EuroFinance

At Delivery Hero, a German multinational online food ordering and food delivery company, innovation in treasury is not an abstract ambition—it is built, tested and applied in practice. At the 34th EuroFinance International Treasury Management conference in Budapest, Christian Schmahl, senior director, Treasury, and Can Akcal, senior manager, Treasury, shared how the global food delivery company has used artificial intelligence, automation and data-driven thinking to reshape its treasury operations. Their story involves experimentation, collaboration and a willingness to rethink how treasury work is done.

Delivery Hero operates on a large scale, across roughly 70 countries and serving close to two billion people worldwide, and the global treasury team plays a central role in managing liquidity, risk, users and payments across a complex structure.

“You see riders with our brands on the streets in many parts of the world,” Schmahl said. “But behind that, there is a very complex cash flow cycle that runs through our treasury accounts.” It was this scale and complexity that pushed the treasury team to rethink how technology could support its work. What followed was what the team described as a “hat trick” — three distinct but connected use cases of AI and automation that delivered tangible value.

Also read: Innovation, inspiration and real-world case studies: Treasury’s insights from Budapest

A treasury GPT built for real use

The first innovation was the development of a treasury-specific GPT tool, embedded directly into its internal messaging platform. Unlike a generic chatbot, this tool was trained using Delivery Hero’s own treasury policies, manuals, training materials and FAQs.

“We have many treasury teams across regions, and they constantly ask questions about processes and policies,” said Akcal. “So we worked with our transformation and automation team to build something that could answer those questions instantly.”

The tool allows employees to ask questions in their own language, with responses generated in the same language. “It’s not a chicken translation,” Akcal noted. “We tested it in Turkish, German and Korean, and colleagues confirmed the quality was very strong.”

The system continuously learns and is monitored using a scoring mechanism. “If the score is around 0.7, that’s already good. Today, we’re closer to 0.9 on average,” he said. “It’s acting like us — like a treasury hero.”

The impact has been significant. Questions that once came through tickets or emails are now answered instantly. “There are no more questions from the regions,” Akcal explained. “They ask the bot directly, in their own language, and get an answer in seconds—resulting in no more questions from regions, no language barrier, and a sustainable high-quality solution.”

Also read: AI’s bright spot for trade amid global tensions

Automating repetitive work with RPA

The second goal in the hat trick focused on robotic process automation (RPA). Schmahl described how treasury processes like intercompany funding were once handled manually, often involving repetitive copying and pasting between platforms.

“I remember when I took over the process, I did it myself just to understand it,” he said. “After a few months, I thought, this can’t be right. I went to the transformation and automation team and said— we need to automate this.”

Within weeks, they implemented an RPA solution. The system automatically transferred data from working files, eliminating manual input and significantly reducing errors. “What used to take me four hours now takes seconds,” Schmahl said. “And it doesn’t make mistakes.”

The benefits extended beyond efficiency. “It’s about quality and reliability,” he added. “And it frees up time so people can focus on more meaningful work.”

An early use case was a daily cash report for the CFO. Previously, this involved manually compiling tables and sending emails each morning. Today, the process is fully automated. “At 9am, the CFO gets the report automatically,” Schmahl said. “Signed: your treasury bot.”

Across the group, Delivery Hero now operates about 150 RPA bots, saving approximately 17,000 hours per month. “It’s only the beginning,” he said. “But you have to start somewhere, communicate it internally, and let people see what’s possible.”

Also read: AI handles the numbers but treasury handle the why

Using generative AI for thought leadership

Completing the hattrick is using generative AI to support thought leadership. Schmahl described how he was invited to contribute an article to Harvard Business Manager but struggled to find the time.

“I’m a treasury expert, not a journalist,” he said. “Writing in that style is not easy.”

Instead, the team worked with their transformation and automation colleagues to build a structured prompt that captured context, audience, tone and format. “All the content was mine — the ideas, the facts, the experience,” he said. “But the way it was written came from the AI.”

The result was an article produced in seconds, after careful prompting and refinement. “I still reviewed it many times before submitting it,” he said. “But the quality was impressive. It wrote in a journalistic style that I couldn’t replicate quickly.”

The article went on to win an internal AI innovation award. “It didn’t change treasury operations directly,” Schmahl noted. “But it helped us understand how powerful this technology can be when used properly.”

Treasury as an innovation leader

For Schmahl, the broader impact goes beyond efficiency or automation. It is about positioning the treasury as an innovation leader within the organisation.

He added that curiosity and openness are critical. “The average age in my team is quite young. They know this technology is here to stay. The box is open — it won’t close again.”

What surprised him most was how treasury, often seen as conservative, became an early adopter. “We were the first team to run RPA within the Delivery Hero group. We were among the first corporate teams to experiment seriously with AI. That’s not what people usually expect from treasury.”

For Delivery Hero, the message is clear: innovation does not require perfection, but it does require momentum. As Schmahl put it, “You don’t need to start with the biggest process. You just need to start.”