AI for businesses / B2B workflow demo
Explore a B2B sales workflow: from a customer message to a quote, approval and a draft ERP order.
Workflow lab
An interactive simulation using fictional data. No AI model is called, no messages are sent and no ERP connection is made.
01 / 06
Hello, please quote for 40 white NV-100 window vents and 20 KG-200 grilles, delivered to our warehouse within 7 days. Customer: DEMO-104.
Hello, please quote for white NV-100 window vents and 20 KG-200 grilles, delivered to our warehouse within 7 days. Customer: DEMO-104.
The message is not a form. In a live implementation, AI identifies products, quantities, colour and requested delivery time without inventing missing details.
02 / 06
The quantity of vents is missing. The quote is paused while the assistant asks: “How many NV-100 units do you need?”
Quantity confirmed by the customer. The workflow can continue.
Validation rule: a product code, a quantity greater than zero and a customer ID are required before pricing.
03 / 06
| Product | Net unit price | In stock |
|---|---|---|
| NV-100 | 60 | |
| KG-200 | 30 |
Only 12 of the 20 grilles are available. AI does not promise a delivery date for the missing 8. In this scenario, the customer agrees to reduce the order.
Scope updated: 12 grilles, with no order placed for the remaining 8.
Prices and stock come from the catalogue; the discount comes from the customer record. The model is not the source of these values.
04 / 06
Illustrative prices. The discount applies to both products. VAT, shipping and the delivery date must be agreed separately. System rules calculate the amounts, not the language model.
05 / 06
A sales representative checks the items, prices and agreed scope. Approval below applies only to the simulated order draft. It does not send a quote or confirm delivery.
06 / 06
The draft contains order lines, the discount and a record of approval. In a live deployment, an integration submits it through an API, checks the response and prevents duplicate submissions.
Nothing has been sent. The delivery date remains unconfirmed.
How we design the implementation
It identifies intent, suggests product matches and drafts text. Missing information is flagged for clarification.
The catalogue, CRM and ERP provide the data. Rules check completeness, pricing and permitted actions.
Sales staff approve quotes and exceptions. The customer confirms scope changes. Approvals are recorded.
We evaluate a pilot using real enquiries: line-item accuracy, corrections, handling time and exceptions requiring human input. This example does not present client results.
First step
