Ecommerce delivery platforms and AI integration | an interview with foodora experts

Experts from ecommerce delivery platform foodora share how they see a successful approach to AI integration.

The latest AI developments have been proving how vulnerable traditional tech structures are. What is foodora’s approach to AI implementation to ensure stability?

Bharath Krishnamachari, Senior Director of Data & Analytics at foodpanda, foodora, Yemeksepeti: One of the biggest lessons of the current AI wave is that traditional technology structures were not designed for the speed, scale, and decentralisation it introduces. Many organisations still rely on centralised governance models and approval processes that just cannot keep up with AI’s quick evolution.

At foodora, we believe stability comes from building the right foundations and governance while enabling innovation at scale, as opposed to slowing AI adoption. Initially, we had a centralised approach, focusing our engineering resources on high-impact AI initiatives such as AI-powered analytics and operational intelligence. While these use cases generated significant value, we also learned that a small central team can quickly become a hold-up when AI demand accelerates across the organisation.

We now approach AI implementation by combining centralised standards and distributed execution. We provide the governance, security controls, and platform foundations centrally, while enabling our teams to build and automate smaller, domain-specific workflows themselves. This way, we can maintain consistency and control where it matters most, while avoiding the rigidity that makes traditional technology structures vulnerable to begin with.

As a delivery platform, what are the biggest challenges you’re facing in this climate, and how are you using AI in your innovation and business strategy to address them?

Bharath Krishnamachari, Senior Director of Data & Analytics at foodpanda, foodora, Yemeksepeti: One of the most important shifts we are watching is how AI changes the relationship between consumers and digital platforms. As AI assistants become more capable and integrated into everyday devices, consumers might stop interacting directly with apps as they currently do. Instead, AI agents could increasingly act on their behalf, be that for reordering groceries, selecting restaurants, or making routine purchasing decisions. 

Platform businesses face a potential disintermediation challenge. If an AI agent becomes the primary interface, differentiation by user experience (UX) alone is harder, and operational excellence grows in importance.

For this reason, we focus our AI strategy on customer-facing innovation but also on fundamentally improving ecosystem operations. We are investing in AI-powered analytics, operational intelligence, automation, and workflow optimisation to help our teams decide faster and serve customers, merchants, and riders more effectively. Additionally, we are empowering our teams to build smaller AI-driven solutions that remove friction from their day-to-day work.

Ultimately, our goal is to use AI to make the entire platform faster, smarter, and more efficient. Consumer expectations continue to evolve, so successful companies will be able to translate AI into measurable operational value throughout their ecosystem.

AI has numerous use cases, and ecommerce, especially, has been taken by storm. Can you walk us through how you’ve implemented it across payments?

Vaibhav Kanth, Director of Product Management, Payments and Fintech at foodora, foodpanda, Yemeksepeti: With our platform that powers foodora, AI acts as the cognitive foundation of our payment ecosystem, in addition to its backend optimisation role. We have long invested in upgrading our orchestration layer from rules to machine learning (ML) and now to an AI-driven layer fully powering the transaction lifecycle. This framework uses predictive intelligence to dynamically route payments and to adjust retry strategies in real time, adapting to issuer performance fluctuations and complex regional variables to achieve higher authorisation benchmarks. 

Beyond transaction routing, we are deploying real-time inference channels to optimise the checkout experience. By analysing contextual signals and historical behavioural patterns, we can display the most relevant payment methods and personalised offers. When a transaction fails, an intelligent recovery engine engages immediately. It diagnoses why the transaction failed and guides the user toward the best alternative before they abandon cart, helping reduce our drop-off and order churn. This shifts our approach to the payment layer from reactive damage control to proactive retention.

How are you leveraging AI in fraud prevention, and which data is most important for logistics and customer experience?

Vaibhav Kanth, Director of Product Management, Payments and Fintech at foodora, foodpanda, Yemeksepeti: Fraud prevention at foodora moved from rule-based flagging to real-time ML scoring operating across the full order lifecycle, from signup, discovery, and checkout, to post-delivery. The most important signals aren't obvious: behavioural sequencing, device context, and basket-level anomalies outperform velocity checks alone, particularly in markets with high cash-on-delivery usage, where fraud vectors differ structurally.

The bigger potential was in combining fraud models with logistics intelligence. Real-time rider telemetry and kitchen preparation signals allow us to detect and disrupt fraud when it occurs, rather than after the fact. At scale, AI enables us to observe rider behaviour patterns across thousands of concurrent orders (something no manual process could match), helping us to serve customers with reliable post-order fulfilment, while systematically closing the gaps that bad actors exploit.

We approach this by tracking successfully-prevented fraud against declined good orders. A model optimised purely for fraud reduction will quietly destroy conversion and alienate your best customers. We invest most of our model governance effort in getting that ratio right across multiple markets with very different user profiles, payment behaviours, and fraud vectors.

What is needed regarding AI ownership and governance for the platform industry to keep evolving safely?

Nan Hao Maguire, Chief Information and Security Officer at foodora, foodpanda, Yemeksepeti: In my view, AI ownership and governance need to evolve beyond a single team or department’s responsibility. As AI becomes embedded across platforms, product, engineering, data science, security, and legal, business leaders shape how these systems get designed, deployed, and managed day-to-day. The real challenge I see is making sure that this distributed ownership doesn't turn into fragmented accountability, where everyone touches AI, but no one feels fully responsible for it.

I think organisations need a common governance framework: clear responsibilities, standardised controls, and shared visibility across the whole AI lifecycle. Critically, that framework must be anchored to business objectives, not treated as a separate compliance exercise. Governance that's disconnected from what the business is trying to achieve tends to either get ignored or become a blocker. When tied directly to business outcomes, teams understand why the guardrails exist, and governance is seen as an enabler rather than a constraint. I also don't think governance should rely on manual oversight or central approval boards for every decision. It works best if built directly into the platforms, workflows, and deployment processes the teams already use, invisible until needed.

For platform businesses like foodora, this means creating clear accountability for data usage, model behaviour, security, compliance, and business outcomes, while still enabling teams to innovate quickly and in ways that advance the business. Organisations scaling AI safely can combine decentralised innovation and centralised standards, all anchored to a shared set of business priorities. Done well, it turns governance from a constraint into a business enabler, bringing the confidence, transparency, and resilience needed to accelerate AI responsibly, drive innovation, improve efficiency, and build a real competitive advantage.

About foodora

foodora is a delivery platform, operating in 5 countries in Europe – Austria, Czechia, Hungary, Norway and Sweden. foodora’s mission is to deliver an amazing, fast, affordable experience connecting customers with businesses and riders, giving everyone more time to pursue what matters most to them. foodora delivers a variety of products including groceries, household products and restaurant meals within 60 minutes or less. foodora is part of Delivery Hero, the world’s leading local delivery platform.