Why a screen-reading AI that clicks for your users matters more than another ad tool
Every cross-border seller I know is fighting the same silent leak: they spend thousands on traffic, perfect their listings, dial in their ads — and then watch a 60–80% early churn rate eat the math alive. You can fix your supply chain, optimize your Amazon PPC, and build a beautiful Shopify store, but if a first-time buyer can’t figure out your checkout flow or a new B2B customer can’t navigate your portal, all that acquisition spend is just kindling. We’ve outsourced every part of the funnel to software except the part where a human actually helps another human get value. That’s the gap Nimbia is going after, and it’s why I’m writing about it here instead of yet another AI ad-copy generator. The product itself is an AI agent that embeds in your app, watches your user’s screen, talks to them like a patient CS rep, and clicks through your UI on their behalf. For a cross-border operator, that’s not a novelty — that’s a conversion-rate lever you can’t buy with more traffic.
The problem: your self-serve docs are a graveyard, and your CS team is a VIP lounge
Let me paint the picture the maker, Joris Machielse, lays out in his launch comments. His first B2B company had him personally onboarding every customer one-on-one. His next product was in user onboarding itself — demo videos, similar to Supademo. So he’s spent years inside the exact problem every SaaS and every DTC brand with a subscription portal hits: live onboarding calls convert, but they don’t scale. You only have so many hours, so you give the “top 10%” of customers — the ones paying you the most — actual human help from Customer Success. Everyone else gets a link to self-service docs and a prayer. And then you wonder why 60–80% of those users churn early, as Machielse notes in his comment on the launch page.
Sound familiar? It should. That’s not just a SaaS problem. That’s the exact dynamic playing out on your Amazon listings, your Shopify store, your TikTok Shop. Think about the last time you bought from a new brand’s site and hit a confusing size guide or a checkout wall. Did you email them? No. You left. You didn’t read their FAQ. You didn’t watch their demo video. You just bounced. The “self-serve docs” approach to cross-border e-commerce is the reason your return rate spikes on products that need a modicum of setup, and why your subscription box cancellations cluster in the first 30 days. People don’t churn because they hate your product. They churn because they never got value out of it, and nobody was there to walk them through.
What Nimbia does differently is collapse that gap. It’s not a chatbot that gives canned answers. It’s an AI that runs a live, free-form conversation with your user, sees their actual screen, and can click on things for them — like a human CS agent screensharing with a customer, except it’s there for every single user, not just the top decile. The maker describes it as an experiment he ran with GPT-4o about two years ago: he gave the model a goal inside his app plus a screenshot, asked for the next step, performed it himself, sent a new screenshot, and repeated. It walked through the entire flow on the first attempt. That’s the seed. Eighteen months later, it’s a product you add with a single line of JavaScript, trained on your product and even on recordings of your past onboarding calls.
How it differs from the incumbent graveyard: tours, tooltips, and chatbots
To understand why Nimbia matters, you have to look at what it’s replacing. The legacy stack for user onboarding is a pile of band-aids: product tours, tooltip overlays, help docs, and increasingly, chat widgets that are really just search boxes for FAQs. The maker calls this out directly — he says his product “obliterates the legacy approach of product tours, help docs and tooltip overlays.” And he’s right, but let me be more specific about why, because this is where cross-border sellers should pay attention.
Product tours, like what Appcues or Userpilot offer, are scripted. They show a user five steps in a fixed order and assume that’s what everyone needs. But real users don’t follow scripts. A new customer on your wholesale portal doesn’t want a tour of the dashboard; they want to upload their resale certificate and place a bulk order. A new subscriber on your DTC site doesn’t want a tooltip about your loyalty program; they want to change their delivery date. Tours and tooltips are the equivalent of a salesperson who gives the same pitch to every customer regardless of what they actually asked for. They’re better than nothing, but only barely.
