# Five rungs, and you stay in charge on every one | BYOM blog

URL: https://byom.co/blog/the-ladder  
Markdown: https://byom.co/blog/the-ladder.md  
Last updated: 2026-10-02

> How an ecommerce brand grows in stages, with ONS, Baymard, Deloitte and the EU AI Act on each rung, from using AI you already have to jobs that repeat with checks.

By Kina (Checked by the BYOM team). Published 2026-09-22. 9 minute read. Series: Growth.

## Key takeaways

- The ONS puts UK business AI use at around 35% in June 2026, and says most of it is shallow: about 1.6 technologies per adopter.
- Baymard's 50 study average puts cart abandonment at 70.22%, and the top reasons are problems a team can fix one at a time.
- A short checklist used at fixed moments cut major surgical complications from 11% to 7% in the WHO study, and the same logic suits repeat jobs in a shop.

Every growing brand reaches the point where the work multiplies faster than the team. Product pages need rewriting, the inbox fills, prices need checking, and the hours that should go into the next range go into admin.

A ladder is a useful way to think about the climb, because each rung gives you more reach without asking you to hand over the business. This post takes the five rungs one at a time and puts a piece of outside evidence on each, so you can judge where your shop stands for yourself.

The platform gives a sense of the scale of the audience. Shopify's own company page says businesses on Shopify have made more than $1 trillion in sales, that more than 675 million unique shoppers bought from Shopify stores in 2023, and that about 10% of US ecommerce sales are processed through it. Those are Shopify's own figures, in dollars and mostly American, but they explain why shopping assistants are being built to read Shopify stores.

- **28.8%** of UK retail sales were made online in August 2026, up from 28.4% in July (ONS, retail sales, August 2026)
- **35%** of UK businesses with 10 or more employees used at least one AI technology in June 2026, up from around 12% in late 2023 (ONS, AI in UK businesses, July 2026)
- **675 million+** unique shoppers bought from Shopify stores in 2023 (Shopify, company page)

## Rung one: connect the AI you already use

Most founders already have an AI assistant open for part of the day, and the numbers say they are not unusual. The ONS reported in July 2026 that about 35% of UK businesses with 10 or more employees use at least one AI technology, nearly three times the level of late 2023. Large language models were the most widely used category, at 18% of those businesses.

The same release adds a caution. The ONS says AI adoption to date has been relatively shallow: the average number of technologies per adopting business rose only from around 1.4 to around 1.6 since late 2023. Having an assistant is the common case. Having it connected to the system where the work happens is the rarer one, and it is the step that turns an answer in a chat window into a change on a store.

That step carries risk, because a connected assistant can reach store data. The sensible way to take it is to give the assistant read access first, see what it gets right, and only then let it propose changes.

A practical test for this rung is to list the three things you asked an assistant to do last week and note where the result went. If it went into a document that someone then pasted into Shopify by hand, the assistant is outside the work. The cost is not the typing. It is that the pasted version is the one nobody reviews against the original, so mistakes in copying surface only when a customer notices.

## Rung two: get seen by AI shoppers

Shoppers are starting to ask AI assistants what to buy. Deloitte's 2025 Holiday Retail Survey of 4,000 US consumers found that 33% planned to use generative AI in their shopping journey, more than double the 2024 figure, and 26% said they trusted it more than they did six months earlier. The uses they named were finding deals, summarising reviews and building shopping lists. These are US figures, so read them as a direction and not as a UK measurement.

A shopper who asks for the best deal is asking software to compare stores on facts. The facts have to be on the page. Baymard Institute's 2026 research on product pages found that 44% of sites do not display or link to a return policy there, though 60% of users expect to find it, and that 67% leave out a shipping cost estimate. A store that states its returns window and delivery cost in plain sentences is easier for a person to trust and easier for software to quote.

