Agentrolux

Dette eksempel er ved at blive opdateret. Teksten nedenfor er stadig på engelsk.

Example · Retail & e-commerceStock and reorder agent

Never run out of your best sellers, or tie up cash in slow stock.

A stock and reorder agent for small shops. Every morning it reads sales, forecasts the next three weeks, and drafts the purchase orders. You approve anything above your limit.

See how it works Talk about your shop
Built for 10–500 employeesWorks with your shop platformSimulated demo below

Monday brief · 07:02

Simulated example

  • Kiln & Co · 96 mugs · $720

    Within this week's limit

    Draft

  • Paper Union · 72 notebooks · $245

    Within this week's limit

    Draft

  • Northwood · 60 oak boards · $2,040

    Above your $1,500 order limit

    Your yes

2 drafted · 1 waiting · 2 products under a week of cover

Nothing is sent to a supplier

01 — The problem

Three small failures that cost real money.

None of them look dramatic on a Tuesday. Together they decide whether the month is profitable.

01

The best seller runs out

The shelf is empty on a Saturday. The customer buys the alternative, or leaves and orders online. You find out on Monday.

See what this looks like at scale
02

Cash sleeps in the stockroom

Slow stock ties up money that could pay for what actually sells. The discount you need later is a second loss.

See the overstock cases
03

One person remembers to reorder

The spreadsheet lives in someone's head. When they are on holiday, nothing gets ordered — and nobody can check why a number was chosen.

See the checkpoints
02 — How the agent works

Six steps, every morning. Same order, every time.

Select any step to see what happens, and an example with real-looking numbers. The human checkpoints are outlined.

Stage 01 of 06

Read sales and stock

The agent reads yesterday's sales from the shop, the website and any marketplaces, then checks stock on hand and orders already on their way.

Example

18 mugs on the shelf, 5 sold yesterday, 12 more due on Friday.

Select any stage for the plain-language explanation. Motion is off when your device asks for less.

Text version of this flow

Data comes in from sales (shop, website, marketplaces), stock on hand, and the supplier catalogue. Six stages follow:

  1. Read sales and stock: The agent reads yesterday's sales from the shop, the website and any marketplaces, then checks stock on hand and orders already on their way.
  2. Forecast demand: For each product it forecasts the next 21 days. Weekends, planned promotions and the last two weeks of sales are taken into account, and the answer is a range.
  3. Calculate reorder: It subtracts stock on hand and stock already ordered, adds the supplier's lead time plus seven safety days, and rounds up to the supplier's minimum order quantity.
  4. Check spending limit (human checkpoint): Every draft is checked against your per-order limit, your weekly limit, your approved suppliers and your do-not-reorder list. Anything above a limit waits for you.
  5. Draft purchase order: It writes the draft in your system with the reason attached, ready for you to approve or change. The agent never sends an order to a supplier.
  6. Report to owner (human checkpoint): You get a short weekly summary: what was drafted and why, what needs you, what is at risk of running out, and what is overstocked.

Outputs: a draft purchase order, a weekly summary for the owner, and an audit log with every input and decision. The agent never sends an order to a supplier.

01

Read sales and stock

Yesterday's sales from every channel, stock on hand, and orders already on the way. If a day is missing, it says so instead of guessing.

02

Forecast demand

Twenty-one days ahead, as a low-to-high range. Weekday rhythm, planned promotions and the last two weeks are included.

03

Calculate reorder

Demand for the lead time plus seven safety days, minus stock on hand and on order. Rounded up to the supplier's minimum.

04

Check your limits

Per-order limit, weekly limit, approved suppliers, do-not-reorder list, and data quality. Anything unclear waits for you.

05

Draft purchase orders

One draft per supplier with the reasoning attached. Nothing is sent. Your supplier sees nothing until you approve.

06

Report to you

A short Monday summary: what was drafted, what needs you, what is nearly out, and what is overstocked.

