As a Director at Glovo, a Delivery Hero subsidiary with over 4,000 employees in 23 countries, explai co-founder Dirk faced a challenge: No matter how large his team was, even with 250 developers and data experts, the workload was even bigger. At the same time, business departments were pressing for real-time insights into Glovo's three-sided marketplace with millions of monthly food orders, prepared and delivered by hundreds of thousands of restaurants and drivers.
As one of the first companies in Europe, his team piloted an AI-based data chatbot in early 2023 to manage the flood of requests. In September 2024, Prosus, creator of the data agent "Toqan" and Delivery Hero's largest shareholder, published project results at Barcelona's OLX Masterclass after 1.5 years of productive operation.
The results are impressive.
Toqan answered 3 million questions from 21,000 users across all lines of business in company-wide chat channels (Figure 1). More than three-quarters of the answers were helpful, and the rate of hallucinated results (users pressing the "Pinocchio emoji") dropped from 9.2% to 1.5%. Users save 47 minutes per day.
It's interesting to see how management expectations and user experiences diverge (Figure 2). Upper management hopes that AI will automate entire areas of responsibility with corresponding staff reductions. This myth is often created by consulting and software giants — Accenture alone earns over 1 billion USD per quarter with AI promises.
The practice looks different: Time savings happen in chunks at many different points in time, through ever more skillful use by employees who thus keep more focus and gain autonomy.
The following Figure 3 shows some of the typical questions: They range from company-wide to very case-specific. Two years ago, under Dirk's aegis, many of these required time-consuming support beyond standard reports.
Users of the Toqan data agent report that they receive data insights 74% faster in 6 minutes instead of the previous 23 minutes (Figure 4).
Employees thus don't become redundant, but rather more valuable, because they can now work more substantively and strategically without being blocked in their workflow by waiting for colleagues.
Dirk wished he could have offered his teams such an acceleration and simplification technology during his 20-year data career. Many promising projects wouldn't have fallen by the wayside, and both his own team and internal customers would have been more satisfied.
This doesn't have to happen to you in 2025. What potential do data agents have in your company? Find out, and contact us to test explai on your data.