NatWest's AI Adoption Report looked at 1,400 UK SMEs and mid-market businesses. 44% already use AI. Only 6% of those users have reached the stage where AI is changing how the business actually runs. Almost a third of users (29%) are still piloting or building awareness. Firms with more than 100 employees use AI at 67%. Smaller firms sit at 36%.
Closing that gap comes down to putting AI on jobs that already happen in the warehouse and the accounts system, and keeping a person in charge of anything that can move stock or money.
Shallow AI looks busy and changes little
Writing product copy, summarising a meeting, or asking a chatbot what a report means: that is useful. It does not change goods in, picking, despatch or the invoice that goes to Xero. Those are the jobs that eat the day in a small operation.
A transforming job is one you can measure
Pick one boring, repeated job with a clear owner today. Examples I would start with:
- Days of cover alerts. Compare recent sales to supplier lead time. Draft a PO for a person to approve. Do not auto-send it.
- Exception picks. Surface the orders that failed scan-verify or needed a substitution yesterday, so someone can fix the location or the kit build.
- Late supplier chase. Draft the follow-up from the PO and the promised date. A person sends it.
If you cannot say what done looks like in a number (POs raised on time, short picks down, invoices matching Xero), you are still in pilot.
Write the boundary before you connect it
Anything that can change stock, an order or an invoice needs an explicit allow-list. We published how we think about that here: Before your AI touches stock or invoices, write down what it can touch.
NatWest's customer research points the same way. 81% of people say being able to reach a real person when they need one is the most important trust factor, and around three quarters want to know when AI is being used.
Shopify's new batch fulfilment does not replace this
Shopify has just added batch fulfilment: group up to 250 fulfilments, print a pick list, buy labels, then scan to verify at packing. That is a solid ops upgrade for merchants who ship from Shopify. It still does not decide stock truth across a warehouse, a courier and Xero. That is the job of a proper Shopify WMS integration. AI that only lives in the chat box will not do it either.
Start smaller than you think
The report says firms at the transforming stage report much bigger time savings than early explorers. You do not get there by buying another dashboard. You get there by giving AI one real job, measuring it for a fortnight, then adding the next. That is how we build Bud.
If you want help choosing that first job for your setup, give me a shout.
Sources
- NatWest Group, AI Adoption Report press release, 25 Sep 2026: natwestgroup.com/news-and-insights/news-room/press-releases/2026/ai-growth-gap-emerging-across-uk-businesses-as-smaller-firms-ris.html
- Shopify Changelog, Fulfill orders in batches, 16 Oct 2026: changelog.shopify.com/posts/fulfill-orders-in-batches