03 — See it run

A made-up shop, 25 products, one Monday.

Step through the run, or press play. Every decision below comes from the same rules described on this page. This is a simulation, not a client result.

SR

Monday reorder run

25 products · 30 days · 3 channels

Simulated example

Step 01 of 06

It reads yesterday's numbers.

Sales from the shop till, the website and two marketplaces. Stock on hand, orders already on their way, and each supplier's lead time. No typing, no spreadsheet.

25

products

30

days of sales

3,094

units sold

4

orders on the way

Revenue last 30 days · $56,663 · 9 suppliers

Simulated example: a made-up shop with 25 products. Not a real client, and not a forecast of your results. Every decision above comes from the same rules described on this page.

04 — You stay in control

The agent drafts. You approve anything above your limit.

Automation is useful only if you can see it and stop it. These are the parts a person keeps, in this workflow.

Nothing is sent to a supplier without you. The agent can create a draft; it cannot place an order.

  • You set the limits

    A maximum for one order and a maximum for the week. The agent checks both before it drafts anything.

  • You approve the exceptions

    Any order over a limit, any new supplier, and any product on your do-not-reorder list.

  • You can pause it

    For one product or the whole shop. Drafts stay where they are, and nothing is lost.

  • You can read the reasoning

    Every decision records the inputs, the forecast range, and why it drafted, held or asked.

05 — Documented failures

What happens when reordering is not a system.

Public cases, with the number and the source. Each one also says what a stock agent would have done differently — not that it would have saved the company.

01 · 2013–2015

Target Canada: The shelves were empty when shoppers arrived

$5.4B

pre-tax loss on exit

Target opened 124 Canadian stores at once. Inventory planning could not keep up, shelves were bare early on, and shoppers who wanted to buy could not. Fortune's post-mortem names empty shelves and stockouts as a central cause. Target closed all 133 stores and cut about 17,600 jobs.

Target Corporation, 15 Jan 2015 Fortune, 15 Jan 2015

A stock agent would have flagged the fastest-selling lines in the first weeks and drafted replenishment orders before the shelves emptied — inside a spending limit the owner sets.

02 · 2022

Nike: Stock arrived all at once, and stayed

+44%

inventory, to $9.7B

Retailers ordered early during supply-chain delays. When shipping times improved, the stock landed together. Nike's inventory rose 44% to $9.7 billion, it said it would liquidate late seasonal product more aggressively, and quarterly net income fell 22%.

CNBC, 29 Sep 2022

The agent reads incoming orders and slowing sell-through together, so it would have flagged the build-up while the stock was still on its way — before the markdowns.

03 · 2022

Walmart: The wrong stock, marked down at scale

11–13%

expected full-year EPS decline

Customers shifted from general merchandise to food. Clothing stayed on shelves and in warehouses, needing more markdowns. Walmart cut its profit outlook, citing pricing actions to improve inventory levels. Its US inventory had been up about 33% year on year, and its US CEO said about a fifth was stock he wished he did not have.

Walmart Inc., 25 Jul 2022 CNBC, 25 Jul 2022

The agent would have seen apparel sell-through falling and stopped topping it up, freeing the cash for what customers were actually buying.

04 · 2023

Industry pattern

The pattern across retail: This is not a big-company problem

$1.77T

lost to stockouts and overstock worldwide

IHL Group estimates that inventory distortion — out-of-stocks plus overstocks — cost retailers about $1.77 trillion worldwide in 2023. It is an industry estimate, not one company's loss, and Sensormatic, which published the summary, sells retail technology.

IHL Group / Sensormatic, 2023

The same arithmetic applies to a shop with 25 products: a weekly run, a forecast range, and a draft order you approve. That is the whole workflow.

Each company case is a public report of what happened, not a claim that a stock agent would have prevented it. Causes were mixed, and the sources say so. The last card is an industry estimate, not a company case.

06 — Facts about retail

The bigger picture, with the source attached.

