What We Learned From Launching ROBOSALES.AI
Sep 30, 2025
5 min read
Article
What We Learned From Launching ROBOSALES.AI
When we launched ROBOSALES.AI, our goal was simple: turn online shopping conversations into conversions. After running thousands of customer interactions, we’ve gathered valuable lessons on how Gen-AI sales assistants really work in practice.
The Conversations That Matter
Our early deployment of ROBOSALES.AI was specifically for a sporting goods e-shop as an AI tennis sales assistant and advisor, which is why so many of these examples revolve around rackets. Tennis proved to be the perfect proving ground: a high-traffic category with complex choices and strong demand for expert guidance. It highlighted exactly where conversational AI can deliver the most value.
The majority of customer questions fall into two clear categories. Some are smart search requests, like “Find me a racket for my playing style” or “Show me something similar to the racket I already use.” Others are general advice requests, such as “What should I look for when choosing a racket for my 5-year-old daughter?” or “I often play from the baseline, what kind of racket suits me?”
Alongside these practical shopping questions, we’ve also noticed customers leaning on the assistant’s broader knowledge, the kind of information that isn’t in the product catalog or provided by the e-shop at all, but comes from the underlying language model. Instead of asking only about specific products, they’ll bring up topics like training tips, equipment guidance, or how to choose between different model series. A parent might want to understand what features matter most when buying a first racket for a child, while a player may ask about the differences between control- and power-oriented frames. These aren’t database lookups; they’re open-ended conversations that blend general expertise with commercial recommendations, making the assistant feel more like a coach or in-store consultant than a search bar.
It’s no surprise: products with the highest traffic and the most nuance are where customers want the most guidance. But here’s what’s interesting: once shoppers start engaging, they don’t stop at a single click. On average, customers who interact with our gen AI sales assistant click three to four times during a single conversation. That means they’re not just curious, they’re actively exploring, comparing, and moving toward a decision.
The Results: Engagement That Drives Sales
The numbers speak for themselves. In around 60% of conversations, ROBOSALES.AI recommends a particular product (rest are general questions). In 2/3 of those cases, the customer clicks on at least one recommended product. And when they do click, they rarely stop at one, they average 3-4 product clicks in a single conversation.
This behavior is important. It shows that once customers begin exploring through ROBOSALES.AI, they don’t treat it as a one-off interaction but as a trusted guide. They dig deeper, compare options, and take steps closer to purchase. That’s not just customer support; it’s a sales engine in action.
The Human Factor in AI Conversations
One of the most fascinating insights has been how customers treat the assistant almost like a person. Many begin their interaction with a question such as “What can you do?”, curious about the capabilities rather than the catalog itself. Others go even further, writing to ROBOSALES.AI as if they were composing a formal email, long, carefully structured paragraphs, polite closings, even signing their names at the end.
What’s striking is that despite “AI” being in the assistant’s name, customers often assume they are speaking to a human on the other side. This shows us two things: expectations are high, and more importantly, people are comfortable engaging in a natural, human-like way. That kind of behavior is exactly what builds trust and drives conversion.

The Behind-the-Scenes Reality
Getting to these results hasn’t been without its challenges. A few key takeaways from implementation:
Data quality is everything.
Product feeds, images, and descriptions must be clean and consistent. “Shit in, shit out” applies: if data is poor, recommendations lose their accuracy, and customers quickly lose confidence.
Each category is unique.
Every product category has its own business rules that require analysis and fine-tuning. We have to become experts in each category, before launching, to make sure the recommendations are correct. Tennis rackets, sneakers and apparel are all different from one another.
Trust is fragile.
We’ve learned that trust is incredibly fragile. Shoppers come into a conversation with curiosity, but also with skepticism, they want to see if the assistant can really deliver useful, relevant answers. The moment ROBOSALES.AI suggests a product that’s poorly tagged, irrelevant to the query, or missing clear information, confidence erodes almost instantly. And once a customer decides the assistant can’t be trusted, they rarely give it a second chance.
Testing only gets you so far.
Before launch, we spent months preparing ROBOSALES.AI by designing and testing more than a thousand conversation scenarios. We tried to imagine every possible question a customer could ask, from the straightforward “Find me a racket under $200” to the more nuanced “Which racket helps me generate more spin if I have a two-handed backhand?” The goal was to cover the widest range of situations so the assistant would respond reliably and build trust from day one.
But real-world customers have a way of humbling even the most comprehensive test plans. As soon as the system went live, we started seeing interactions we hadn’t anticipated. Some customers wrote to the assistant as if they were crafting a formal email, complete with long introductions, multiple paragraphs of context, polite closings, and even signing their name at the end. Others sent highly professional queries that read more like business correspondence than a quick shopping question. And because many of these messages were so long and detailed, they pushed the assistant into handling complexity beyond what we had initially expected.
This has been both a challenge and a revelation. On one hand, it proved that no amount of pre-launch testing can fully predict how real people will engage with AI. On the other, it showed us just how naturally shoppers slip into human-style communication when the experience feels conversational. Instead of short prompts like you might use with ChatGPT, customers often write in flowing sentences, sharing their needs and preferences as if they were speaking to a sales consultant in-store. That level of depth is exactly what allows ROBOSALES.AI to shine, because when the conversation is richer, the recommendations can be more accurate and personal.

What This Means for E-Shops
At its core, ROBOSALES.AI is more than a chatbot, it’s a growth tool for e-commerce. By engaging shoppers in natural conversations, it keeps them active on-site longer and moves them deeper into the product catalog. By guiding search with both product data and broader expertise, it improves discoverability and helps customers find the right fit faster. And by responding in a way that feels human, it builds trust and creates a relationship that goes beyond a single transaction.
For e-shops, this translates into higher engagement, smarter product search, stronger customer relationships, and ultimately, more sales. That’s the promise of AI in commerce: not just answering questions, but transforming conversations into conversions.