An agent that runs procurement. Not another dashboard.
The product is live and the first pilot conversations are under way. This page holds the market, what actually works today, and the plan with its three cases.
The problem, in numbers
The German Mittelstand does not have an insight problem. It has an execution problem.
of SMEs still run day-to-day operations mostly manually or semi-automated, with paper, media breaks and disconnected spreadsheet silos.
Bitkom, 2026of German SMEs used an ERP at all in 2025. The EU average is 45 %, and adoption has stagnated.
IfM Bonn, EU comparison 2025companies uses digital systems to optimise purchasing. The most-used system in purchasing is the ERP, with spreadsheets immediately behind.
BME / Onventis, Einkaufsbarometer Mittelstand 2024The bottleneck is not missing software. It is software that displays instead of executing.
The market
The layer we play on is the AI share of supply chain software. The two credible houses disagree about its size, so this page states the range rather than a false precision.
Europe is not a side market in this, and Germany carries the European share. That is also where the stagnating ERP base from the section above sits. That gap is what Synapsio builds into.
What already runs
Synapsio is not a presentation. The product runs at app.synapsio.solutions, holding stock, suppliers, orders and the log in one system.
The agent reads supplier mail and order confirmations, computes cover and reorder point per item, writes real orders by mail, and puts every decision in the log with its reason. Per item it picks a named method, from reorder point through Croston to newsvendor, and explains the choice.
A chat window cannot weigh Croston against newsvendor. That needs the stock, the history and the execution in one system.
A method per item
The agent picks a named replenishment method for each item: reorder point (s,Q), Croston for intermittent demand, newsvendor for perishables, periodic review or multi-echelon. It records the reason and its confidence alongside, and a person can pin any choice.
Scored afterwards
Every ordering decision is stored with the signals behind it and later scored against what actually happened. The system knows which of its own decisions were good ones.
Supplier mail both ways
Order out, reply in: incoming mail is read and the order status follows from it. The supplier signs up for nothing and installs nothing.
Several locations
Stock per site, own warehouse and third-party logistics, transfers with approval, and replenishment that knows where the stock sits.
Bills of material
Consumption is derived from real production runs instead of estimated, including the change history on the bill of material.
Approval and log
A value threshold, approval straight from the mail, and every action the agent takes recorded with its time and its reason.
Tenant separation
Enforced in the database, on every table. A tenant can be deleted completely and verifiably.
A cost ceiling per customer
The agent's compute is metered and capped per company. That is also the mechanism billing sits on.
How Synapsio earns
Two layers. The first carries today, the second arrives with the marketplace.
Monthly, tiered by business volume
Four tiers from entry to enterprise. Each has a monthly limit, and the limit is soft: going past it means a notice and a small overage, not being switched off mid-operation. For software that places orders, a hard stop is not defensible behaviour.
A share of marketplace purchases, later
When a company buys through Synapsio a fee applies, falling as volume grows and reduced for selected industries. This layer belongs to the marketplace and is not built yet. Payment processor fees are separate and are not our revenue.
Prices are set closer to launch. What already stands is the mechanism underneath: usage and budget are metered and enforced per company. Billing is therefore not a second system, only the same foundation with a price attached.
The plan
Three cases for 2030, rather than one number carrying everything.
- Floor €9M
- DACH only, subscription only. No transaction layer, no EU expansion.
- Base €42M
- DACH in depth, EU from 2028, transaction layer live. This is the case we underwrite. Breakeven 2030.
- Ambition €115M
- Full EU, marketplace effect, and settlement between systems.
The early ramp stays deliberately measured: around twelve paying customers by the end of 2026. Synapsio is sold by a sales team, not by self-registration, and the fast ramp curves from the AI world come from products that roll themselves out. We raise the ambition in year five, not the velocity in year one.
Peak cumulative funding need before breakeven: around €3.5M.
Funding
Product, first pilots, proof in live operation.
Sales in DACH, transaction layer.
EU expansion.
Deck and conversation
The pitch deck is open as a PDF. For the detailed numbers, the financial model and the state of the pilots, a direct conversation works better.