Chatbots, meanwhile, are the other side of the coin. They’re conversational, but they’re also mostly useless. Most are trained on your help docs, so they give answers that sound plausible but miss the user’s actual context. They can’t see the screen. They can’t say, “Oh, I see you’re stuck on the payment wall — let me click ‘Apply Coupon’ for you.” They just link to another article. Nimbia’s difference is that it’s grounded — the maker explicitly says it’s “grounded of course so it doesn’t hallucinate” in his response about free-form conversations. It’s not generating answers from a knowledge base; it’s looking at what’s on the user’s screen, understanding their intent, and acting on it. That’s the difference between a support ticket and a screenshare with a friendly rep who just fixes the thing for you.
Why Amazon sellers should care more than Shopify ones
Here’s a hot take: most of the conversation around Nimbia is going to be about SaaS onboarding, because that’s the launch audience. But cross-border sellers on Amazon should be paying even closer attention, because you have the worst onboarding problem of all — and you don’t even control the platform. Your Amazon listing is your front door, but the “help” you can give a confused customer is limited to a bullet point and a product image. You can’t embed a script. You can’t add a tour. You can’t even send them a follow-up email if they abandon the listing (Amazon owns that relationship). The only help you can offer is the listing itself.
That means the lesson from Nimbia for Amazon sellers isn’t “install this on your listing” — you can’t. It’s a deeper lesson about the physics of conversion: a user who gets personalized, real-time help converts dramatically better than one who gets static content. The maker’s first company grew 40% faster with this approach, and he says the first company running Nimbia is growing 40% faster today. For an Amazon seller, that’s the argument for doing whatever you can to move buyers toward a channel you control — your own Shopify site, your email list, your brand registry content — where you can actually deploy tools like this. Amazon is a rental. Your own storefront is where you can build the kind of guided experience that turns a first-time buyer into a repeat customer.
What cross-border sellers can borrow from Nimbia’s approach
You don’t have to be a SaaS company to steal the playbook here. The core insight — give every user the help that only your top 10% of customers get — applies to any business where a customer needs to do something after purchase. Let me give you three concrete ways to borrow this thinking.
First, record your best onboarding calls and turn them into training data. The maker says Nimbia can be trained on recordings of your past onboarding calls. That’s a brilliant move, and you can do a version of it without any AI. Go back through your support tickets, your email threads, your chat logs. Find the five questions that come up over and over for new customers. Then build your product content around those questions — not around what you think is important. Your FAQ should be a mirror of your customers’ actual confusion, not a reflection of your product roadmap. If you’re on Shopify, that means rewriting your product descriptions to preempt those questions. If you’re on Amazon, that means using every bullet point to answer the objections your support team hears daily. The words your customers use to describe their problems — the maker mentions “hearing what people actually get stuck on, the words they use for their problem” — are gold. You’re probably not mining them.
Second, think about the “single line of JavaScript” as a metaphor for reducing friction everywhere. Nimbia’s install is trivially easy — that’s a feature. The harder it is to get help, the less likely people will use it. Cross-border sellers have the opposite instinct: we hide support behind contact forms, we bury return policies, we make customers jump through hoops to get a human. Stop that. Put your phone number on your checkout page. Put a live chat widget on your product pages. Make your return policy a one-click label generator. The easier you make it for a customer to get help, the fewer returns you’ll process — because most returns aren’t malicious, they’re just confused buyers who didn’t get what they expected.
Third, segment your help by customer value, not by customer size. The maker’s point about the top 10% getting human help while the rest get docs is the default for most businesses. But the math is backwards. Your churn risk isn’t concentrated in your biggest customers — they have procurement teams and contracts. Your churn risk is in the long tail: the small buyers, the first-timers, the ones who are one confusing step away from abandoning you forever. For a cross-border seller, that’s the customer who buys one item, gets confused about international shipping, and never comes back. Give that segment the best experience, not the worst.
Where the math breaks: the feedback loop problem
There’s a weakness in Nimbia’s pitch that the commenters on Product Hunt immediately spotted, and it’s worth examining because it tells you something about where AI-assisted support will and won’t replace humans. Joris van der Steuijt asks the question that should be on every operator’s mind: when the AI runs the call instead of you, how do you still get the “oh interesting, they always struggle here” insight? He notes that live calls gave him feedback on the product itself — “hearing what people actually get stuck on, the words they use for their problem, stuff you’d never get from a support ticket.” That’s a real loss.
