Stop reacting.
Logistics has always been a reactive sport: something runs short, and someone scrambles. Osmiron learns from everything your factory has already done, sees what it is about to do, and gets material moving before the line ever has to wait.
- 01 Normal
- 02 Forming
- 03 Moving first
- 04 Prevented
Moving first. Spare blanks go from the store to Weld before anyone has to ask.
01 / The whole factory
It doesn’t need to understand your machines. It understands your factory.
If you have sat through a factory software pilot, you know how it usually goes: months spent modelling every spindle and conveyor, and a model that is out of date the week it goes live. Osmiron works one level up.
When a machine finishes a job, Osmiron doesn’t care how. It cares that something is now available, somewhere, at about some time, and that something else is going to need it.
That is enough to reason about any factory, whether it makes gearboxes or yoghurt. Because it reasons about the whole flow rather than one machine, it sees the knock-on effects that no single machine’s data can show.
- Work
- Material
- People
- Places
- Time
- Dependencies
- Everything Osmiron needs to know about a factory
02 / What it learns
It learns what nobody wrote down.
Every factory already knows what is about to go wrong. It is in how long a pick takes on a Friday afternoon, in the aisle everyone quietly avoids, in the delivery that always turns up a little late. Most of that gets logged, charted and filed, and used to explain yesterday. Osmiron uses it to change what happens next.
Nobody tells Osmiron that nights take the long way round, that one cell slows as the hall warms up, or who is quickest on the jobs nobody else wants. It picks those things up from the signals, ties them to what the factory was doing at the time, and keeps them.
Every worker, machine, aisle and shift gets a character, and that character is allowed to change. So when Osmiron looks ahead, it expects what your factory actually does, not what the plan says it should.
Machine · Press 3
Cycle time drifts after a few hours running.
Aisle · C4
Slows every pass between 14:00 and 16:00.
Worker · Picker 12
Fastest on mixed-bin picks. Still getting faster.
Shift · Nights
Pulls more from Line 2 than the plan assumes.
Illustrative. None of these are programmed in; Osmiron finds them in your factory’s own signals.
03 / Learning by doing
Every decision is also a lesson.
Osmiron doesn’t wait for a finished model before it is useful, and it never considers the model finished. It decides in real time, under real uncertainty, then watches what actually happens and adjusts.
Better expectations make better decisions. Better decisions make better data. It is the idea behind reinforcement learning, pointed at a factory floor instead of a game board. The longer it runs, the less it has to guess.
- 01
See
Signals from people, machines and the systems you already run.
- 02
Expect
What the next few hours look like, not just the next second.
- 03
Decide
What should happen next, for the factory as a whole, not one station.
- 04
Learn
What actually happened, folded back in before the next decision.
- ↺ and again, all day, every day
On the floor
One clear answer to one question: what should I do next? Every worker gets a plan built around what they do best, and everything else stays out of the way.
In the office
The factory as it actually behaves: where time goes, where material piles up, and why. Down to sentences like “this corner of the layout makes every route worse.”
04 / Rehearsal
Break it here, not there.
Once Osmiron knows how your factory behaves, you can ask it the questions you would never risk finding out for real. Take workers away, delay a delivery, close an aisle, double an order, and watch where it gives first.
Every answer starts from your factory as it is today, with everything it has learned about it. Find the breaking point while it is still a simulation, and know whether the extra forklift, the new hire or the layout change will pay off before you commit to it.
- ›What if demand jumps next month?
- ›What if Line 3 goes down on a Tuesday morning?
- ›What if two people are off during the biggest order of the year?
- ›What if this cell moves next to the dock?
- ›Would another forklift actually help, or just add traffic?
05 / The first weeks
Honest from the first day. Better every week after.
Getting started doesn’t take a year-long rollout. Once Osmiron knows the layout of your floor and starts receiving signals, it is working, and it gets better from there.
From the first day you get a clear overview of your factory and the best plans Osmiron can make with what it has seen so far. It never pretends to know more than it does.
- 01
Map
We walk your floor with you and pin the places that matter to known GPS coordinates. It is a walk around the factory, not a modelling project.
- 02
Connect
We plug in the data your factory already has. Osmiron doesn’t need much, and the worker app fills in the rest.
