What eight systems can tell an agent, and what they cannot.

For commerce, nearly all of them. Every row except Salesforce CRM can tell an agent what is on hand right now, and every row can take a real order in the system of record rather than handing back a checkout link. That half of the problem is close to finished.

For supply chain automation, five of the ten. Answering "when does more arrive" from a documented API is possible on Odoo, NetSuite, SAP S/4HANA, Oracle Fusion and Dynamics 365 Finance and Operations. On the other five the data is partial or absent, so a job that reacts to resupply has nothing to read.

The catch is that four of those five need an IT ticket, a plan upgrade or a security review before an agent gets near them. NetSuite is the exception. That is the whole shape of the market right now, and the sections below are why.

The systems an agent can reach fastest hold the least. Shopify and WooCommerce connect through a single approval screen, and neither carries a supplier, a purchase order or a lead time anywhere in its documented schema. SAP S/4HANA carries the richest supplier and net-price data of anything we read, and getting access to it is a committee decision measured in weeks.

So "can an agent tell me when this is back in stock" gets answered no in two different ways, and the difference decides what you do about it. On a storefront the data does not exist. On an ERP it exists and the agent cannot reach it.

The pattern is a tendency, not a law, and the exceptions are the useful part. Odoo has the best resupply data of any mid-market system and puts external API access behind a pricing plan. Dynamics 365 Business Central onboards in about half an hour with no partner and no Azure work, and answers resupply only partially. NetSuite sits close behind on both counts, with machine-to-machine credentials that never need re-consenting. Those three are where the tradeoff is smallest.

Each cell is scored Excellent, Good, Partial or Missing against what the vendor documents, not against what a partner app can add on top.

One person read the vendor documentation, pinned to a version: the Shopify Admin API 2026-07, the Odoo 19.0 source tree, learn.microsoft.com for Business Central and Finance and Operations, docs.oracle.com for NetSuite and Fusion, and the SAP Business Accelerator Hub. Where a claim could not be confirmed in the vendor's own words it is not on this page.

A cell moves when a vendor ships an API, not when a vendor announces a direction. Every row carries the date it was last read, and the whole table is re-read quarterly. If you run one of these systems and a cell is wrong, tell us and it changes with your name on the correction.

It scores documented APIs, so it understates any system where an extension can close the gap. ATUM gives WooCommerce a real supplier and purchase order API. One AL extension closes all three Business Central gaps at once. Neither is what a merchant has by default, which is the thing an agent has to plan for.

It also says nothing about licensing cost, and licensing decides more here than capability does. Oracle Global Order Promising is the best promised-date answer we found and a separate line item.

And it cannot score a stack that is one of a kind. No two in-house systems are the same, which is why the last row of the table is not a score.