Use cases by industry13 use cases

AI automation for manufacturers and wholesale distributors

The typical scene is an inside sales team retyping purchase orders. Customers send PDFs, Excel order forms, email text and WhatsApp photos of handwritten lists, each with their own part numbers, and six people key them into SAP Business One, Business Central, abas or proALPHA line by line. WizCommerce, which sells order-entry software, puts manual entry at about 20 minutes and $10 to $12 per order. Treat that as the vendor's figure and count your own.

The pattern repeats upstream. Estimators read drawings and STEP files to answer RFQs; CloudNC says its quote agent brings a 2 to 4 hour RFQ down to 5 to 9 minutes, a vendor claim for CNC machining, not for sheet metal, assemblies or castings. Buyers chase suppliers for order confirmations in three languages. Order entry automation, RFQ-to-quote drafting and a supplier expediting agent cover those three desks.

Deadlines are arriving from outside too. Germany requires businesses above EUR 800,000 in prior-year turnover to issue structured e-invoices for domestic B2B sales from 1 January 2027, and everyone from 2028. Many industry ERPs cannot produce XRechnung or ZUGFeRD, which is what e-invoice issuing solves without replacing the ERP.

What I would build for manufacturing and distribution

The desks where a mid-sized manufacturer loses hours

  • Customer part numbers

    Every customer uses its own codes, units and pack sizes. A cross-reference table, plus fuzzy and semantic matching where it falls short, maps them to your SKUs, and every correction a rep makes becomes a new mapping instead of a repeated mistake.

  • Purchase orders that differ from the quote

    A PO that changes price, quantity, date or drawing revision against the quote is caught before the ERP order exists, not when the wrong part is already on the machine.

  • RFQ packs with mixed revisions

    Drawings arrive in several revisions, in inches and millimeters, without surface treatments. The model lists every assumption it made; machine time comes from routing history and a cost model in code, and the estimator sets the price.

  • Supplier compliance paperwork

    CBAM emissions data, REACH and RoHS declarations and conflict-minerals documents must be requested from suppliers, checked for completeness and filed per part. It is chasing work with a validation step, which suits an agent well.

  • An ERP nobody can ask

    SAP Business One has no native AI agent and no official MCP server. A scoped MCP server lets sales ask Claude or Copilot about stock and orders without being handed the database.

Where a person signs off, and when EDI or a product is enough

  • Substitutions and price changes

    The system never approves a substitute part or a changed price on its own. Discontinued items, off-contract prices and low-confidence lines wait in a queue for inside sales.

  • Customers already on EDI

    Large customers sending EDI through SPS Commerce or a VAN do not need AI at all. The build targets the long tail that still emails PDFs and spreadsheets.

  • Food service ordering

    Choco has raised $328M and runs order and voice agents for food distributors. For a food wholesaler on a system it supports, buying is the faster route.

  • Machining quotes

    Paperless Parts and CloudNC already serve CNC job shops. Custom quoting pays for sheet metal, assemblies, castings and European ERPs those tools do not connect to.

Software these builds usually connect to

The systems do not get replaced. The build sits across them, reads from them through their APIs, and writes results back.

SAP Business One / Microsoft Dynamics 365 Business Central / NetSuite / abas / proALPHA / Epicor Prophet 21 / Infor CSD / Odoo / Sage / Acumatica / SolidWorks / SPS Commerce

Frequently asked questions

Can AI enter customer purchase orders into our ERP?

Yes, for orders that arrive as PDFs, spreadsheets, email text or photos. The model extracts the lines; mapping to your SKUs and checking price and stock are code; the sales order is created through the ERP's API, such as the SAP Business One Service Layer, keyed so a resent email cannot create a second order. Uncertain lines go to a rep with the source highlighted.

Is there an AI agent for SAP Business One?

Not from SAP. Business One has no native AI agent and no official MCP server. The practical route is a custom integration through the Service Layer or the DI API, with a secure connector when the system runs on-premises or with a hosting partner. Order entry, supplier confirmations and stock questions are the usual first builds, because that is where B1 users retype the most.

Do we have to replace our ERP to issue e-invoices in Germany from 2027?

Usually not. If your vendor offers an XRechnung or ZUGFeRD update at a sensible price, take it. If not, a converter layer reads invoice data from the database, an export or, as a last resort, the print output, maps it to the EN 16931 model, validates it with the KoSIT validator and sends it by email or Peppol. Most of that is plain code; AI helps with field mapping and reading print files.

Where should a manufacturer start with AI?

Where retyping is heaviest and the rules are clearest, usually customer order entry or the shared orders inbox. Run it in draft mode first so inside sales reviews every order it creates, measure accuracy per customer, and switch customers to automatic entry one at a time as their formats prove stable. Quoting and supplier chasing come next, because they reuse the same extraction and ERP connection.

Also browse: Operations and logistics · Finance and accounting · SAP Business One · NetSuite · all use cases