Every number here links to its publisher, shows the sample or method, and names who paid for the research. We do not quote a figure we have not opened.

01
Source

IHL Group / Sensormatic, Inventory Distortion

2023

Model of worldwide out-of-stocks plus overstocks: $1.77 trillion in 2023, down $172B from 2022.

Sponsored by Sensormatic, which sells retail technology. An estimate, not a census.

Open the source

$1.77T

Lost to stockouts and overstock worldwide in 2023

IHL Group's estimate of inventory distortion: empty shelves and excess stock together. An industry model, not one shop's loss.

IHL Group / Sensormatic · 2023
02
Source

NRF + Happy Returns, 2025 Retail Returns Landscape

2025

Industry projection: $849.9B of returns in 2025; 19.3% of online sales expected to be returned; 9% of returns fraudulent.

Industry-wide projection, not a small-shop measurement.

Open the source

19.3%

of online sales expected to be returned in 2025

Returns are part of stock planning, not an afterthought. Build them into the cash you hold back.

NRF + Happy Returns · 2025
03
Source

Toast, Voice of the Retail Industry

2026

Blind survey of 340 US convenience, bottle-shop and grocery operators with fewer than 16 locations. 88% believe AI will help them work more efficiently; 21% rank inventory management among their top three difficulties.

Toast sells retail software. These are beliefs, not measured efficiency gains.

Open the source

21%

of small US retailers rank inventory among their hardest jobs

In Toast's 2026 survey of 340 operators, inventory management rose six points year on year as a top-three difficulty.

Toast · 2026
04
Source

TechNet / Morning Consult, Small Business E-Commerce & AI Survey

2026

Online survey of 1,108 US small-business owners who sell online, 25 February–9 March 2026. 71% use AI tools regularly; 52% daily.

Industry-sponsored (TechNet's members include Amazon, eBay, Etsy and Shopify). Self-reported use, not measured results.

Open the source

71%

of small US online sellers use AI tools regularly

More than half use them daily. The tools are already in the business; the question is whether they are doing the boring, useful work.

TechNet / Morning Consult · 2026
05
Source

Faire, UK independent retailer survey

2025

Poll of 300 UK independent retailers by OnePoll. 83% had begun adopting AI; about one third use it daily.

Faire is a wholesale marketplace. UK-only, vendor-commissioned sample.

Open the source

83%

of UK independent retailers had started using AI

About a third used it every day, and users reported saving around 135 minutes a day. UK-only, and the sample was commissioned by a marketplace.

Faire · 2025
06
Source

MarketsandMarkets, AI in Retail Market

2026

Market forecast: USD 31.12B in 2024 to USD 164.74B by 2030, 32.0% annual growth. Covers personalisation, recommendations, assistants and supply-chain AI.

A forecast from a research firm, not a measured result. The category is much broader than this one workflow.

Open the source

$164.74B

Forecast size of AI in retail by 2030

A market forecast at 32% annual growth, and a broad category: recommendations, assistants and supply-chain tools all count.

MarketsandMarkets · 2026

These findings come from other businesses and other researchers. They do not promise what will happen in your shop. We measure your own four-week baseline before the agent drafts anything.

07 — Numbers that help you plan

Three numbers worth knowing before a call.

One comes from a published survey. The other two are yours: where no reliable average exists, we ask you instead of inventing one.

Sourced benchmark

19.3%

of online sales were expected to be returned in 2025.

Set aside room for returns when you plan stock. A reorder agent should read refunds as well as sales.

NRF + Happy Returns, 2025

Your number

hours

That is about 208 hours a year — roughly 26 working days.

Put it in the calculator

Your number

That is 12 counts a year. If each one takes half a day, that is 6 working days.

No published average; your count is the useful number

08 — Before and after

The same week, two different jobs.

Drag the handle to compare a manual week with the agent week. This is an illustration of the change, not a measurement of time saved.

