Q&A: Arber Sejdiji, Co-founder & CEO, Zenline AI

Zenline AI builds agentic assortment intelligence for retailers. It delivers margin-improving, product-level recommendations on what to list, delist, reprice, and where private-label opportunities exist by continuously analysing the full product range, internal data, the competitive landscape, trends, and shopper behaviour. Zenline is industry-agnostic, serving sectors from cosmetics to DIY.

Ex-BCG Arber Sejdiji is the CEO and Co-Founder of Zenline AI, a retail startup developing AI agents for retailers.

These agents are already deployed at some of Europe’s largest retailers, supporting margin improvement through agentic assortment and pricing recommendations.

Can you tell us a bit about your background?

I studied at ETH Zurich and conducted AI research in Cambridge. After my studies, I then worked in strategic consulting at Boston Consulting Group. Here I kept seeing the same problem: even the largest global retailers were leaving significant margin on the table because assortment decisions were being made manually, on incomplete data, and too slowly. Category managers would often focus on just 5% of their portfolio, where the data was clean and the potential obvious, while the remaining 95% went largely unmanaged.

Zenline was founded to solve exactly that problem. And solving this can unlock 2-4% more margin.

What does your company do? / What is your USP?

Zenline builds AI agents for Assortment Intelligence. That’s something completely new, as agents were technically not capable of doing that six months ago. What we are doing is enabled by the major advances in AI over the past year. Our agents combine margin data, sell-through figures, stock levels, competitor assortments and shopper signals or trend analysis to produce concrete, product-level recommendations: what to delist, how to reprice, where the substitution gaps are, and which trends are gaining share with competitors.

The core USP is simple: BI tools deliver data, our agents deliver ready-to-act-on recommendations. An analysis that would take a category manager several weeks can be produced in a few minutes, and our agents do it continuously across the entire range, not just the 5% of SKUs where the data happens to be clean. Category managers can finally focus on evaluating recommendations and executing, instead of spending most of their time preparing and interpreting data.

What’s special about the platform and your approach?

The shift from analytics to agentic AI is what distinguishes us. Most retail technology stops at visualisation: it shows you what the data looks like and leaves the interpretation and decision to the category manager. Our agents go further, taking over actual workflows of a category manager. They identify optimisation potential against the retailer’s own strategy and KPIs, analyse substitution options, price sensitivity and variant overlaps, and output specific recommendations that can be simulated, react and reported on directly.

For example: reduce the price of product A by 0.50 dollars and sales increase; replace product B with product C for a better margin at the same shopper demand; trending product D is gaining at your biggest competitor, launch it. Those are not insights for an analyst to work with. Those are decisions for a buyer to evaluate.

What advantage does it add?

Three things. First, speed: decisions that previously took weeks are produced in minutes, and the system runs continuously so the analysis is always current. Second, coverage: the agents work across the entire assortment, not just the SKUs where internal data happens to be clean and reliable. Third, measurable commercial impact: in pilot projects across beauty, food, DIY and furniture, we consistently see 3 to 5 percent EBIT uplift through better decisions on assortment, pricing and variant management. The combination of speed, full coverage and quantified outcome is what makes the difference in practice.

How does a product implementation actually look like and how do you measure success?

We deliberately built for a fast start, if a retailer reaches out, we are ready to kick things off the next week. Our agents can generate first recommendations using only external data, competitor websites and market signals, before internal data is uploaded. There are also no months-long ERP integration projects — a customer sends us a file with all their data, and agents load it into the tool, mapping everything to the right place automatically. We typically begin with a four-week pilot focused on a defined category or set of categories. Ultimately, what matters is whether the recommendations lead to better buying decisions, and we track the financial outcomes of acted-on recommendations as a performance indicator.

How are retailers using your systems to gain competitive advantage and what does best practice look like? Can you share a case study with us?

A good example is a facial skincare category we analysed for a European retailer. More than half of the products in the category had no suggested substitute that could be shown to shoppers. When any of those products went out of stock or was discontinued, the customer simply left. Our agents mapped the full range by ingredients, skin type and pack size and identified the substitution gaps. The recommendations covered which private label products could fill those gaps most effectively, and which branded SKUs were strong candidates for delisting based on low repeat purchase and poor margin contribution. The result was a cleaner, more navigable category with better private label placement and a measurable improvement in both customer satisfaction and margin. This kind of analysis, when done manually, takes weeks. The agent produced it in minutes and the team could act on it immediately.

What challenges are retailers facing in 2026?

Three things stand out:

First, speed: assortment decisions that used to be made every six months now need to happen weekly. Shopper behaviour shifts overnight, trends arrive via social media before they register in sales data, and competitors react faster than ever. Retailers who cannot keep up with that cadence lose ground structurally.

Second, the courage to reduce complexity: studies show that 40 to 60 percent of SKUs are unproductive, yet they stay “on the shelf”. Retailers who do not actively manage their range down create overloaded assortments that confuse shoppers, waste logistics capacity and erode margin.

Third, AI-ready leadership: the technology exists, the pilots prove the value, but adoption requires management willing to work with new tools and to generate trust in AI due to this.

Retailers have a 12 to 18 month window to catch up. After that, the competitive AI gap becomes too large to close.

What is on the horizon for you as a company?

We’re aiming to lead retailers into the agentic economy, which means becoming the leading platform for AI-driven assortment decisions.

………………………………………….

To find out how Zenline AI can help your retail operation, visit them online here or connect with them here .