Most finance teams I speak to in the UAE are busy with e-invoicing right now. Picking a service provider, mapping fields, testing with their ERP. A fair question keeps coming up in those conversations: if invoices are about to arrive as structured data, do we still need AI in accounts payable at all?

My answer is yes, but not where most people expect. E-invoicing removes a lot of the work AI was being sold for. It also leaves a set of problems that rules alone handle badly. Here is how I would split them.

Where the dates stand

These are from the Ministry of Finance and the Federal Tax Authority, as of the time I am writing this. Check the official pages before you plan around them, because they have changed once already.

  • The pilot and voluntary use started on 1 July 2026.
  • Businesses with annual revenue of AED 50 million or more must appoint an Accredited Service Provider by 30 October 2026 and go live by 1 January 2027. The provider deadline was moved from 31 July 2026. The go-live date was not.
  • Businesses below AED 50 million must appoint a provider by 31 March 2027 and go live by 1 July 2027.
  • In-scope government entities must appoint a provider by 31 March 2027 and go live by 1 October 2027.

The system covers business-to-business and business-to-government invoices, and it applies to the buyer as well as the seller. If you receive invoices, you need a provider too.

What e-invoicing takes off the table

Reading a PDF invoice is the classic AI demo in finance. Pull out the supplier name, the TRN, the line items, the VAT and the total. It is also the part that e-invoicing largely replaces. Once a supplier is live, the invoice reaches you as structured data with the required fields already in place.

For those invoices, you don't need a model to guess what the total is. You need plain rules: check the fields, match the TRN to your supplier master, check the VAT maths, and post. Rules are cheaper, faster and easier to audit than any model. If a vendor is still pitching you AI to read invoices that already arrive as data, ask what it does that a validation rule doesn't.

Where AI still helps

The structure tells you what the supplier billed. It doesn't tell you whether you should pay it. That gap is where most of the real effort in accounts payable sits, and where AI earns its place.

Matching to what you actually ordered. An invoice line says "maintenance services, Q3". Your purchase order says "HVAC AMC, Tower B". A goods received note covers half of it. Rules break down on this kind of fuzzy match. A model that reads the invoice, the order and the receipt together, and explains why it thinks they match, can clear a lot of these and send the rest to a person with the reason attached.

Coding and cost centres. Structured data gives you clean line items. It doesn't pick the right GL account, cost centre or project for a line that says "consulting". Teams usually do this from memory and old invoices. A model can learn from your past coding and suggest the line, with a person approving anything below a confidence level you set.

Exceptions and supplier emails. Price differences, partial deliveries, credit notes that don't line up, a supplier asking why they haven't been paid. These arrive as emails and attachments in Arabic and English. Sorting them, pulling out what is being asked and drafting a reply is language work, which is what these models are good at.

The invoices that still come the old way. The rollout is phased. Smaller suppliers have until mid 2027. Until then, and for anything outside the system, some invoices will still arrive as PDFs and scans. For a lot of companies that will be a shrinking share of volume, not zero.

Where plain rules are better

I would not use AI for any of these:

  • Checking that required fields are present and in the right format.
  • VAT calculations and totals.
  • Matching an invoice to a purchase order when the numbers and references line up exactly.
  • Duplicate checks on invoice number, supplier and amount.
  • Approval routing by amount and department.

If the answer is always the same given the same input, write a rule. You will get the same answer every time, you can show an auditor exactly why, and it costs almost nothing to run. AI belongs where the input is messy and the answer needs judgement.

Do it in the right order

My advice to finance teams is simple. Get e-invoicing live first. It is a legal deadline and a defined project. Don't let an AI project slow it down.

Then look at what is left. Once structured invoices are flowing, you can see clearly which exceptions eat your team's week: the fuzzy matches, the coding questions, the supplier emails. That is a much better starting point for AI than a PDF reader you are about to stop needing.

When we run a pilot on accounts payable, that is how we start. We measure how long the exceptions take today, build tests from your real invoices and emails, and try more than one model on them. Sometimes the answer is that a few extra rules in your ERP solve most of it, and we say so.

If your e-invoicing project is moving and you want to know what to do with the exceptions afterwards, start a chat with us on the site, or book a call through the Applied Labs page. Bring a month of exceptions and we'll tell you which ones are worth automating.