I’m shopping for a bike right now. An electric one, to ride the streets of Paris.
I live on the Belleville hill, I cross the whole city to get to work. I’d spend more than an hour a day on my bike, and I like neither vibrating on cobblestones nor jolting over a pothole. I’m not exactly careful, I can afford very little maintenance. Some weekends I ride out to the forest. I have budget, but I won’t pay the price of a car for a bike. The chain on my previous electric mountain bike kept breaking. I don’t want to lock my wheels to the frame, or rub my hands against a chain. I’m scatterbrained, I want built-in lights.
The customer knows their problem long before they know their product criteria. No way I can express my problem in a search bar, no more than through filters. Yet my need translates:
- Electric hybrid bike, city and trails
- Automatic belt drive, low maintenance, nothing to break
- Mid-drive motor, more comfort, no rear-wheel theft and no grease on my hands
- Full-suspension frame ideal against potholes, decent front suspension acceptable
- Lights built into the frame
- Under 3,000 euros
In 2003 the salesman at a specialty shop would have pointed me in the right direction within minutes. In 2020 the in-store salesman went from expert to sales rep, and alone on the internet, I have to become an ebike expert by binge-watching Youtube. No expertise, blank-page anxiety in front of the search bar, incompetence in front of pedantic filters. I push my search back to later. It will take weeks, sometimes months.
In 2026, a shopping assistant is an online advisor, expert and trustworthy
A shopping assistant helps me find the right model or models for me.
It compares products by translating the jargon into words I understand: bye “7-cog drivetrain vs Enviolo belt”, hello “manual shifting vs automatic gears”.
It handles reassurance and answers my worries: “does this bike have enough range for a weekly charge?”, “how long does the insurance run?”
It finds me the best deal for my need: “you don’t need the rack version or the latest Bosch motor, and with the welcome code you land inside your budget”.
Usage still in transition, and complementary
Enough bikes, over to tech. Search bars still work very well. “Google Pixel 9” will show the right models in under a second.
After 30 years of learning to speak in keywords, most of your customers will keep using the magnifying glass to find what they’re looking for. After 30 years of spending hours browsing merchandising pages for inspiration, we still enjoy it. After 20 years of enduring dumb, pre-scripted, frozen chatbots, we have learned to distrust the floating bubbles at the bottom of the page.
Shopping assistants add latency to searches but deliver a confident, ultra-personalized experience. Nothing like the late chatbots.
They complete the optimized experience of our pages by serving customers left behind for too long, the ones who don’t know what they’re looking for, but who know what they want.
What do you gain?
Contrary to what the advertisers tell you, low adoption (5%), and incremental conversion nowhere near x3. Conversation conversion rates are very high, but that is also selection bias: the discussions will come from your most engaged customers.
Yet Amazon is doubling down on Alexa Shopping and spreading its AI across its whole experience.
Why? First because the selection bias is a blessing, the shopping assistant has become a privileged communication channel with your best customers.
Then because despite the selection bias, the analysis of our conversations shows that our AI strongly helps the user with their purchase in 75% of cases.
Also because, like it or not, 20% of customers now run their research on an AI before buying. And a conversation on your store is the guarantee that the recommendations will not come from your competitor’s catalog.
Finally because these conversations uncover the needs, priorities, doubts and criteria of your customers that no click-analytics tool could ever offer. A gold mine of information for your product development, your marketing teams, and your personalization efforts.
What is our approach?
We started building our shopping assistant the day GPT4 shipped. We are e-commerce product and marketing experts, discreet members of the Syndikat collective.
We built Probedots for operators: we believe managing an assistant happens in a web interface, not in lines of code. Conversation review, conversation analytics dashboards, prompt optimization, the essentials are there.
Being good marketers, simple agent creation for your campaigns (hello Christmas agent!), agent A/B tests for optimization, CRM retargeting, 1:1 personalization of your site post-conversation, GEO, we have that too.
But we are lucid: we are well placed to know that with Claude, you could rebuild all of it yourself in a week.
The difference is Syndikat. We are ourselves shopping assistant PMs for our clients, and to maximize your conversation KPIs, you are in the right place.
Alex Gorius, Founder @ Probedots