BeforeAfter
Illustrative week

A manual week

  1. Mon 07:00Open the spreadsheet. Yesterday's till export is missing two days.
  2. Mon 09:30Count the best sellers by hand. Two are already lower than they should be.
  3. TueThe popular candle sells out. There is no time to order until Thursday.
  4. ThuEmergency order from a pricier supplier to cover the weekend.
  5. FriTwo hours matching invoices to what actually arrived.

With the stock agent

  1. Mon 07:00The agent has already read 30 days of sales, stock and orders on the way.
  2. Mon 07:02Three drafts are ready. One is over your limit and waits for your yes.
  3. Mon 07:05You approve two, change one quantity, and get on with the day.
  4. TueThe candle is still on the shelf. The order is already on its way.
  5. FriA ten-minute read of the weekly summary. No spreadsheet.

Drag the handle, or use the arrow keys. This is an illustration of how the week changes, not a measurement of saved time — the pilot measures your own baseline.

09 — Your version

Four questions about your own week.

Keep your answers. They are the brief for the first version we would build for you.

  1. How many products do you reorder in a normal month?

  2. What tells you a product is about to run out: sales, lead time, or a supplier date?

  3. What is your approval limit for a purchase order, and who approves above it?

  4. Which stock decisions would you never hand over: price, clearance, or a new supplier?

Next

Put your own figures into the calculator below. It runs in your browser, and nothing is sent or stored.

Put your week into the calculator

Bring the four answers to the free one-hour call. We will use them to sketch the first version, and you keep the written plan either way.

Talk about your business
10 — Plan the numbers

Move the sliders. See the shape of the saving.

This is a planning tool, not a quote. It shows how the arithmetic works and which inputs matter most. Every formula is printed below the result.

25

Count every size or colour as its own product.

5 h

Checking stock, counting, emailing suppliers.

$42

Total sales ÷ number of orders.

$38,000

All channels together, before returns.

8%

If you are unsure, start at 5% and change it later.

55%

What the stock itself costs you.

Planning estimate

Not a forecast

Hours back each month

13.0

about 156 hours a year

Sales recovered from fewer stockouts

$608–$1,216

from $3,040 at risk each month

Cash freed from slow stock

$941–$1,881

one-off release, once the stock sells down

Every output is an estimate that depends on your inputs and the assumptions below. Nothing is stored or sent.

How we calculated this
How we calculated this — every formula and assumption+

Hours saved

5 hours a week × 4.33 weeks × 0.6 = 12.99 hours a month.

We assume the agent removes 60% of the manual reordering time. That share is a planning assumption, not a measured benchmark.

Sales recovered

$38,000 × 8% = $3,040 at risk. We assume 20–40% of that is recoverable: $608–$1,216.

Recovery depends on why customers left and whether they come back. The 20–40% band is our planning assumption.

Cash freed

$38,000 × 55% = $20,900 cost of goods. × 1.5 months of stock × 20% slow-moving × 15–30% released = $941–$1,881.

One-off, not monthly. Assumes 1.5 months of stock, a fifth of it slow-moving, and a 15–30% reduction in that slow stock.

What this is not

Not a promised result, not a quote, and not financial advice. The pilot measures your own four-week baseline before anything changes, so the numbers you see afterwards are yours.

11 — Build this next

Other workflows we can build for a shop.

Start with the one that costs you the most. These use the same pattern: read, decide, draft, ask, report.

Returns and refund triage

Read the request, check the order and policy, approve small refunds, and send the customer a clear status.

Ask about this workflow

Product descriptions with brand rules

Draft copy that follows your tone and required details, for you to approve before it goes live.

Ask about this workflow

Order-status chat

Answer 'where is my order?' with the real tracking status, and hand exceptions to a person.

Ask about this workflow

Abandoned-cart follow-up

One helpful reminder, written in your voice, with a real reason to come back — never a discount spiral.

Ask about this workflow

Review replies for approval

Draft replies to Google and marketplace reviews, flag the ones that need care, and wait for your yes.

