Let me start with something everyone knows.
You walk into an appliance store. You wander between the washing machines, come back a second time to the same model, look at the price tag, pull a face. And at some point a salesperson comes over. Not at the door, not an hour later – exactly when you looked up from the price.
That person wasn’t reading a report. They were looking at you. And without any system at all, they knew things no system would have told them: that you’re hesitating, that you’re comparing, that you’re about to walk out empty-handed – or buy, if someone clears up one doubt.
That’s all of advisory selling in a single move. And it’s exactly what we lost when we moved our stores online.
A shop floor with nobody working it
E-commerce delivered everything it promised. Reach. Open around the clock. A catalogue with no floor-space limit. Customers from the whole country instead of one city. The store grew a thousand times over.
And it went blind.
Because in an online store nobody is watching the customer. They come in, walk the aisles, come back a third time to the same washing machine, get annoyed, leave – and on the other side there’s nobody who noticed any of it. There’s a server writing events to a database.
The funny part is that these companies have a customer service team. Often a sizeable one. Nobody ever gave those people anything to look at.
Instead of eyesight we got three prosthetics, and every one of them is worse than looking:
The report. Analytics tells you what happened – yesterday. It’s CCTV footage watched the next day. All of it true, and there’s nothing you can do with it, because those people are already gone. A report answers “how many were there?” A salesperson has a different question: “who do I talk to right now?”
Static “recommended products”. That’s rearranging the shelves. It works on a crowd, it doesn’t know the specific person standing in the aisle, and it doesn’t change when she changes her mind.
The chatbot. This one is the worst, because it pretends to solve the actual problem. A chatbot isn’t the salesperson on the floor – it’s the guy with flyers hired to stand in the doorway. It doesn’t see the customer, it doesn’t know who’s close to deciding, it opens with the same line to everyone after thirty seconds. Which is why everyone has learned to close it on reflex.
That’s the problem. Now for where our answer came from – because it didn’t come from a slide deck about AI.
How we ended up here
For more than twenty years we’ve been writing software for companies that sell in the real world. FMCG manufacturers and distributors, wholesalers, field service companies.
Step back and look at it, and all our products do one thing. They give the person in the field information they wouldn’t otherwise have.
AMPER MSF is the app for the field sales rep walking into a store with a tablet. It holds that customer’s history, their prices, their outstanding balance, their last orders. It works offline, because out at a wholesaler on the edge of town, signal is an opinion, not a fact. The human sells – the app just makes sure they walk into the conversation prepared.
AMPER B2B is the portal where a trade customer places orders themselves, at two in the morning, with their own pricing and their own stock levels. The rep stops retyping orders out of emails and starts doing the job they were actually hired for.
AMPER Flow does the same for field service: standing in front of the unit, the technician knows what equipment it is, when it was last serviced, what has failed on it before, and whether it’s still under warranty. Again – the tool doesn’t fix the air conditioning. A human does, one who finally has the full picture.
AMPER B2C is our e-commerce platform, developed as open source, because we think a store should belong to whoever runs it.
And this is the point where something started to grate.
Because in B2B, with the field rep and the service technician, for twenty years we’d been doing exactly one thing: giving a human the context to have a better conversation with another human. And in B2C – where the customers are most numerous – our own store was every bit as blind as anyone else’s.
That’s the friction Live Assisted Sales came out of.
So what is it, exactly
Live Assisted Sales shows your team the shop floor in real time.
Not yesterday’s report, and not a crowd. People, right now, in the store. And next to each one, what the salesperson standing between the shelves would see:
- who’s here – including the anonymous, logged-out ones, which is most of your traffic;
- who’s close to deciding – a purchase intent score from 0 to 100, computed live from behaviour;
- why – the top three reasons in plain language: “added to cart”, “fourth visit to the same product”, “high cart value”;
- what mood they’re in – from happy to frustrated, the way you’d read someone’s face;
- what to do now – one concrete prompt, not a chart. Reach out. Hold them. Offer a product.
That third point – the “why” – cost us the most work, and I think it’s the most important one. Because the worst thing you can hand a salesperson is a number with no explanation. Nobody trusts a black box, and a salesperson who has once been burned by an algorithm will never look at it again. The salesperson on the floor can justify their hunch. Our system had to be able to do the same.
The argument about the send button
Now the thing we argued about longest.
The AI assistant in the chat window prepares a draft reply – tailored to that customer, their cart and the flow of the conversation, grounded in the real catalogue. And that’s where it stops. A human always hits send.
You can build it the other way. Technically there’s nothing stopping you from giving the AI a send button and having an autonomous salesperson. It’s just that sooner or later that salesperson will invent a price, promise a delivery date you can’t hit, or write something nobody on your team would ever say. And the customer buying eight thousand zloty worth of furniture will remember precisely that.
So, no. Someone buying anything pricier than a bar of chocolate wants to talk to a human. They just want that human to already know what they’re about to ask.
Incidentally: what for us was a product decision turned out to sit rather comfortably with the AI Act. But the order was the other way round – first we decided it was simply better this way.
How it fits with the rest
Live Assisted Sales isn’t a separate island. It’s the missing piece of the same arrangement.
The field rep got context in AMPER MSF. The service technician got it in AMPER Flow. The trade customer serves themselves in AMPER B2B. Now the consultant handling a retail customer gets the same thing: information they wouldn’t otherwise have, at the moment they need it.
The starting point was very concrete, actually. The AMPER platform has had Live Shopping for a long time – a live conversation with a salesperson inside the online store. It worked well, but it had the same flaw as every chat on earth: the consultant found out the customer existed the moment they typed their first message. Which is usually a few minutes too late – and in the case of everyone who never typed anything, never at all. Live Assisted Sales is what happens when you give that chat eyesight.
Technically it holds together too. The reference integration is our own AMPER B2C – it’s what everything was built on, and it sets the event standard. The integrations for WooCommerce, Shopify, PrestaShop and Magento reproduce it one to one, so a store on any of them connects in a few minutes, with no developer and no copying API keys around. And if someone runs their store on something else entirely – that’s exactly the work we’ve been doing under “custom projects” for two decades.
There’s one more reason we built this rather than someone bigger from across the ocean. Solutions from the US have to be stitched together from three or four tools today, and even then they have no European data residency. We have consent, retention and the right to erasure built into the product, and the data stays in Europe. For a store in the EU that stops being a formality and starts being a commercial argument.
One last thing
When you give an online store its eyesight back, something happens that we didn’t expect: it sees more than a physical store ever could.
A salesperson on the floor can handle three customers at once – here one person sees all of them and knows who to approach first. Salespeople forget – the store remembers that this person was here two days ago and left a full cart. A store manager will never find out whether moving that shelf helped – and we hold back ten percent of customers as a control group, with no assistance, and after a month we hand you a number instead of a hunch.
That last one is our favourite sentence in this whole story, as it happens. Not “we believe it works”. Just: this much more.
Phew – that was a long one. Thank you for making it to the end, dear Reader.
*Want to see your own shop floor live? We’ll show you Live Assisted Sales on data close to your own store – twenty minutes, no strings attached. Contact us.
