← All Projects

Agentic Partner Desk for Industrial Connectivity

Enquiries in five languages become correct quotes and checked purchase orders, with code setting every price and a manager approving every commitment.

Agentic Partner Desk for Industrial Connectivity

A sales and order desk for a channel partner of an industrial connector maker. One agent sorts the inbox and drafts quotes, one reads purchase orders and tracks orders, and one answers technical questions from published product data only. A voice agent can run any of it. The model reads and writes the words. Code writes every part number, quantity and price, and a manager approves discounts, large quotes and new delivery promises.

Distributors of technical products spend hours turning loosely written enquiries into correct quotes, and more hours checking that purchase orders match what was quoted. Customers write in their own language. A wrong part, a missed pack size or an unapproved discount costs margin and trust, so the work cannot simply be handed to a model.

The inbox reads enquiries in five languages and ranks them by value and urgency. In the quote flow the model extracts the requirement, and code matches the set, prices it and checks it against margin and approval rules. A purchase order check compares each line to the quote, and a datasheet agent says "not stated" when the published data is silent. Every action goes into an audit trail, and a guided demo runs the whole workflow live.

The desk reads enquiries in five languages, prices every quote in code, and holds discounts, large quotes and new delivery promises for a manager. The test purchase order carried three planted mistakes. The check found all three, plus a fourth, a lead time, that I had not planted.

Status
Working system
Client
Confidential

Built to the Governed Cognitive Layer: the model reads and drafts, deterministic checks decide what it may do, and a named person approves anything that commits the business.

Mid-market takeaway

Let software compare documents line by line. A purchase order checked against its quote by code catches the errors people skim past, including ones nobody planted. Keep the model on the words, keep the numbers in code, and send discounts and delivery promises to a named person.

Working system

Case study, talk deck and live demo

The case study, the talk deck and the working desk sit behind a password. I walk partner and sales teams through it live.

Ask for a walkthrough →

The same design rules behind this build, applied to your business in half a day.

Book the WorkshopDiscuss a build like this

Screens from the system

Quote pipeline showing which steps the model does and which are done by code or a person, with the extracted requirement from a Japanese enquiry
The model reads the email. Code matches, prices and checks. A person approves.
Excerpt from the audit trail showing agent reads and checks alongside a person's discount and manager approval
An excerpt from the audit trail: agent steps and human decisions side by side.
Inbox sorted by priority with enquiries in Vietnamese, English, Japanese and Thai, each with an estimated value and next step
Enquiries in several languages, sorted by value and urgency, with the next step for each.

Demo data throughout. Customer names, accounts and payees are fictional.

Terence Kok
Before You Go

I seeded the test purchase order with three mistakes to see whether the check would catch them. It caught all three and then flagged a fourth I had not planted: a lead time that could not meet the customer's date. A person checking that order at the end of a long day would probably have missed it, and the customer would have found out from a late delivery.

Terence Kok