Back to Blog
Published:
Last Updated:
Fresh Content
Legal AI in the ERPChapter 18

How does AI contract review work in Dynamics 365 Business Central?

9 min read
1,875 words
high priority
Ali Ahmed

Ali Ahmed

AI Solutions Engineer, Cognilium AI

TL;DR

Business Central holds the commitment, not the contract. AI review reads the document your ERP never sees, then checks the terms that did reach it.

Business Central holds the commitment, not the contract. It knows the vendor, the agreed price, the delivery terms and the dates. It has never read the agreement those came from.

AI contract review reads the document your ERP never sees — then checks whether the terms that reached Business Central are the terms you actually negotiated.

What does Business Central actually hold?

The operative half of the agreement, expressed as records a buyer acts on:

In Business CentralNot in Business Central
Vendor card, payment terms, delivery termsThe signed agreement itself
Agreed prices and line discountsLiability caps and indemnity
Purchase orders drawing on those termsTermination and notice provisions
Dates that drive receipt and invoicingData-protection commitments
Who approved what, and whenWhy any of it was agreed

Microsoft's own wording makes the split explicit. A purchase invoice exists to "record your agreement with a vendor to purchase products on certain delivery and payment terms"record the terms of the agreement, not hold the agreement. And a blanket purchase order exists to "manage your commitment to a vendor to purchase large quantities delivered in several shipments over time."

That left column is the contract doing its job. Every purchase order priced against a negotiated rate is the agreement operating, whether or not anyone has opened the PDF since signature.

The closest Business Central gets to pointing back at the paperwork is a reference number: on purchase documents you "can specify a document number that refers to the vendor's numbering system." A pointer, not a document — and it points at their filing, not yours.

And the right column never reaches the system. It stays in a file, in an inbox, on a drive — which is fine until somebody needs to know what the liability position was on a supplier who has just failed.

So what is the actual problem?

Nothing reconciles the left column against the document that produced it.

Microsoft's purchasing documentation describes how the terms are recorded. Across those purchasing and vendor pages it describes no check that what was recorded matches what was signed — and that is the correct division of labour, not an omission: the ERP is a system of record, and it records what it was given.

So the terms are typed in once, by a person, from an agreement they read at the time. After that the record operates independently:

  • The agreement is amended and the vendor card is not updated.
  • A price is renegotiated by email and never reaches the line.
  • A payment term is entered as thirty days because that is the default, and the contract said sixty.
  • The person who negotiated it leaves, and the reason behind an unusual term leaves with them.

None of that produces an error. The system is not wrong — it is faithfully executing what it was told, which stopped matching the agreement some time ago.

This is the gap AI contract review closes in an ERP, and it is a different job from reading contracts faster.

Where does review connect to Business Central?

Three points, and they are deliberately narrow:

① Before signature — read the draft against your playbook. Standard playbook-driven review: the clause is measured against your position, your fallback and your red line. Business Central is not involved yet.

② At entry — check what gets typed in. The review already knows the payment terms, delivery terms and prices the agreement contains. Those are the same fields a buyer is about to enter. Comparing the two at the moment of entry catches the typo, the default and the stale carry-over.

③ After the fact — reconcile the record against the document. Periodically, for agreements already live. This is the one nothing prompts you to do, and it is where the value sits — because terms drift silently and no error is ever raised.

Point ② is the cheapest and point ③ is the most valuable. Neither requires the ERP to store the contract, which is the design mistake to avoid: Business Central should hold the terms, and something else should hold the document.

If you are still deciding what to buy at all, the routes are native Business Central, an AppSource contract app, or an enterprise CLM platform — what each one actually solves.

On the enterprise tier the same three points exist against a richer object — a purchase agreement with a validity period, four commitment types and a policy that breaks the link when a price changes: how the two tiers compare.

Why does the Dynamics ecosystem write about billing instead?

Because that is what the existing extensions do — and it is worth understanding, because it explains what is missing.

Search contract management Business Central or contract management Dynamics 365 and page one is recurring billing, subscription management, renewal scheduling, contract entry — Rand Group, mhance, valprovia, ntribs, d-velop and MSDynamicsWorld, read 10 September 2026. Those are real problems and the apps that solve them are legitimate. They are all about the commercial mechanics of a contract you have already agreed.

Not one of those page-one results reads the agreement. They manage what happens after somebody typed the terms in — they do not check whether what was typed matches what was signed. That is a statement about two SERPs on one date, not about every extension on AppSource — but those two SERPs are what a buyer looking for this actually sees.

Microsoft has already shipped the read-extract-match pattern here — the Payables Agent pulls terms out of an inbound PDF and checks them against a purchase order: what it reads and what it skips. It is pointed at invoices, not agreements.

And it is not only the partners. Business Central itself ships vendor subscription contracts, and Microsoft's own description says what they are for: "A vendor contract maps liabilities to suppliers or manufacturers", whose lines "contain detailed information about the billing of the components of the contract." Microsoft's contract object on this tier is a billing object — which is the same split, stated by the vendor rather than inferred from a SERP.

*That is the difference between contract administration and contract intelligence:*

  • Administration — schedule the billing, track the renewal, raise the order. Well served.
  • Intelligence — what did we agree, is the record faithful to it, and where are we exposed? Not served on either of those SERPs.

What should you check first?

Before evaluating anything, run this on ten agreements. It costs an afternoon:

  1. Pick ten live vendor agreements with real spend behind them.
  2. Open the signed document and the vendor card side by side.
  3. Compare four fields: payment terms, delivery terms, agreed price, renewal or notice date.
  4. Count the mismatches.

