The future of retail tech: Smarter AI, safer rollouts, and real impact

Staying ahead in e-commerce means constantly innovating your technology without compromising reliability. It's a delicate balancing act, and few understand this better than our own Luis Ribeiro, Vice President of Technology and Services at ZEOS. With over two decades of experience, Luis stands at the forefront of building scalable tech solutions for Europe's fashion logistics ecosystem.

In an interview with TechPros.io, he shares how ZEOS approaches this challenge while preparing for the future of retail technology.

The industry has largely embraced composability, and for good reason; it promises flexibility and agility. However, as Luis points out, "extreme composability creates significant integration challenges" that not every company is equipped to handle. Instead, he champions pragmatic composability.

This approach means clearly understanding and separating core business problems into distinct domains: think catalogue management, inventory, or order management. Sometimes this involves vendor technology, other times it means complementing with custom-built software. It’s not about making everything composable; it’s about strategically applying it where it adds tangible business value and flexibility, ensuring technology truly serves a purpose.

AI has been a cornerstone of retail for some years, predominantly in areas like personalised recommendations and customer segmentation. Yet, Luis highlights a significant new trend: "We’re seeing a new trend with technologies like vector databases bringing new capabilities and algorithms, reinforcing a revolution in personalisation."

This isn’t just an incremental improvement; it’s fundamentally changing the speed of innovation. Previously, training models could take days or even weeks. Now, updates are happening in hours, potentially even instantly. This acceleration leads to better customer experiences through real-time insights, directly driving improved conversion rates.

Beyond personalisation, sophisticated AI is democratising decision-making support across retail planning functions and supply chain management. Predictive analytics, trading insights, and "returns autopilots" are just a few examples where AI is optimising complex operations. The advent of Generative AI is set to further expand the toolkit and algorithms applied in business, promising even more profound transformations.

Black Friday, Christmas, and other peak shopping moments can make or break retail performance. At Zalando, we’ve designed our architecture to treat these surges as “business as usual.” Our systems are built on microservices that are as stateless as possible, allowing individual components to scale independently. Combined with auto-scaling across most domains, this gives us the elasticity to absorb sudden spikes seamlessly. But technology alone isn’t enough: ahead of peak periods, we double down on observability, scenario planning, and recurrent load testing to identify and eliminate bottlenecks before they impact performance. By blending auto-scaling flexibility with proactive capacity planning, we ensure that even the heaviest traffic days run smoothly — protecting the customer experience and giving brands confidence during the most critical moments of the retail calendar.

Luis' insights underscore a holistic, pragmatic approach to intelligent retail architecture. By focusing on strategic composability, harnessing the power of new AI technologies like vector databases, meticulously managing change, and making informed build-or-buy decisions, he and his team continue to build resilient, scalable, and innovative solutions that drive success in the dynamic world of e-commerce.

So what can you take away from this approach for your own tech strategy? We’ve broken it down into the following checklist.

Composability with purpose

• Don’t make everything composable — focus on where it delivers business value and flexibility.

• Extreme composability risks creating unnecessary integration challenges.

AI as a retail game-changer

• Already embedded in personalisation (recommendations, segmentation).

• New tools (e.g. vector databases) enable real-time insights and instant updates, boosting conversion.

• Sophisticated AI now supports retail planning, supply chain, returns management, and order allocation.

Modernisation approaches

• Strangler Fig pattern: Gradually replace legacy systems without disruption.

• Avoid big-bang replacements; use incremental rollouts instead.

Safe experimentation & rollouts

• Use feature flags to control activation of new capabilities.

• Apply canary deployments — test with a small user group before full rollout.

Balancing custom vs. packaged solutions

• Custom-build only when it creates competitive differentiation.

• Use packaged software (ERP, finance, HR, CRM) for commoditised areas to avoid reinventing the wheel.

Change management focus

• Biggest challenge isn’t technology but people and processes.

• Prioritise process resilience over chasing technical perfection.