Earlier this month I was back in Osnabrück for the third DSAG-Handelstage. Two days. Around 400 CIOs, domain leads and implementation partners from German-speaking retail. 32 exhibitors. One community that runs its business on SAP and wants to know what actually works.
That is why I value DSAG. It is not a showroom but a place to see AI running in real operations instead of on a roadmap slide. This year I had the chance to show exactly that on stage, together with Dmitry Proskurin, Head of Supply Chain Development at Parfümerie Douglas.
Thomas Henzler, DSAG board member for sales, production and logistics, opened the event with SAP's Autonomous Enterprise. In that model, people set the goals and guardrails, and AI agents prepare, coordinate and execute large parts of daily operations. Forward-looking control of inventory and goods flows was named explicitly as an opportunity for retail.
Then came the reality check. "Autonomous processes don't come from deploying AI agents alone," Henzler said (translated from German). They need clean data, consistent master data, end-to-end processes and integrated systems first. For retailers with grown, heavily customized SAP landscapes, that is a substantial transformation effort.
He is right. And his point raises a question the room kept circling back to: if autonomy depends on the data and planning behind it, what happens to the stock no forecast was ever built to steer?
On stage with DOUGLAS
DOUGLAS does not have a planning problem. It runs around 1,900 stores in 22 countries with its core systems on SAP, omnichannel warehouses that serve both stores and online customers, and a dedicated forecasting and replenishment tool. The DOUGLAS slides were precise about where that tool does its job: the permanent assortment.
That scope is the whole story. In beauty retail, the permanent assortment is only part of what sits in a store. Around a quarter of the DOUGLAS portfolio changes every year. Novelties, promotions, presentation stock and the long tail are driven by commercial decisions, not by forecasts. No planning system, however good, was built to steer them store by store once they land.
So DOUGLAS added a second layer downstream. One line on their slides said it plainly:
YDISTRI complements the state of the art upstream supply chain.
Upstream, stock flows from suppliers through the warehouses into stores. Downstream, YDISTRI moves it from store to store once the plan has done its job and local demand starts to diverge. That is second allocation.
Two real cases from our session showed what that looks like at SKU level.
One store had sold 10 units of a premium serum in twelve months, with 42 still on hand. At that pace, the stock would have lasted around 1,500 days. Another store sold 196 units of the same serum over the same period, and its stock there lasted about a month. We moved 23 units. On paper, this product was a fast mover. In the wrong store, it behaved exactly like dead stock.
The second case was starker. A fragrance had not sold a single unit in twelve months at one location, with 11 on the shelf. Elsewhere in the network, the same fragrance had sold 64 units and was down to its last 4. Ten units changed stores.
The product was never the problem. The location was.
The pattern held across everything we redistributed. Most of the stock came from store positions holding more than a year of inventory, many with no sales at all. The large majority went to positions that were already out of stock or would sell through within 90 days. For DOUGLAS, the result is a 12% reduction in working capital requirements at the same level of product availability, with the system live within three months. No new purchase orders were needed to get there. Every unit moved was stock DOUGLAS already owned.
Autonomy doesn’t have to wait
DSAG's closing summary described an industry in motion: retailers testing concrete AI scenarios, rethinking processes and building the foundations to scale what works. For most of the Autonomous Enterprise, that is the right sequence. Inventory does not have to wait for it.
Store-to-store decisions are small, frequent and measurable. Each move either sells or it doesn't, within weeks. Second allocation runs on the sales and stock data every retailer already has, next to the SAP landscape already in place, with no rip and replace. That makes inventory the natural place for AI to prove itself in daily operations before it takes on bigger decisions.
The gap Henzler described between vision and practice is real. In inventory, it is shorter than most roadmaps assume. And waiting has a price. Every product sitting in the wrong store is a markdown with a date on it, and every unit sold at full price is one that never turns into waste.
Thanks, and what's next
Thank you, Dmitry, for sharing the stage with me and giving the DSAG community an open look at DOUGLAS's experience, results and learnings. A customer telling the story in their own words is worth more than any slide we could build.
Thank you to Stefan Binkowski and the SAP team for the support and the organization, and to DSAG for a format where practitioners talk to practitioners.
And thank you to everyone who stopped by Stand 23 to talk with Mario Megela and me. Those conversations are only getting started.
If you want the full DOUGLAS story, we sat down with their supply chain team to talk it through in detail.