The Importance of Human Judgment in Analytics
No matter how automated analytical processes become at Grab, human judgment always plays a critical role. Humans are responsible for framing the problem, defining metrics, and making the final decision.
12The Future of Analytics
AI-assisted analytics at Grab will continue to evolve, making decision-making processes faster and more effective. This technology will enable analysts to take on more strategic roles.
Original article: Grab Tech — Grab'da Yapay Zeka Analitiği Nasıl Dönüştürüyor?