LO
Get told when a verified record lands on this occupation →
Retail salesperson / shop assistant
The person on the shop floor who finds out what someone actually wants, knows what is in the back, and turns a browse into a purchase — which is a different job from taking the money.
This is not a probability of losing your job. It combines how much of the role's task load is exposed to automation with how far adoption has actually gone — useful for comparing occupations on one consistent basis, and for nothing else.
Covers selling on a shop floor — advice, fitting, demonstrating, upselling. Cashiering is a separate occupation on this site with a different exposure, and warehouse and stockroom work is separate again. The single biggest variable this page cannot generalise is whether the store's economics come from footfall or from being a collection point for online orders, because that decides what the job even is.
The evidence base holds verified records for other occupations, but not one for this one yet. Until it does, the analysis below is reasoning about task structure and known technical capability — for this job in particular it is not backed by traceable sources, and we would rather say so than cite things we have not verified. An empty section here is a gap in our coverage, not a finding about the work.
What is actually changing#
The unit of analysis is the task, not the job title. A role is not replaced — its task mix shifts.
Is this your job? Say so and this page narrows to your share of it.
A job title is a bundle of tasks bought together, and no two people hold the same bundle. Nothing is sent anywhere — it stays in this browser.
Working out what they actually want
Still human-led≈ Platform inferenceReading someone who says they are just looking, asking the two questions that reveal the real constraint, and recommending against the expensive option when it is wrong.
Online recommendation engines have had twenty years and enormous data on this exact problem, and the thing they still cannot do is the thing a good assistant does in thirty seconds: notice that the stated requirement is not the real one. That correction depends on watching a person handle the object, which is not an input any recommender receives.
This holding does not mean the floor keeps its people: the same conversation can be moved to a chat window staffed by fewer people covering more stores, and in that move the task survives while the post does not. What protects this task is proximity to the object, so the exposure rises the moment the store stops holding stock.
Knowing what is actually in the back
Automating≈ Platform inferenceAnswering whether it exists in that size, when the next delivery lands, and which other branch has one — accurately enough that someone will wait for it.
This is a database lookup that used to require a person because the database was unreliable and the person compensated. As inventory systems got accurate enough to expose to customers directly, the compensating knowledge stopped being scarce — and a customer with a phone can now answer it without asking anyone.
Losing this task does not free the assistant's time in a way the assistant benefits from, because it was the reason to approach them. It was the errand that started most conversations, and a floor where nobody needs to ask anything is a floor where selling has to start some other way — which nobody has designed.
Keeping the floor standing up
Still human-led≈ Platform inferenceReplenishing, folding, tidying, moving the display, and handling the customer who is upset about something that is not your fault.
Shelf-scanning robots have been deployed and withdrawn repeatedly by large chains, and the pattern in the withdrawals is consistent: they audit the shelf well and cannot fix it, so a person still walks the same aisle. Handling an angry customer is not a task anyone has attempted to automate in a physical store, because the whole point is that someone is accountable in person.
Safe from automation and safe from cost are different things. The clearest threat to these hours is not a robot, it is the algorithmic rota: shifts cut to the half hour against predicted footfall, which reduces the same labour without removing a single task. That change has already happened in large chains and shows up in nobody's automation index.
Taking the money and the return
Automating≈ Platform inferencePayment, refunds, exchanges, the exception that needs a manager's code.
Self-service payment is the longest-running automation in retail and the direction is not in question; what is in question is how far it goes, and the site holds a verified record of a chain removing self-checkout from almost all its stores. The mechanism is well understood: the saving is real and so is the shrinkage, and which dominates is a per-format calculation rather than a technology fact.
This task moving says little about this occupation, because in most shops it is not where the hours are — cashiering is a separate job here for exactly that reason. Read a self-checkout rollout as a fact about the cashier count, not about whether anyone is left to help you find a size.
Which technologies matter here#
Four separate signals. They are deliberately not added together — a job exposed to two technologies is not twice as exposed.