Share of online sales is the reason this rung is worth the effort. In August 2026 the ONS put the online share of UK retail sales at 28.8%, with online spending 8.9% higher than a year earlier. The channel is large and still growing, and the way a shopper reaches it is changing.

Baymard's product page review adds detail you can act on. It found that 42% of users try to judge a product's size from images alone, while 37% of sites provide no image that shows the product in scale. It found that 57% of sites use dropdown menus for size selection rather than buttons that show every option. None of that requires new technology. It requires someone to go through the catalogue and state the dimensions in words.

## Rung three: hand real work to an operator

The third rung is about handing over work that has a clear right answer and a clear owner. Baymard's average cart abandonment rate, drawn from 50 studies, is 70.22%. Some of that is browsing: 42% of US shoppers in its survey said they abandoned a cart because they were just browsing. The rest has causes a team can name.

- 40% of shoppers who abandoned a cart cited extra costs that were too high, such as shipping, tax and fees.
- 20% said delivery was too slow.
- 18% said the site required them to create an account.
- 17% said checkout was too long or complicated.
- 13% said the returns policy was unsatisfactory, and 12% could not see the total cost upfront.

Each line is a repeatable job. Show the delivery cost on the product page. Check the shipping times against what the courier is doing. Test guest checkout. Read the returns policy as a customer would. Baymard estimates that roughly 35.26% of conversion rate gains are available from checkout design improvements alone. A merchant with two people has more of these jobs than hours, which is the case for handing the reading, drafting and checking to an operator and keeping the decisions.

Consider what a week of handed over work looks like in practice. On Monday the operator reads last week's product pages and lists those without a stated delivery cost. On Tuesday it drafts the text for the ten with the most traffic. On Wednesday you read the drafts and decide. On Thursday the approved changes go onto the store. On Friday it checks the pages again. You spent perhaps an hour, on judgement, and the reading, drafting and checking took none of your time.

## Rung four: approve every change

Reach without control is how a small mistake becomes a large one. Article 14 of the EU AI Act, which deals with high risk AI systems, is a useful statement of what real oversight means, even though an online shop's assistant is not automatically in that category and the Act is not UK law. It requires that such systems can be effectively overseen by natural persons while in use. The questions below are worth asking of any tool, human or software, that can change your store: who can approve a change, whether that person can be the one who proposed it, and whether the record of what changed survives.

The article spells out what the person must be able to do. They must properly understand the system's capacities and limitations and monitor how it runs. They must stay aware of the tendency to rely too much on its output. They must be able to interpret that output, and to decide in any particular situation not to use the system or to disregard, override or reverse its output. They must be able to stop it with a stop button or similar procedure.

The phrase about automatic over reliance is the one a busy merchant should take seriously. If every proposed change looks the same and takes one click to accept, approval turns into habit and stops being a check. A review step works when it shows you the specific thing that will change. Shopify made the same point in its bulk inventory editor in December 2025: the editor now shows original and new values side by side, with increases in green and decreases in red, so that you can see you are replacing quantities and not adjusting them, and it warns you if the inventory changed while you were editing.

> decide, in any particular situation, not to use the high risk AI system or to otherwise disregard, override or reverse the output. EU AI Act, Article 14

Older control practice points the same way. Separation of duties is described as having more than one person required to complete a task, with the classic example of two signatures on a cheque. In accounting systems, companies routinely stop the same person from both receiving payments and approving write offs. Applied to a shop, the person or system that proposes a price change should not be the one that approves it.

## Rung five: put the work on rails

The top rung is repetition. A job that runs the same way every week, with a check at each stage, beats a heroic one off effort because the checks catch the same classes of slip every time. The best known evidence is from surgery, and it should be read with that in mind: a hospital theatre is not a Shopify store.

In the WHO surgical safety checklist study, run from October 2007 to September 2008 in eight hospitals in Seattle, Toronto, London, Auckland, Amman, New Delhi, Manila and Ifakara, 7,688 patients were followed, 3,733 before the checklist and 3,955 after. Major complications fell from 11% to 7% and inpatient deaths after major operations fell from 1.5% to 0.8%. The checklist took minutes and was used at three set moments.

