450 Daily Sales Enquiries, Automated.

THE CHALLENGE

Volume Made The Judgement Work Impossible.

A national manufacturer's sales operation was processing roughly 450 inbound enquiries per day through a mixed-channel funnel - web forms, email, distributor portals, phone notes captured by reception. Each enquiry went through three manual steps: classification (product line, urgency, lead grade), routing (which sales pod, which territory, which technical specialist), and a first-touch response.

The team handling that work was experienced, but at 450 enquiries a day the queue ran 24-hour latency in busy periods. Distributor relationships started to fray. Web-form leads cooled before anyone replied. The internal estimate was that the funnel was leaking somewhere between 10% and 20% of qualified leads to slow response alone.

A previous attempt at rule-based automation had stalled - too many edge cases, too many product categories, too many distributor-specific quirks. The team had concluded the work needed human judgement and that scaling it meant scaling headcount.

THE APPROACH

Private AI Plus MCP. Inside The Customer's Environment.

run-e ran a Digit Automate engagement on the TonsleyAI platform - private AI inside the customer's environment, with safe, observable access to the CRM and product data via the Model Context Protocol (MCP).

Discovery mapped the full enquiry funnel and the human classification logic. Private-AI fine-tuning ran on the customer's historical enquiry data, product taxonomy and distributor records - entirely inside the customer's cloud tenant. No data left the environment.

An MCP-orchestrated workflow ingests each enquiry, classifies it against the product taxonomy, looks up the distributor and territory context, routes to the correct sales pod, and drafts the first-touch response. Every prompt, every tool call, every routing decision is logged - the supervisor team can audit any AI decision. The AI handles the high-confidence routine; edge cases bubble up to a supervisor exception queue. Humans stay in the loop for the work that needs judgement.

THE OUTCOME

From Triage Team To Exception Queue.

• Manual triage collapsed. The full-time team activity of routing 450 daily enquiries became a supervisory exception queue. Most enquiries flow through end-to-end without human touch.

• Latency dropped sharply. First-touch response times moved from a 24-hour worst case to minutes for the high-confidence path.

• Estimated lead leakage closed. The 10-20% slow-response leakage estimate compressed materially.

• Zero data exposure. The architecture means no proprietary product, distributor or customer data ever left the customer's environment. The compliance team signed off without conditions.

• Headcount stayed flat. The supervisor team kept their roles; the routine work that had been crowding out their judgement work was the bit that went.