A mid-sized manufacturer's sales team spent a third of its week qualifying enquiries, looking up prices and writing follow-ups. In three months we turned this routine over to a team of AI agents. The salespeople went back to what they do best: talking to customers.
Starting point
Enquiries arrived by email, through the website and by phone. Someone had to read each one, log it in the CRM, find the customer, check prices and prepare a quote. Average time from enquiry to quote: 3.5 working days.
- 3 salespeople, 1 assistant, over 400 enquiries a month,
- CRM in Dataverse, price lists in Excel on SharePoint,
- discounts approved “in the corridor”, with no audit trail.
Solution architecture
We built on the four-layer architecture described on the AI agents page. The key was giving agents data access via MCP servers with the same permissions people have.
What each agent does
- Qualifier extracts the enquiry details, finds the customer in the CRM and scores the lead.
- Pricer builds the line items from the current price list and the customer's history.
- Assistant prepares the quote and sends it to the salesperson for approval in Teams.
“What surprised us most: the agents don't just make fewer pricing mistakes. They never forget a follow-up either.”
client's head of sales
Results after three months
Time from enquiry to quote fell to 4 hours, and salespeople approve 82% of quotes without changes. The qualification agent's configuration is surprisingly short:
Detailed figures, the brief template and the webinar recording are in the article attachments.