# Primark tests changes on a copy of its warehouse first. Small firms can do a version of that

> Primark will test warehouse changes on digital twins built from its own data before spending money. Here's the small-business version, and why AI needs to understand your business first.

Published: 2026-10-24
Author: Dan Cluderay

A model, or an AI assistant, is only useful if it understands how your business really works. Before you ask any AI to help you make a decision, give it something true to work from.

Primark announced a partnership this week with Kallikor, a supply chain modelling company. The idea is simple. Before Primark changes how a distribution centre runs, it tests the change on a digital copy first.

Layouts, staffing, how stock flows through the site, even changes across the wider network. Each option gets run in the model so the team can compare cost, capacity, service and risk before any money is committed. Primark's Director of Supply Chain, Adam Hoccom, said it lets them test different setups "before we commit capital spend".

The timing makes sense. Primark is getting ready to launch home delivery in Great Britain, and it has bought an automated fulfilment site in Sheffield to support it. That's a lot of change to get right first time.

## The detail that matters

The interesting line isn't about AI at all. Kallikor says the models are built from Primark's own operational data, so the results reflect how its sites actually run, "rather than relying on assumptions or averaged views of the network".

That's the bit most AI projects skip. A model, or an AI assistant, is only useful if it understands how your business really works. If it doesn't, it can't help you decide anything. It can only produce confident guesses.

## You don't need a digital twin to start

A small warehouse or online shop isn't going to commission a simulation of its building. But the principle scales right down. Before you ask any AI to help you make a decision, give it something true to work from.

**1. Write down how an order really flows.** From the customer clicking buy to the parcel leaving. Who touches it, which system it passes through, where it waits. Most businesses find at least one step that only exists in someone's head.

**2. Get the numbers to agree.** If Shopify, your stock system and Xero each tell a different story, any AI reading them will tell you a fourth. I've written about [getting Shopify, Xero and stock to agree](/insights/before-you-connect-ai-to-shopify-and-xero/) before you connect anything else.

**3. Agree what your words mean.** "In stock" can mean on the shelf, on the shelf and not allocated, or on its way from the supplier. "Profit" can mean three different things. People cope with that. AI doesn't, unless someone writes it down.

**4. Decide what it's allowed to change.** Testing ideas is safe. Changing live stock or invoices isn't. I've covered [writing down what your AI can touch](/insights/write-down-what-your-ai-can-touch/) separately.

## Then test before you spend

Once the business is written down properly, you can do a small version of what Primark is doing. Ask "what happens if we move the fast sellers nearer despatch?" or "what if we take on another courier?" and get an answer based on your own orders, not a blog post's averages.

That's what we're building [Bud](/bud/) to do at OnDraft. It's in beta, and it starts by learning how a business works: its language, its systems and where it's going. Other AI tools can then plug into that, rather than each one guessing from scratch.

Whether you use Bud or not, the order of work is the same. Understand the business first. Then let the AI help you decide.

If you want to talk through where your own systems disagree, [get in touch](/contact/).

## Sources

- Kallikor, Primark partners with Kallikor to strengthen supply chain decision capability, 22 Oct 2026: kallikor.ai/primark-partners-with-kallikor-to-strengthen-supply-chain-decision-capability/
- Retail Gazette, Primark taps AI technology to overhaul supply chain decision-making, 23 Oct 2026: retailgazette.co.uk/blog/2026/10/primark-taps-ai-technology-to-overhaul-supply-chain-decision-making/