Retail
The store work nobody has time for
We build the systems that do that work: returns, stock, prices, marketplaces, the weekly pack.
From the dock door inward (receiving, put-away and picking) is a different page: Warehousing →
A returns bay of unbooked stock, and a file that still thinks it sold.
Returns supervisorReplenishment that doesn't fire, and counts done twice because the first disagreed.
Store managerOversell penalties, returns that never reach sellable stock, and deductions finance cannot reconcile.
Marketplace leadA weekly pack assembled by hand from POS, ERP and marketplace exports.
Head of trading
What changes first
Returns move again
Items are routed and back in sellable stock the same day.
Shelves and records agree
The lines that quietly stopped selling raise one count instead of a stocktake.
Monday is a decision, not a rebuild
The pack assembles itself and names the few things that need someone to choose.
Where the work leaks, and what we build
Returns
The returns bay
A returns desk on your own systems reads the reason, routes the item, prepares the refund, and sends the odd ones to a person.
A person releases every refund above your threshold, and every abuse flag.
Stock
The shelf says empty, the system says otherwise
The system watches sales against book stock store by store, asks for one count instead of a stocktake, and drafts the order.
The store manager accepts the count task; a person releases the order.
Marketplace
Several marketplace panels, one stock number
One console over your marketplace channels reconciles stock, warns before a dispatch deadline slips, matches every deduction to your ledger, and drafts the dispute.
A person files every dispute and approves every push to a marketplace.
Reporting
The Monday pack rebuilt by hand
We build the pack as an agentic system on your data: it assembles itself, explains the variance, and flags what needs a decision.
The person who owned the pack reviews it, then spot-checks.
A week with the system
- Monday
- The pack is built. It asks for counts on the stores that stopped selling.
- Tuesday
- A competitor drops a price. You see the margin and you decide.
- Wednesday
- Weekend returns are routed; the ones that look like abuse wait for a person.
- Thursday
- Peak. Order-status calls are answered; the damaged-delivery call goes to a person.
- Friday
- Finance opens the deduction report. Every line is matched, and the disputes are drafted.
Illustrative example. Nothing here is a measured result.
What we would watch with you
| What we watch | How it is measured | Baseline |
|---|---|---|
| Return-to-sellable cycle time | Returns-bay scan to sellable stock, read from the ledger. | Most chains cannot measure it today. |
| On-shelf availability on the top lines | Gap rate on the top-selling lines, weekly, per store. | Needs a store-executable target, since most causes are store-level. |
| Marketplace cancellation and late-dispatch rate | Platform-reported seller metrics, per marketplace, monthly. | Each platform defines it differently, so state it per platform. |
| Price-decision latency | From an observed market change to a recorded human decision. | The regulatory frame requires a human decision point. |
| Hours of manual reporting removed | Named people and hours per week, measured before and after. | The easiest baseline to establish, and the one owners feel. |
Questions retailers ask
Will it change prices on its own?
No. A person approves every price change. The system watches the market, shows what moved and what it costs you, proposes an action, and records what you decided. That record is the point.
Does our data have to leave our building?
No. We can run the whole system on hardware you own, inside your network, including anything touching store cameras or customer records. Where a cloud service genuinely helps, we say so before you decide.
Do you leave once it is live?
No. We stay. We monitor it in production, fix what breaks, change what the stores tell us is wrong, and re-version it until it hits the targets we set together.
Retail
Tell us what Monday looks like.
Send us one process and we will tell you honestly whether an agentic system helps.
Talk to us