Atlas
AI employees / Atlas · Inventory & operations

Atlas

“SKU-4412 is down to eight units against a reorder point of fifteen. At this rate you're out on Thursday.”

3×
counter queries served per hour
2.2s
median stock query
12,840
SKUs tracked live
0
stockouts last month
The process he owns

Anyone who needs a stock figure should be able to trust the one they are given.

He owns everything between stock moving and somebody acting on a number. He works from the movements themselves rather than a periodic count, and he is measured on the variances he raises as much as on the answers he gives. Adjusting stock stays human. What changes is that the figure on the screen and the figure on the shelf stop drifting apart.

The artifact

The reorder review, exactly as Atlas filed it.

Every line carries the check that produced it, and anything held stays visible with its reason attached. Structure and logic from a live deployment; values changed.

ConsoleTue 24 Jun 15:38
Atlas — reorder review · aisle b · mx-central
AtlasInventory & opsBuyer review
Reorder review · aisle B · MX-Central94SKUs3held for a person11:00filed at
SKU-3478Oil filter, heavy duty · 42 on hand · reorder point 15Clear
SKU-4412Brake pad set, rear · 8 on hand · reorder point 15Reorder
SKU-4488Coolant hose 18 mm · 6 on hand · reorder point 10Reorder
SKU-4501Wiper blade 22" · 74 counted, system says 77Variance
90 othersWithin rangeClear
Counter note · WhatsApp voice · 11:06

«Oye Atlas, ¿tenemos filtros 3478? El cliente está esperando.»

Atlas’s decision

He answers in the language the voice note came in: 42 units, aisle B-14, MX-Central, in 2.2 seconds. He does not correct the 4501 variance himself — a three-unit gap goes to the warehouse lead for a physical recount.

review B-0f7 · closed 12:15 by J. Beltrán, Warehouse Lead2 reorders raised as drafts · 1 variance sent for recountFull audit trail →

What Atlas does.

Every capability trained on your SKU catalogue, your locations and your real lead times — not a default template.

01

Tracks stock across every location

Receipts, transfers and sales watched as they happen, so the number is current rather than as of last Friday.

02

Finds the part from how people describe it

Partial codes, descriptions, local nicknames or a voice note — nobody has to memorise a SKU.

“oil filters 3478” → SKU-3478 · 42 pcs · aisle B-14 · MX-Central
03

Warns you before the shelf is empty

Flags an item against its reorder point with the lead time factored in, not once it has already run out.

▲ SKU-4412 · 8 pcs · reorder point 15 — suggested order 120 (6-day lead time)
04

Keeps the ERP honest

Aligns ERP records with actual warehouse movement and surfaces the variance instead of overwriting it.

05

Answers hands-free

Counter and floor staff ask by voice on WhatsApp while they are carrying something.

06

Recalculates reorder points

From real consumption and the lead times your buyer knows are optimistic, per location.

Sound like what your warehouse needs?
How he gets specialized

He doesn’t arrive knowing your operation. He is taught it, on your records, before he touches anything live.

Four stages. You see his calls next to your team’s before anything is delegated.

Weeks 1–2

Diagnostic

Someone from our forward-deployment team works your counter and stock queue by hand for two weeks — the real queries, the real variances, the parts that never scan.

Weeks 3–4

Your catalogue, ingested

SKUs, aliases, locations, and the lead times your buyer knows are optimistic. Permissions scoped per system in the same pass.

Weeks 5–8

Shadow mode

Six months of movements, adjusting nothing. You compare his reorder calls against what your buyer actually ordered, and the differences tune the model.

Month 3

Live

He answers, flags and drafts. Releasing a purchase order stays with your buyer.

First usable results in weeks · full operation ≈ 3 months
Reach

No new platform. He works inside what your team already opens every morning.

The floor asks by voice because their hands are full. Nobody types a SKU into a dashboard while a customer waits at the counter.

Systems

Connected on the systems you already run

SAPOdooCustom ERPWMSPOS systemsBarcode / QR+ any system with an API
How he is asked

Any channel the floor already uses

WhatsApp textWhatsApp voiceBarcode scannerWeb dashboard+ any channel you already use
In production in

Where Atlas is already deployed

RetailWholesaleManufacturingAuto partsAgriculture+ more
Languages
Spanish · English
Same rules, either language, in one run
Regions
Mexico · US · LatAm
Data stays in the region you nominate
Availability
24 / 7 / 365
Work picked up as it lands, not in a batch
Time to launch
≈ 3 months
First usable results inside the first weeks
Scope, escalation and error

What he can reach, what he can’t, and what happens when he is wrong.

Every permission is set by you, per system, and visible on the trust page. Atlas reads every movement and adjusts nothing on his own authority.

Permissions, as deployed
SystemReadWrite
ERP · inventoryyesadjustments queued
WMS · locationsyesno
POS · sales feedyesno
Purchase ordersyesdrafts only
Payment systemsnono
Release a purchase ordern/anever
When the numbers disagree
01 · Hold

A variance is flagged with the movement history behind it, never silently reconciled to whichever number looks more likely.

02 · Escalate to a named person

Within 60 seconds to the warehouse lead for that site, with the last movements on that SKU attached.

03 · When he’s wrong

He’s wrong on about 1% of queries — usually a mislabelled location or a part with two names. The correction writes back into the catalogue so it answers correctly next time.

04 · The line he doesn’t cross

Atlas cannot adjust stock on his own, release a purchase order or write off shrinkage. Those stay with your warehouse lead and your buyer.

Results

The workload, redistributed.

What a warehouse team was carrying alone across three sites, and what it looks like after Atlas has been through shadow mode and gone live.

Before
Your team, alone.
Manual counting hours / week55
Error / miss rate18%
Process coverage40%
With Atlas
The same team, on the movements that matter.
Manual counting hours / week8
Error / miss rate1%
Process coverage100%
One wholesale distributor, three warehouses, 12,840 SKUs · six months either side of go-live · the monthly two-day stock stoppage ended.
3×
Retail

Counter service, three times the throughput

From six minutes to two per inventory query, which tripled counter capacity and stopped customers walking out mid-queue.

Auto parts chain · MX · NDA
2 days → 0
Wholesale

Stock audits without stopping the warehouse

A two-day manual count every month was replaced by continuous accuracy, with variances raised as they appear.

Wholesale distributor · MX · NDA
3 sites
Manufacturing

One number across three warehouses

Unified visibility and automatic transfer tracking, so a part missing in one plant is found in another before it is reordered.

Manufacturing plant · MX · NDA

Put Atlas on your warehouse floor.

Two weeks of diagnostic on your real stock data, with our forward-deployment team on site.

Talk to us on WhatsApp
No commitment · 2-week diagnostic · clear proposal