AI Sales Agent is a sales tool, not just a support chat. To fully harness its potential, it’s worth knowing how to configure, monitor and optimise it correctly. Here are some best practices.
1. Set up the chat as an advisor, not a helpdesk
AI Sales Agent in edrone is optimised for product recommendations and supporting purchasing decisions – not for after-sales service. Ensure that:
Content in the knowledge base focuses on helping customers choose a product
Welcome bubbles encourage a conversation about the customer’s needs, not just questions about orders
Redirection to a human (escalation) is configured for after-sales enquiries
2. Monitor reports regularly
Go to Reports → AI Sales Agent. The report has two tabs and a date range picker.
The Sales tab shows whether the chat brings sales:
With and without chat – compares the conversion rate and average order value of customers who talked to the chat with visits without a conversation. It also shows how soon customers buy after their last chat interaction: up to 1 hour, 1–24 hours or 1–7 days.
Sales from chat – sales from orders in which the customer clicked a product the chat recommended or added it to the cart within 7 days. A funnel below goes from conversations to recommended products, clicks and orders.
Orders from chat – a sales breakdown, including orders from customers who talked to the chat but didn’t click anything, and a list of recent orders. Each order shows what the customer did, a link to the conversation and the order details.
Most recommended products – products with their recommendations, clicks and CTR.
The Conversations tab shows what customers ask about and when they write:
When customers write – conversations and sales outside your team’s working hours, conversations the chat handled without handing them over to your team, and conversations by hour and by weekday. If you haven’t set working hours, the after-hours figures show as unavailable.
What your customers ask about – the most common conversation topics.
Conversations with the chat – every conversation with the customer’s question and its result: order, handed over, or products recommended.
Keep in mind that the comparison with visits without a conversation shows a correlation, not proof that the chat caused the difference.
To share the results with the shop owner or a manager, use Show off your results. It opens a one-page summary of the selected period that you can download as a PDF.
Tip: Check the reports at least once a week, especially at the start. This is the only way to optimise performance.
3. Analyse conversation transcripts
In edrone Inbox, you’ll find full transcripts of all conversations. It’s a goldmine of information:
What questions do customers ask most often? → Add answers to the knowledge base
Which products does AI Sales Agent struggle with? → Update the feed and descriptions
Where do customers drop out of the conversation? → Think about what can be improved
You don’t have to read every transcript to spot frequent questions. In Reports → AI Sales Agent, What your customers ask about shows the most common topics. Conversations with the chat lists each conversation with the customer’s question and its result. Use the filters to narrow the list, open a conversation in Inbox or export the list to CSV.
4. Test different welcome bubbles
The content of welcome bubbles has a big impact on how many people open the chat. Test different versions:
Version | Example |
Problem-focused | "Not sure what to choose? I can help!" |
Benefit-focused | "I’ll find the perfect product for you" |
Urgency | "This week’s bestsellers – ask me!" |
Compare the number of conversations before and after changing the chat bubble.
5. Tailor the chat to campaigns and seasons
Chat can be particularly effective when combined with active campaigns:
Black Friday / festive season – add information about promotions and bestsellers for the period to the knowledge base
New collections – update the feed and shopping guides before the launch
Email campaigns – if you’re sending emails about a specific category, set the chat to activate on that category’s page
6. Don’t disable chat on mobile
Over half of the traffic in most shops comes from mobile devices. Disabling chat on mobile means missing out on a large proportion of potential conversations and conversions.
7. Patience and iteration
The chat learns from what you provide it with. The first few weeks are a calibration period – track results, expand the knowledge base and optimise settings. Results improve as the quality of data and the number of conversations increase.