- 03
Learn
Osmiron starts learning from the first signal. The first week is the slowest, because it is testing the waters and getting to know your factory.
- 04
Improve
The gains build as new situations come up and Osmiron sees how your factory handles them.
06 / Pricing
We only get paid out of what you save.
Osmiron is priced so that no factory ends a month behind. Pay a fixed rate set by the size of your factory, or pay nothing up front and give us a share of what Osmiron actually saves you.
Every time Osmiron stops something from happening, it adds the cost it avoided to a running tally. At the end of the month we take our share of that tally. If it saved you nothing, you owe nothing.
Every line of the tally is in the console: what Osmiron saw, what it did, and what it would have cost. If you don’t agree with one, reject it and it comes off the bill.
The terms are shaped with each factory: a locked share, or a larger one with a ceiling so you always know the worst case.
Fixed
A set monthly rate, based on the size of your factory. The same bill every month, whatever happens on the floor.
Share of savings
A share of what Osmiron saved you that month, agreed with you up front. Nothing saved, nothing owed.
- 3 MarLine 2 supplied before it ran dry€6,400
- 11 MarLate delivery kept off Line 1€2,150
- 19 MarPress 3 drift caught before the night shift€3,800
- 24 Mar · rejected by the factoryPicks rerouted around aisle C4€900
- Saved
- €12,350
- Osmiron’s share
- €2,470
- You keep
- €9,880
07 / Questions
Fair questions, straight answers.
- Do our workers need new devices?
- No. The Osmiron app runs on any phone or tablet. It shows every worker their plan as the floor changes, shows how the shift is going, and doubles as the sensors Osmiron learns from. If you already track locations, Osmiron can connect to that instead.
- Where does our data live?
- Wherever you want it. Osmiron runs on your own site or in our cloud, in your region: EU factories stay in EU data centres, US factories in US ones. Each factory’s data is locked to that factory, and it is only ever used to make Osmiron better at running that factory, never for anyone else.
- Does it replace our current systems?
- No. Osmiron works alongside the systems you already run, and we are adding connections into the ones factories use most. You can also start with Osmiron purely as a place to rehearse: test your plans against how your floor actually behaves, and see what it would do for your turnaround before anything changes for real.
- Does it run the factory on its own?
- No, and it isn’t meant to. Osmiron is there to help the people who run the factory see further and decide better. It proposes, and the people in charge always have the final say.
- What happens when Osmiron is wrong?
- It will be sometimes, especially early on. Every miss, and every time someone overrides a plan, becomes something it learns from. It only ever tests limits inside the safety bounds your factory sets.
08 / Why we exist
The oldest problem we never solved.
For as long as there have been people, we have been trying to solve logistics, long before anyone had a word for it. It meant gathering enough food to last through the winter, getting the harvest to market before it spoiled, and making sure the right tools were in the right hands at the right time. The industrial revolution didn’t solve the problem. It made it bigger, faster and far less forgiving.
Every generation has got better at it, and every generation has run into the same wall. The warning signs were almost always there, in how the work was actually moving, but no one could see all of it at once. So the problem was only ever understood afterwards, once the harvest had spoiled or the line had stopped.
The people who run factories today are some of the most capable operators there have ever been. They know things no dataset will ever hold. They hear when a press sounds wrong long before any sensor notices. They know who needs to leave early because their kid is sick. They can tell when a plan that looks right on paper is going to fail on the floor. That judgement is what keeps the world turning.
What they have never had is a way to see ahead. Osmiron learns the things nobody ever wrote down, and some nobody knew were there, down to the smallest detail: who hits their stride after lunch and who slows down, or which small delay at the dock always turns into a big one on the line. Nobody has to tell it any of this. It notices, and it puts every one of those details to work, giving each person and machine the job they are best at in that moment. Put thousands of them together, and Osmiron can see where the day is heading hours before it gets there, and what could be done about it while there is still time.
The decisions stay with the people running the floor. For the first time, they get to see tomorrow before it happens.
We are a small team in Stockholm building that system. If you run a factory, tell us where your floor stops, and we will be straight with you about whether Osmiron could have seen it coming. If you are an investor, or an engineer who likes hard problems with physical consequences, we would like to hear from you too.