Operator field note refined 2026-10-07. This is a decision method, not a product-wide survey.
When a contractor asks whether their software is “AI-ready,” they usually mean “does it have an API.” We used to think so too. In implementation work, we compare the records available through each permitted route before building a financial view across systems. The lesson was humbling: the presence of an API alone cannot establish whether it supplies the records for the intended decision.
Take one large construction ERP. Its REST API is documented and modern-looking. In practice, under a real customer’s credentials, endpoint access and field coverage differed from the available database view. That tenant observation did not establish a limitation of every installation or the current product. The useful test was to compare the required due-date and payment fields with the native report, then choose the permitted route that supplied them.
A different failure can occur in the field that links jobs to invoices. An invoice response may contain the financial amount without the job identifier used by the operating system. Test the current edition, enabled fields and export before choosing a bridge. A successful request with the wrong identifier coverage is still insufficient for a job-level margin calculation.
And then there are the systems with no pull API at all. For a browser-backed report, retain the date window, filters and file rather than infer coverage from a completed download. Adjacent windows must have compatible definitions; repeated headers and subtotal rows must be identified before combining them. Whether another route is available depends on the current product and account.
So the real ingestion picture, across the systems we work with, is a patchwork: API where it is genuinely good, direct database reads where the API is hollow, scheduled report or CSV delivery where the API hides fields or costs too much, and browser automation where there is no API at all. A serious data layer uses all four and reconciles between them. That is the unglamorous work nobody demos.
How common is this? We do not have a measured industry denominator for unavailable APIs or unusable exports. The software index records cited interfaces and unknowns tool by tool. One difficult tenant is a reason to test the route, not a basis for a category-wide percentage.
This is why we score a system’s AI-readiness by a different question than most buyers ask. Not “does it have an API,” and definitely not “does it have an AI feature.” The question is: how cleanly can a practitioner get complete, trustworthy data out of it, by any method? Answer that, and you know whether AI can ever do anything useful on top.
Use this in an implementation review
- Choose the required record and field population before choosing API, export or browser access.
- Compare a permitted extraction with the native report, including date filters, totals and identifiers.
- Retain the original export and distinguish detail, repeated headers and subtotal rows.
Specific systems to inspect
Operator analysis informed by scoped implementation workflows. The checks explain how to evaluate a financial decision; they are not an industry survey, a capability certification or a claim that every account behaves alike. See the full software index.