The maker’s answer is that they “mine feedback through the transcripts and call recordings.” That’s true, but it’s a different kind of feedback. A transcript tells you what the user said. It doesn’t tell you the emotional texture — the hesitation, the frustration, the moment when they almost gave up and then didn’t. A human CS rep on a live call feels that. They know when a customer is about to churn and can pivot. An AI, however well-trained, is still processing patterns, not reading a room.
For cross-border sellers, this is the cautionary tale. You can automate the help, but you can’t automate the learning. If you put an AI on your front line, you need a systematic way to review its transcripts, find the friction points, and feed those back into your product and your listings. Otherwise you’re just automating the part of the job that’s easy — showing up — and losing the part that’s hard — understanding why people struggle. The maker says the agent “will build a relationship with everyone, and will remember context for the next time they chat with someone.” That’s a nice feature, but it’s not the same as a founder who personally feels the pain of a customer’s confusion. Keep the humans in the loop, even if the AI is doing the talking.
The privacy and security ceiling
Another commenter, Anastasiia, asks a sharp question about what happens when a user gets redirected to a payment provider and back — does the session continue or get lost? The maker’s answer is revealing: the agent is embedded in the front-end code of the application, so if the user leaves that app, it won’t be available. For third-party SSO or payment redirects, Nimbia won’t help, and the maker says that’s intentional for security. They also note that the AI “explicitly cannot see payment information; only the page context around it,” because one of their customers has to be HIPAA compliant.
That’s the right call, but it’s also a ceiling. For cross-border e-commerce, the payment wall is exactly where most users get stuck — currency conversion, international address fields, failed card attempts, customs and duty calculations. If your AI helper goes dark right at the moment of maximum friction, you’ve automated the easy part of the funnel and left the hard part manual. The maker says they might add a different mechanism for that in the future, but it depends on customer needs. For now, if you’re evaluating Nimbia for a checkout flow, know that it won’t be there when the user hits Stripe or PayPal. That’s not a dealbreaker — it’s just a reminder that AI assistance has boundaries, and you should design your funnel to handle the handoff gracefully.
What I’d watch / test next
If I were running a cross-border operation, here’s what I’d do this week, based on what Nimbia is showing us:
Audit your own “top 10% vs. the rest” gap. List your top 20 customers by revenue. Now ask: what help do they get that your other 200 customers don’t? Write down the five things your CS team does for them that you don’t do for everyone else. That’s your activation checklist. If you can’t automate it, at least make it a self-serve flow.
Mine your support transcripts for the “words they use.” Pull your last 50 support tickets and your last 20 sales call recordings. Extract the exact phrases customers use to describe their problems. Rewrite your product descriptions, your FAQ, and your onboarding emails to use those words — not your marketing language. That’s the feedback loop the Product Hunt commenter was worried about losing, and you can build it without any AI.
If you run a SaaS or a portal-based DTC model, get on Nimbia’s waitlist. It’s a single line of JavaScript, and the first company running it is growing 40% faster, per the maker. That’s a small sample size, but the mechanism — giving every user the help only your top 10% got — is sound. Test it on your onboarding flow, not your checkout, and measure activation rate, not just conversion.
For Amazon sellers: don’t try to automate help you can’t control. Instead, move the relationship. Use your inserts, your brand registry, and your email capture to pull buyers toward a channel where you can actually deploy tools like this. The future of cross-border e-commerce isn’t just better ads — it’s better help at the exact moment a customer needs it. That’s a channel you have to own.
The takeaway isn’t “buy this tool.” It’s that the era of static help docs and canned tours is over. The next competitive advantage in cross-border e-commerce isn’t a faster ship time or a cheaper price — it’s being there for your customer when they’re stuck, in the way that only your best human reps used to be. Nimbia is one version of that future. The question is what you’re building.
