Confirmed
Scanned, weighed or reported by a machine, with the time and the source.
Facts you can trust. Intelligence you control.
WarehouseAI builds its picture of your warehouse from scans, deliveries and machine reports, and it tells you how sure it is. It then uses those facts and your procedures to decide what happens next, but only within the authority you give it, and it records why.
Decision record DR-4471 (illustrative)Acted
Loading lane D2 is blocked. Route R42 leaves in 30 minutes and its packages are waiting at D2.
Allowed toAct within limits
RecordedFacts, options, rule and result
Left to the machinesHow they physically move
Five kinds of fact, never blended.
“Item P123 scanned on shelf B08 at 10:02” tells you where it was last seen. If something suggests it was moved without a scan, WarehouseAI flags its location as uncertain rather than guessing.
Likely
Worked out from other information, such as an item someone probably moved without scanning. Never shown as confirmed.
Scheduled
Work that has been assigned but not yet done. A scheduled move is not a finished one.
Conflicting
Two reports disagree. Both are kept until a rule or a person settles it.
Unknown
Shown as unknown, because knowing what you don’t know prevents bad decisions.
A good-looking screen must never show a move as done before it really is.
A complete history you can replay.
Every change is kept, so you can see what happened, when, and why the record says what it does. That matters in a client dispute, an audit or an investigation.
- Corrections are added next to the original record. History is never overwritten.
- The order in which messages arrive doesn’t override what actually happened on the floor.
- Before acting on a decision, WarehouseAI checks the facts again in case something changed.
- Replay any moment to see what was known then and why a location or status changed.
- AI never invents stock or changes the record of what physically happened.
Your rules first. AI only where the rules run out.
Everyday decisions follow fixed rules and proven calculations, so they are fast, cheap and predictable. Unusual situations, competing priorities or uncertainty go to AI or to a person. Robots never have to wait on AI to keep moving.
Check the factsWhat is true right now, and which procedures apply
UnderstandWork out what kind of situation this is
Weigh the optionsRule out anything that breaks a hard limit
DecideAct within its authority, or hold and ask
ConfirmHand out the work and check the result
You decide how much freedom each decision gets.
For example, a warehouse can let WarehouseAI re-slot stock on its own while still requiring approval for disposals, stock write-offs or a major reorganisation. New equipment or a new AI model never gets more authority on its own.
Watch onlyWarehouseAI watches and keeps a record, but makes no suggestions.
SuggestIt recommends an action and shows why, and a person decides.
Ask firstIt prepares the action, and a named role approves it before anything moves.
Act within limitsIt acts on its own within the limits you set and reports what it did.
Your procedures, followed every time.
Standard operating procedures, client-specific rules and approved exceptions live in one library. Each has an owner, a scope and the dates it applies, so the right procedure is used for the right client, location and time.
- The procedures applied are the ones for this client, location, task and date.
- Rules suggested from past activity stay suggestions until someone approves them.
- A one-off exception isn’t quietly turned into a permanent rule.
- Every decision records which version of a procedure it followed.
If AI goes down, the warehouse keeps running.
Routine work carries on under your rules. Unusual, important decisions are held or passed to a person with an explanation. An outage is never treated as an approval.
- You are not locked into one AI provider. It can be changed without disrupting the operation.
- AI running costs are tracked for each process, with spending limits.
- A cheaper AI model is used only if it passes the same quality checks.
- Your stock records and your equipment’s safety systems stay in charge, whatever AI model is in use.
Try changes before the floor sees them.
Test a new layout, a busier peak or a machine breakdown against a copy of your real warehouse. The test never touches real stock or real equipment.
- Compare scenarios such as a robot out of service, a bigger order wave, a new slotting plan or a blocked aisle.
- See the assumptions, the expected output and delays, and how much the result depends on uncertain inputs.
- Adopting a tested plan re-checks today’s facts first and turns it into normal, reviewed work.
- New equipment is rehearsed in simulation before it is switched on in your building.
Bring us the warehouse you have.
Design partners get the first versions built and tested on their own floor, products and order history. You start with guided manual work and automate from there.