Ask about this workflow

Supplier invoice matching

Compare each invoice with the order and delivery, flag differences, and queue the rest for payment.

Ask about this workflow

Stock-count reconciliation

Turn a count sheet into a corrected stock file, with the differences explained before anything changes.

Ask about this workflow

Dedicated pages for these workflows are not built yet. Each card starts the conversation with the right context.

11 — Questions retailers ask

Straight answers, no sales fog.

If something is missing, ask on the call. We will tell you honestly whether this is a fit.

Book a free 1-hour call
Do I need to change my shop platform?

No. The agent reads from the systems you already use: Shopify, WooCommerce, marketplaces, and a till or accounting export. If a connection does not exist, we start with a daily file export and add a direct connection only if the volume justifies it.

Who sees my sales data?

Only the people you name. Data is held in your own accounts, encrypted in transit and at rest, and never used to train an AI model. The agent reads product-level sales and stock; it does not need customer names or addresses.

What if the forecast is wrong?

It will be wrong sometimes. That is why it shows a range and why you approve anything above your limit. The weekly summary compares what it forecast with what sold, so the error is visible and the settings can be adjusted.

Can I stop the agent?

Yes, at any time, for one product or the whole shop. Pausing leaves every draft exactly where it is. Nothing is sent to a supplier unless you approve it.

How long does setup take?

A first version usually takes about two weeks: connect the data, agree the limits, and run it in parallel with your normal process for a week before it drafts anything on its own.

What does it cost?

Every shop is different, so we do not publish a price. We scope it on a free 1-hour call and give you a written plan you keep, whether or not you go ahead.

Does it place orders without me?

No. It drafts purchase orders and asks for approval above your limit. Even an order within the limit stays a draft until you approve it in your own system.

I sell on marketplaces too. Does that work?

Yes. Sales from each channel are read into one stock picture, so the same unit is not counted twice. Where a marketplace holds its own stock, we treat it as a separate pool unless you tell us otherwise.

12 — Start here

Tell us about your shop. We'll show you which part to automate first.

One hour, no slides. Bring a rough idea of your best sellers and the stock that keeps sitting still. You leave with a written plan you keep.

Book a free 1-hour call Email us instead

hello@agentrolux.com

13 — Sources

Every number on this page

Checked October 2026. Vendor-sponsored research is marked as such in the source panel next to each figure. Sample sizes and limits are recorded in the page notes.

For technical teams

Integrations, data handling, guardrails and the full specification

Integration options

  • Shopify, WooCommerce and BigCommerce: read orders and stock; write draft purchase orders where the platform supports it.
  • Amazon, eBay and Etsy: read sales through their APIs or a consolidated daily feed.
  • Accounting and inventory: Xero, QuickBooks, Cin7 and Linnworks via CSV first, API later.
  • Approvals: email, Slack or Teams. The owner replies approve, hold, or change quantity.

Data handling

  • Credentials live in the server environment, never in the browser.
  • Product-level sales and stock only; no customer names or addresses are read.
  • Data is stored per shop, encrypted, and deleted 90 days after a run unless the owner asks to keep it.
  • No customer data is used to train a model.

Guardrail configuration

  • Per-order and weekly spend caps, enforced in code rather than in a prompt.
  • Supplier allow-list and a per-product do-not-reorder list.
  • Read-only sales, stock and open-order access; write access only to draft purchase orders.
  • Data-quality stops: missing days, negative stock, and any day above four times the median.

Audit log format

run_id, at, sku, action,
inputs { on_hand, on_order, lead_time_days,
         moq, unit_cost, forecast_low,
         forecast_high, weekly_spent },
decision { qty, cost, status },
reason, approval, outcome
Download the workflow specification (.md)

Includes the decision rules, defaults, failure handling, metrics and the 32-scenario evaluation set.

Figures checked October 2026. This page describes a simulated workflow, not a client result.

Sample limits are shown with every figure