The number you get is your answer. If it is zero, your process is disciplined and you do not need AI for this. If it is not zero — and it is rarely zero — you have measured your own exposure in an afternoon, and you know which of the three connection points above is worth building.

We would rather you ran that exercise than bought anything, because a team that has counted its own mismatches asks much better questions afterwards.

About Cognilium Cognilium builds AI optimization apps for Microsoft Dynamics 365 — companion apps that optimize the pricing, inventory, warehouse and planning decisions your ERP manages but can't optimize. Dynamics is your system of record. Cognilium is your system of intelligence. https://cognilium.ai · https://www.linkedin.com/company/37180269/

Legal AI Ops. We transform legal workflows with agentic AI, copilots, agentic workflows and decision intelligence — built into core workflows rather than beside them, to raise productivity and cut operational overhead. Contract Review Copilot is the contract-review app in that family. It ships as Paralegent AI, in production today. How we build Legal AI Ops — custom AI capabilities on top of legal work, against your playbook and your Dynamics 365.

Run the ten-agreement check and bring the mismatches to a 15-minute call. If there are none, we will tell you so and you will have lost fifteen minutes.

Sources

Sources and fact-check
#§ClaimTierSourceVerdict
11A purchase invoice records the agreement's delivery and payment terms; a blanket order manages the commitmentT1 — verbatim ×2purchasing-manage-purchasing (ms.date 2026-06-17), fetched 2026-09-10 — Quote to order for purchase invoices and Work With Blanket Purchase Orders rowsPASS — load-bearing
1b1External document numbers reference the vendor's numbering systemT1 — verbatimSame page, External document numbersPASS
4b4A BC vendor subscription contract "maps liabilities" and its lines carry "detailed information about the billing"T1 — verbatim ×2srb/working-with-contracts/vendor-contracts (ms.date 2026-05-04), fetched 2026-09-10PASS — upgrades §4 from a SERP inference to a Microsoft statement
1c1Business Central therefore holds terms, not the agreementT2 — ours, read off claims 1 and 1b rather than assertedPASS
22Terms are entered once and then operate independently; drift produces no errorT2 — ours, an operational argumentInternal definitionPASS
33The three connection pointsT2 — ours, our own design positionParalegant_TECHNICAL_PROFILE.md §1, §11PASS
43The ERP should hold terms, a separate system the documentT2 — ours, consistent with ch3Internal definitionPASS
54The Dynamics ecosystem's contract content is billing, subscriptions, renewals and entry — not agreement readingT2 — ours, and it is EVIDENCED: a SERP read on 2026-09-10 returned Rand Group, mhance, valprovia, ntribs, d-velop, MSDynamicsWorld and an AppSource listing, all billing/administrationlegal-doc/THE-CLAIM.md §3.2PASS — load-bearing
64Administration vs intelligence as the distinctionT2 — oursInternal definitionPASS
75The ten-agreement checkT2 — ours, a method. No claim about what it will find — "rarely zero" is stated as our expectation, not a measurementInternal definitionPASS

Tier summary: 4 × T1 (all verbatim), 8 × T2 — 0 × T4.

Source defect found and closed during writing. purchasing-manage-purchasing was listed in sources: before it had been opened — the exact failure the WhatsApp retraction taught. It was then fetched (2026-09-10, `ms.date` 2026-06-17) and three verbatim quotations now carry §1. Recorded rather than quietly fixed.

No figures. No mismatch rate, no time saved, no customer. "Rarely zero" in §5 is our expectation and row 7 says so.

Disclosure: the CTA offers to tell the reader they have no problem.

Share this article

Ali Ahmed

Ali Ahmed

AI Solutions Engineer, Cognilium AI

Ali Ahmed is an AI Solutions Engineer at Cognilium AI.

Applied AI AgentsAgentic SystemsRetrieval-Augmented Generation (RAG)LLM Product Engineering
Next in this series
How do contracts work in Business Central vs Finance & Supply Chain Management?
Chapter 19 · 10 min
In short

Key takeaways

  • Business Central holds the commitment, not the contract — vendor, prices, terms and dates, but never the agreement they came from.
  • The terms are typed in once and then operate independently, so amendments, renegotiations and defaults silently pull the record away from the document.
  • Three connection points: read the draft before signature, check the fields at entry, and reconcile the record afterwards. The last is the most valuable and the one nothing prompts you to do.
  • The ERP should hold the terms; something else should hold the document. Storing contracts in the ERP is the design mistake.
  • Search results for Dynamics contract management are administration, not intelligence — billing, renewals and entry are well covered; "is the record faithful to what we signed" is not answered there.
  • Count your own mismatches on ten agreements first. It costs an afternoon and it answers the question better than any demo.
What goes wrong

Common mistakes to avoid

  • Treating the ERP as a contract repository. It holds the terms that act, not the document that explains them.
  • Assuming entry was correct because nothing errored. A wrong payment term is a valid payment term.
  • Buying contract administration when the problem is intelligence. Billing apps do not read agreements.
  • Reconciling once at go-live. Drift is continuous, so the check has to be too.

Frequently Asked Questions

Find answers to common questions about the topics covered in this article.

Still have questions?

Get in touch with our team for personalized assistance.

Contact Us

Still have a question this did not answer?

The person who wrote this article answers these. Describe your setup and what you are stuck on — you will get a straight answer, including where we think the approach is wrong.