How it got here#
The index is not a static number. This is where it would have sat at each capability checkpoint since ChatGPT — reconstructed, and labelled as such.
A moderate climb driven by one task leaving rather than by anything arriving on the shop floor: knowing what is in the back used to require a person, because the inventory system was unreliable and the assistant compensated. As those systems became accurate enough to show customers directly, that compensating knowledge stopped being scarce — and it was the errand that started most conversations. The curve flattens from 2025 because what remains is advisory work anchored to physical stock, which no recommender receives as an input. Read the height as the loss of a reason to approach someone, not as a machine doing the selling — and note the change this curve cannot show at all, which is shifts cut to the half hour against a footfall forecast.
A flat line is not a forecast of safety. It says which tasks automation has reached so far — the occupations that moved least here are the ones where the constraint is physical or regulatory, and both of those can change.
Recent changes#
No verified events recorded yet.
This section will fill from the monitoring pipeline as events are collected, de-duplicated, graded and linked to the tasks above. An empty list here means we have not verified anything — it does not mean nothing is happening.
"We found no news" is not the same as "you are safe."
What this means for you#
This is still one of the easiest jobs in any economy to get, and that has not changed. What has changed is where it leads: the stock-knowledge rung that used to make someone indispensable in their second year has been flattened by a customer-facing inventory app. Aim at the categories where the product has to be tried, fitted or explained, because those are the ones where the floor keeps its people.
Your leverage is the conversation, and the risk is that it gets relocated rather than replaced — into a chat channel where one person covers eight stores. If your employer is building that channel, the question to ask is who staffs it, because that is the same job with a different postcode and usually a different pay band.
Your options#
Four directions, each with its real constraints and one thing you can test this week. Continuing as you are is a legitimate choice — it just has to be a chosen one.
Move to a category that must be tried on
Where the product has to be fitted, demonstrated or sized, the sale cannot happen without proximity, and proximity is what this occupation has.
Those categories usually pay on commission, so the income becomes less predictable even when it is higher.
Count how many of your sales this week happened only because the customer handled the item. That is the part of your job that is anchored to the floor.
Follow the stock, not the till
Stores that have become collection points for online orders need someone who owns accuracy between the app and the shelf, and that person is paid more than the floor.
It is a back-of-house job, so the parts of the work you may actually like — the customers — go away.
Pick ten items and compare the app's count to the shelf. The gap you find is the size of the problem nobody owns.
Common questions#
Not in the way the question implies, and the more useful split is between the two halves of the job. Answering what is in stock has already largely gone, because the customer can look it up — that was the errand that started most conversations. Working out what someone actually wants has not, because it depends on watching them handle the thing, which no recommender receives as an input. The realistic risk is not a robot on the floor; it is the conversation being relocated to a chat channel where one person covers several stores.
No date. The signal for this job is whether your store still holds stock a customer can touch, because everything durable about the role is anchored to that. Watch the floor space given to collection lockers versus display, and watch whether your shifts are being set to the half hour against a footfall forecast — the second is already reducing these hours in large chains without removing a single task, and it will reach you before any robot does.
Large chains have deployed and withdrawn them more than once, and the pattern in the withdrawals is the useful part: the robot audits the shelf well and cannot fix it, so a person still walks the same aisle with a trolley. What a scanning robot changes is who knows the shelf is wrong, not who corrects it — and correcting it is where the hours are.
No, and treating them as one is how people get the wrong answer about both. Cashiering is a transaction task with the longest-running automation in retail pointed straight at it. Selling on a floor is an advisory task anchored to physical stock. Many people do both in the same shift, which is exactly why the exposure of the shift is not the exposure of either task — and why this site keeps them on separate pages.
Method and sources#
- Assessment date
- 2026-09-14
- Basis of the task judgements
- 0 evidence-backed · 4 platform inference · 0 not enough evidence
- Verified events
- 0