- **11% to 7%** major complications before and after the checklist (Harvard Gazette on the WHO study, 2009)
- **1.5% to 0.8%** inpatient deaths after major operations (Harvard Gazette on the WHO study, 2009)
- **7,688** patients in eight hospitals (Harvard Gazette on the WHO study, 2009)

A later review of 20 studies adds the caveat that matters for a shop. It says the checklist should be understood not merely as a list of items to be checked off, but as an instrument for improving communication, teamwork and safety culture, and that results depend on leadership and structured training. A checklist ticked without attention does little.

For a store, the same logic gives you a short list of jobs that repeat, each with a check at the start, one before anything touches the live store, and one at the end. Weekly price checks, new product setup and policy review all suit it.

Start with one job and run it for a month before adding a second. Pick the one where a mistake would cost you most or embarrass you most, because that is where a check pays back soonest. Write the three checks on one page. After each run, add to the list anything that went wrong that the checks did not catch. A checklist that never changes is not being used.

| Rung | Evidence in the merchant's world | What to check |
| --- | --- | --- |
| Connect | ONS: 35% of businesses use AI, mostly shallowly | Is the assistant connected to where the work happens? |
| Get seen | Deloitte: 33% plan to use AI to shop; Baymard: 44% hide returns | Are returns and delivery stated in text? |
| Hand over work | Baymard: 70.22% abandonment, extra costs top reason | Which repeat job costs you the most hours? |
| Approve changes | EU AI Act Article 14: override and stop | Can one person approve their own change? |
| Rails | WHO checklist: complications 11% to 7% | Which weekly job has no check? |

## Where to start

The BYOM Shopify app runs a readiness scan on your installed store and shows how AI shoppers see your products, why a product is invisible to them and what to fix. Each fix arrives as a proposal, and the store changes only when you confirm it.

A Rail is a job Kina does for you on repeat. You approve before anything touches the store, each Rail has its own checks, and every run ends with a record of what happened.

Related: [What BYOM is](https://byom.co/blog/what-byom-is).

## Sources

- [ONS, retail sales, Great Britain, August 2026](https://www.ons.gov.uk/businessindustryandtrade/retailindustry/bulletins/retailsales/latest)
- [ONS, artificial intelligence in UK businesses, 2023 to 2026, published July 2026](https://www.ons.gov.uk/businessindustryandtrade/business/businessservices/articles/artificialintelligenceinukbusinesses/2023to2026)
- [Shopify, company page with merchant statistics, 2026](https://www.shopify.com/about)
- [Deloitte, 2025 holiday retail survey](https://www.deloitte.com/us/en/insights/industry/retail-distribution/holiday-retail-sales-consumer-survey.html)
- [Baymard Institute, cart abandonment rate statistics, 2026](https://baymard.com/lists/cart-abandonment-rate)
- [Baymard Institute, current state of ecommerce product page UX, 2026](https://baymard.com/blog/current-state-ecommerce-product-page-ux)
- [EU AI Act, Article 14, human oversight](https://artificialintelligenceact.eu/article/14/)
- [Wikipedia, separation of duties](https://en.wikipedia.org/wiki/Separation_of_duties)
- [Shopify changelog, prevent accidental inventory overwrite in bulk editor, December 2025](https://changelog.shopify.com/posts/prevent-accidental-inventory-overwrite-in-bulk-editor)
- [Harvard Gazette, surgical safety checklist drops deaths and complications by more than one third, 2009](https://news.harvard.edu/gazette/story/2009/01/surgical-safety-checklist-drops-deaths-and-complications-by-more-than-one-third/)
- [PubMed Central, review of studies of the WHO surgical safety checklist](https://pmc.ncbi.nlm.nih.gov/articles/PMC3489074/)