AI for sales
From customer email to a Dynamics-ready sales order
A system that interprets emails and PDFs, applies commercial and logistics rules and prepares traceable orders without turning AI into a black box.
Built capability, continuously evolvingDozens of coordinated rules
Document, decision and order remain connected
Human review for exceptions
The challenge
The difficulty lies in the decisions
Creating an order is not about extracting four fields. The system must recognise intent, identify the customer and reference, resolve products and dimensions, split deliveries, prevent duplicates and decide what can be automated and what needs review.
Intent
Distinguish a new order from a query or an action on an existing order.
Customer and reference
Resolve company, customer order reference and possible collisions before creating records.
Product and format
Check codes, families, weights, dimensions and units, including unresolved items.
Delivery
Interpret dates and addresses and separate lines that require different orders.
The solution
Architecture and rules serving the process
AI understands; rules decide
AI structures the document and proposes an interpretation. Sensitive decisions go through explicit, testable and versioned rules.
Per-line creation and duplicate control
Each line can produce its own order. Existing references and collisions are checked before anything is created.
Exceptions remain visible
When an item or condition cannot be resolved safely, the system keeps the context and routes it to review rather than inventing an answer.
Actions on existing orders
A request to change, delay or cancel is recorded against the relevant order and is not mistaken for a new sale.
End-to-end flow
01 · Input
Email, PDF and sender context
02 · Interpretation
Proposed data and detected intent
03 · Decision
Customer, product, reference and delivery rules
04 · IDAX
Review, traceability and order preparation
05 · Dynamics
Controlled synchronisation with the ERP
Operational value
What changes in practice
A prepared order linked to the original document and its decisions.
Visible exceptions so the team reviews only what needs attention.
Lower risk of duplicates or acting on an incorrect interpretation.
Reusable rules that can evolve by customer without breaking the general process.
An anonymised case: the engineering and process are explained without revealing confidential data or attributing unpublished client metrics.
Explore IDAX