AI by sector · Manufacturers
How can manufacturers use AI without changing their ERP?
A typical day. Orders arrive as PDFs by email, and each one is re-keyed into the ERP. A supplier lets its lead time slip without warning. Quotes wait on an answer from the shop floor. Invoice entry gets pushed back to Friday. Every one of these workflows already exists in your ERP. What is missing is the labour that feeds it, and a supervised AI agent can supply that work.
In brief
How does AI help a manufacturing SME?
Djtal is the operational AI specialist for Swiss companies. Its supervised agent feeds a manufacturing SME's ERP (incoming orders, quotes, supplier lead times, invoices and master data). Odoo, Dynamics 365 or SAP Business One remains the system of record; sensitive data is mapped first and approval points are agreed in writing at scoping.
What the agent takes on
Five workflows, with a clear line between the agent and your teams.
| Workflow | What the agent takes on | What stays with your teams |
|---|---|---|
| Incoming customer orders (email, PDF, portal) | Reads the order, matches it to the quote and the catalogue, prepares the entry in the ERP and flags price or lead-time discrepancies | Resolves the discrepancies, approves the order |
| Price enquiries and quotes | Prepares the quote from the catalogue, the customer's history and the discount rules | Approves prices and margins, signs off each quote before it goes out |
| Supplier lead-time follow-up | Chases suppliers, updates lead times in the ERP and raises an alert when a delay affects a customer order | Sets priorities, negotiates |
| Supplier invoices | Reads the invoice, matches it to the order and the goods receipt, prepares the allocation | Checks the cases outside the agreed framework, approves payment |
| Master data (items, customers, suppliers) | Detects duplicates and missing fields, prepares up-to-date records | Approves creations and merges |
Gartner (June 2025) forecasts that more than 40% of agentic AI projects will be abandoned by the end of 2027, most often for lack of a profitable use case. The five workflows above are the opposite case. They have daily volume, known rules and a measurable gain within the first weeks, on the kind of process that successful AI projects have in common.
Your ERP stays
Odoo, Dynamics 365 and SAP Business One stay in place, and the agent adapts to them.
A manufacturing SME has years of items, bills of materials, price lists and habits stored in its ERP, and production depends on it every day. Changing it just to add AI would cost more than it gains; the agent works around it instead. It reads the mailbox, the PDFs and the exports, and it writes through APIs where your system provides them (Odoo and Dynamics 365 both offer full ones, and we work with Odoo regularly). Where no interface exists, it prepares the work for your teams.
If your vendor is building AI into its product (Copilot in Dynamics 365, the AI functions in Odoo), we start there. What your licence already includes is switched on first, and a dedicated agent covers the rest, in particular the work between systems. We apply the same logic to Zia at Zoho and to the AI built into ERPs.
The method
Four steps, and the metric comes first.
- 1.
Map the current workflows
Over half a day, your teams show us where the work piles up, the data involved (prices, drawings, customer data) and your ERP's interfaces.
- 2.
A pilot on one pain point
Order entry or supplier chasing, a single workflow to begin with. The metric is written down before any configuration starts.
- 3.
Put figures on the before and after
Entry time per order, errors avoided, supplier delays detected earlier. The numbers from your own order book decide what comes next.
- 4.
Extend to the next workflow
Every added workflow carries over the written scope, the approval at steps that commit the company and the action log. When the engagement ends, your teams hold the keys.
At Les Huiles de Jil, a Swiss brand that sells online, a single agent in production sorts the email inbox, enters supplier invoices into the accounts and enriches the CRM as conversations go on, under the supervision of management. Its approval points match the ones in the table above.
Frequently asked questions
The questions manufacturing SMEs ask about AI.
Can manufacturers use AI without replacing their ERP?
Yes. Your ERP keeps its role, its master data and its years of configuration. Djtal's agent reads emails, order PDFs and exports, then writes through the Odoo or Dynamics 365 APIs, imports or work prepared for your teams.
Where does a manufacturer's sensitive data go when an AI agent is used?
Djtal settles that at scoping, before anything else. We map which data the agent touches (purchase prices, margins, drawings, customer data covered by the Swiss Federal Act on Data Protection, FADP) and where it is processed and hosted, then limit the scope to the data the workflow needs and document it in writing. An agent that prepares quotes does not need to read your manufacturing drawings.
Odoo and Dynamics 365 already have their own AI. Why add an agent?
What your licence includes is switched on first (Copilot in Dynamics 365, the AI functions in Odoo). These assistants help each user within a single screen. The agent covers what they leave untouched, which is the work between systems, such as the order PDF that arrives by email, the chasing of a supplier whose lead time slips and the link between the inbox and the ERP.
What does the agent do on its own, and what do your teams sign off?
The agent reads, keys in, prepares and flags. Anything that commits the company (a quote that goes out, an order confirmation, an accepted price variance) passes through human approval. The autonomy dial is set in writing at scoping and turned up as results come in.
Which workflow should come first, and how quickly will we see results?
The pilot covers a single workflow, often the entry of incoming orders or the chasing of suppliers, with a metric defined beforehand (entry time, errors, delays detected earlier). Results arrive within weeks. The other workflows follow on the strength of that figure.
Companies that want the full picture before a pilot start with the AI strategy audit (workflow map, agent permissions, pilot plan).
Where to begin
Start from this week's orders.
Bring the orders that came in this week. Together we pick the workflow an agent would relieve first and list what your sensitive data needs at scoping. The decision to continue stays with you.
Last updated: