Customers & past enquiries
Earlier RFQs, requirements and quotations provide context for the next enquiry.
Own prototype · logistics & technical procurement
From enquiry to quotation. With AI agents.
Reading emails, matching parts, finding earlier quotations: requests for quotation involve repetitive manual work. This prototype explores connecting those steps in one workflow. RFQ stands for Request for Quotation.
The process
An email starts the workflow.
The agent classifies the enquiry, draws on existing knowledge and prepares the next step. Follow-up questions and quotations remain subject to human review before sending.
- 01 · Incoming message
Email arrives
A new message triggers processing. The agent reads the enquiry and attachments.
- 02 · Classification
Is it an RFQ?
The agent identifies the request and who should handle it. Other messages follow the appropriate workflow.
- 03 · Context & checks
What do we already know?
Check the customer’s earlier RFQs and quotations. Review parts, suppliers and open questions.
Information missing
Prepare follow-up questions
The agent suggests an email with the open questions. A person reviews and sends it.
Customer reply → add context and check again.
Information complete
Draft a quotation
Combine suitable parts, suppliers and terms. Prepare quotation items and a reply to the customer.
A person reviews & approves
Confirm items, prices and delivery times. Only then is the quotation sent to the customer.
The foundation
Shared data instead of scattered information.
The greatest leverage comes from combining capabilities: the agent understands the enquiry. A shared database provides customer history, earlier quotations, parts and suppliers.
Parts & specifications
Items, technical specifications and known alternatives provide the basis for suitable quotation items.
Suppliers & terms
Sources, prices and delivery times are maintained together and checked for currency before approval.
AI helps with understanding and classification. Fixed rules check required fields, calculate totals and control approvals. Deterministic automation and AI each handle what they are suited to.
Target for standard enquiries
< 5 min
From incoming email to approved quotation.
A high-volume process with considerable potential.
When no questions remain and parts and suppliers are known, the target is a response in under five minutes. This requires current terms and someone available to approve promptly.
This is a target for testing, not a measured result. Time to draft and waiting time for approval are considered separately.
The pattern extends beyond logistics and technical procurement: recurring enquiries, a shared data source and clear approvals are useful starting points in purchasing, sales and service too.
Your process
Where does this work repeat in your organisation?
In the introductory call, we look at a specific workflow: which enquiries arrive, what data is available and what must a person decide?