AI for order intake and quoting on a legacy ERP
Can AI agents handle order intake and quotes when your ERP is old or has no API? What they read, what they draft, where a person approves, and how to start.
An old ERP, or one without an API, does not rule out AI for order intake and quoting. AI agents can read incoming orders and quote requests, check them against the customer, the quote and the plan, and draft the order or the quote for a person to approve. What the ERP decides is how the agent reads and writes data, and that is checked before anything is built.
What the work looks like today
At many manufacturers that make to order, orders arrive as emails, PDFs and phone notes. Someone retypes them into the ERP, checks the prices against the quote and the date against the planning sheet, and chases anything that does not match. Quotes depend on the customer and their terms, the product configuration, material prices and capacity. Most of that knowledge sits with a few people, which is why the work slows down when they are away.
What an agent does, and what a person keeps
| Step | The agent | A person |
|---|---|---|
| Read the order or request | Reads the email or attachment and pulls out the customer, lines, quantities and dates | Steps in only when the agent cannot match it |
| Match to records | Matches the customer, the products and the quote the order refers to | Resolves anything the agent flags as unclear |
| Check | Checks prices against the quote and terms, and the delivery date against the plan | Decides on every exception |
| Draft | Drafts the order entry or the quote | Approves before anything is entered or sent |
| Record | Saves what was decided and why | Sets the rules the agent follows |
Anything that leaves the company, such as a quote or a confirmation to a customer, waits for a person by default.
What if the ERP has no API?
Then Ortelian connects through data exports or through read access to the database underneath the ERP. Which of the two fits depends on the ERP, and the access is agreed with your team before anything is built.
Writing back is a bigger step than reading. Starting with the agent reading and drafting, and a person entering, keeps the risk low while the standard is proven on real orders.
Why the agent needs context, not just the email
An order is only right or wrong against the customer's terms, the quote, the product and the plan. Those live in different systems and in people's heads. The agent works from a connected picture of them, with a source on each fact, rather than guessing from the email. The same order is worked through in what is a context graph for AI agents and what is an ontology for AI agents.
How to start
Pick one piece of work, such as order intake for one product line, and agree the standard on real orders before building. With Ortelian that starts with an audit of about two weeks on site, then two week build cycles until the first job meets the standard. The platform is hosted in the EU, and the IP built for you and your data stay with you. To check whether you are ready, see the AI readiness assessment. For other operations work, see AI agents for operations.
Questions people ask
Does this replace our ERP?
No. The ERP, planning tools and spreadsheets keep doing their jobs. Agents work alongside them and write back only where you have agreed they may.
Will the agent send quotes to customers on its own?
Not by default. Approvals and anything that leaves the company wait for a person. Which steps can later run without approval is your decision.
Is this the same as AI quoting software?
No. Quoting software is a product you configure. This is an agent working across your email, ERP and planning, set up with your team inside an engagement. There is no self service version.
Where does the agent run?
On the Ortelian platform, hosted in the EU and operated by Ortelian. Your team keeps working in the tools it already uses.
