How did you decide which AI models to support in the Totebot.ai architecture, and what criteria do you use to choose the “best” model for a specific store?

We wanted to support the most popular AI models – from OpenAI’s GPT, through Claude Sonnet and Haiku from Anthropic, all the way to Gemini Flash and Pro from Google. At the moment, we support a total of fifteen models, but for e-commerce use cases the best performer has been Gemini Flash.

It has become the default choice for all stores on our platform because it offers the best balance between price and answer quality, especially in the domain of catalog search and suggesting relevant products.

In more advanced plans, users can configure and select which model they want to use, depending on their needs.

What are the biggest challenges when integrating Totebot with platforms like Shopify or WooCommerce, particularly in the area of inventory synchronization and real-time updates?

The biggest challenge was to enable a simple 1-click setup for merchants while making sure our AI receives rich and accurate information about each individual product. This is crucial for Totebot to provide the customer with exactly what they need.

When it comes to real-time inventory updates, this hasn’t been an issue – the bot always knows the current stock status and only suggests products that are actually available. This way, the customer never receives a frustrating recommendation for a product that’s sold out.

How do you measure the success or ROI of Totebot AI agents – which metrics matter most to you, and can you share concrete results from the Apipet experience?

Inside the Totebot dashboard there is an advanced analytics module that precisely tracks performance – from which answer led to a purchase to the exact transaction value. This gives us a clear view of the incremental impact of the AI agent, because if the customer hadn’t received a quick answer, they would have had to search the website or contact support, which often leads to abandoned purchases. With Totebot, they get the solution in the shortest possible time, which directly increases conversion.

On top of that, Totebot has significantly relieved customer support – it answers over 200 messages per day, many of which would otherwise end up in email or over the phone. This reduces costs, speeds up response times, and raises service quality.

In the beginning, we wondered whether customers would even want to interact with an AI bot, but practice showed the opposite – more than 90% of conversations carry positive sentiment. That’s a strong signal that users actually appreciate instant and precise answers.

We believe these are just the early steps of the agentic AI era in e-commerce, which will soon become the standard. We’re proud that with Totebot we can combine a better customer experience, advanced support, and a sales agent on steroids.