Skip to content
NuncilyNuncily
AI order processing

AI sales order processing for email orders

Turn orders that arrive by email, PDF or voicemail into checked entries in your order system, with only the unclear ones sent to a person.

  • 01Receive
  • 02Understand
  • 03Collect
  • 04Clarify
  • 05Validate
  • 06Act
  • 07Verify
  • 08Complete

What gets automated

  • Reading orders from email bodies, PDFs and spreadsheet attachments
  • Transcribing voicemail orders and reading them like any other order
  • Matching the customer's wording to your product codes and price lists
  • Checking quantities, prices, delivery dates and account status
  • Creating the order in your ERP or order system
  • Confirming with the customer, and asking about anything unclear

Outcome

Your order desk checks a short list of exceptions instead of keying every order by hand.

When a person is needed

Anything that needs judgment goes to your team with the whole thread and everything collected so far attached. The AI flags; a person decides.

Orders still arrive the way customers like to send them

Most businesses that sell to other businesses have a web shop or a portal. Most of their orders still don't come through it. They arrive as an email with a list in the body, a PDF purchase order, a spreadsheet, a photo of an order sheet, or a voicemail left after the office closed.

Somebody then reads each one, works out what the customer meant, finds the right product codes, checks the price and the account, and types the order into the system. It's careful, repetitive work, and it's the kind of job that gets a job advert of its own.

What the AI does with each order

  1. Reads the order wherever it arrived. Email text, PDF purchase orders, spreadsheet attachments and transcribed voicemails all go through the same path.
  2. Works out what the customer meant. "Two cases of the usual chicken" becomes a specific product code, because the build knows that customer's order history and your catalogue.
  3. Checks the order against your rules. Is the product active? Is the quantity sensible for this customer? Does the price match their price list? Is the account on hold? Is the delivery date one you can meet?
  4. Creates the order in your system. Through the system's API or its import route, exactly as your team would have entered it.
  5. Confirms with the customer. A confirmation goes back on the channel the order came in on, and if something is unclear, the question goes back in the same thread.

What stays with your team

Anything the rules can't settle. An unknown product, a quantity ten times the customer's normal order, a price disagreement, a customer who's over their credit limit: these are flagged, not guessed. Your team sees the original message next to what was extracted, fixes or approves it, and moves on.

Over a pilot, the list of reasons something gets flagged is also the most useful report you'll get about your order process. It shows exactly where the time was going.

Why not just push customers onto a portal?

Because most of them won't move, and the ones who order the most are often the least willing. A chef ordering at eleven at night, a site manager on a phone, a buyer whose own system generates PDF purchase orders: none of them wants a new login. Automating the order desk means you stop asking customers to change, and stop paying people to compensate for it.

How a build works

Order processing is a custom build, because it depends on your catalogue, your customers and your system. It runs in three stages:

  • Pilot. One order channel, a sample of real orders, two to three weeks. We agree the measures up front: orders processed without a touch, time per order, and errors caught.
  • Build. The full set of channels and rules, connected to your system, with your team approving everything during a supervised launch.
  • Run and improve. Monitoring, new rules as your catalogue and customers change, and new workflows when you're ready for them.

See how custom builds work for pricing and what you own at the end.

Common questions

Which systems can you put orders into?

Whatever your orders live in today, whether that's an ERP such as Sage, NetSuite, SAP Business One or Microsoft Dynamics, an accounting package, or a system built for your business. Where the system has an API we use it; where it doesn't, we use the import route it supports.

What happens when the AI isn't sure about an order?

It doesn't guess. An order with an unknown product, an odd quantity or a price that doesn't match goes to your team with the original message and the extracted order side by side, so checking it takes seconds rather than minutes.

Do our customers have to change how they order?

No, and that's the point. Customers keep emailing, sending PDFs, attaching spreadsheets or leaving voicemails. Nothing changes on their side.

Does this replace our order processing team?

It changes what they spend the day on. Instead of typing every order, they check the exceptions, look after customers, and deal with the orders that genuinely need a person.

How do we start?

Send us up to 30 of your real orders for a free run, and we show you what the AI did with each one. If you like what you see, a guaranteed pilot on one order channel follows, usually over two to three weeks. If it misses the measures we agree in writing, you do not pay for it.

Schedule a demo

Bring us the process. We'll show you it handed over.

Tell us how the work arrives today and what has to be true before it's done. We'll tell you straight whether it fits the platform or needs a build.

Loading available times…

Open the calendar in a new tab
  • A conversation about your process, not a slide deck.
  • 30 minutes. We'll show the loop running on a process shaped like yours.
  • No obligation, and your details stay with us.
Prefer to write to us first?

Or email support@nuncily.com