Best AI for CFOs

Research notes on AI in finance work, by task

Index
Appendix A
Built
4 November 2026
Volume
1
This page
No exhibits
  1. Best AI for CFOs
  2. Methodology

AAppendix, A to E

Methodology

How a product gets into a note, what we quote, and what we never do.

A product is included when its maker's own public pages say that it applies AI, machine learning or a language model to a finance task covered by one of the twelve use cases. That is the only test. No vendor paid, asked or was asked to be listed, and no vendor can pay to be described, placed or ordered.

Inclusion

We started from vendors that describe AI features for finance teams on their own sites, then read each vendor's home page and the page it links for its AI product. Vendors whose pages refused automated requests (a 403 or 429 response, or a robots.txt rule) were left out rather than described from memory; so were vendors whose pages, as read, did not say what the AI does. Records describe 93 vendors, read between 29 October 2026 and 3 November 2026. Still left out because their pages refused our requests: SAP, Anrok, Payhawk, Prophix. Still left out because the pages read did not say what the AI does: Microsoft (Copilot for Finance). A vendor appears once on the site, with one record listing all of its AI product lines that we read.

What counts as a source

Product facts come only from the maker's own domain and from SEC EDGAR. What a maker says its AI does, the data and systems it names, where a person reviews its output, how data is handled and who owns the maker are verbatim excerpts of 40 words or fewer, in quotation marks, linked to the page and dated with the day we read it. Three things are ours, written in neutral words from the same pages, and labelled as ours wherever they appear: the one or two sentences that summarise each product (“Our summary”), the use cases a record is filed under, and the short list of data a product names that the use-case tables print. A field the maker's pages do not address says “not stated on the pages read”; we do not fill it in.

A listed vendor whose pages refuse automated requests can instead be described from its latest annual report on EDGAR (Form 10-K, 20-F or 40-F), when the report says what its AI does for finance work. The record then quotes the filing, names the form and the date it was filed, and says that the product pages refused automated reads. No record is made this way now. When the maker's own pages can later be read, the record quotes them and keeps the filing excerpts that still hold, each linked to its source: Workiva.

Some makers' sites refuse simple automated requests (an HTTP 403 or 429, or a browser challenge) while their robots.txt allows the pages. Those pages were read through a rendering fetcher, which loads a page the way a browser does, one request per page; a page that robots.txt disallows was not fetched by any route. Every excerpt was checked by script against the text as fetched, the date read is the day of that fetch, and the link is the address actually fetched. 9 records quote pages read this way: Coupa, Drivetrain, Eftsure, Kyriba, NetSuite, Oracle, Sage, Trustpair, Workiva. The daily recheck cannot re-read these pages, so their read dates stay until a later read succeeds.

Public-company status comes from the SEC's company ticker file. “Listed” links the company's EDGAR filings; “not listed” means only that the maker is not in that file.

The use-case notes cite regulators and standard setters (the SEC, NIST, the Internal Revenue Service, FinCEN, the FBI, the Federal Reserve) and the COSO framework where they have published on the work. The PCAOB and FASB sites refused our automated requests when we read sources for this edition, so their standards are not quoted here.

Alphabetical order, no ranking

Every list and table is in alphabetical order by maker. There are no scores, ratings, stars, ranks, picks, awards or superlative claims, and no affiliate links or sponsored placements. The site's name is a name, not a verdict. Prices appear only when a maker's own pricing page states one and we can cite the date we read it; this edition quotes none.

How filings were searched

We searched the SEC's EDGAR full-text search for 10-K, 10-Q and 8-K filings made between June 2025 and October 2026 that mention AI next to finance work (the close, reconciliations, invoices, collections, forecasting, tax, treasury and similar), then read the matching passages. 19 excerpts are quoted. Two kinds are kept apart on every page: companies describing AI in their own finance work, and listed finance-software companies describing their AI products to investors. A filing is never taken to mean that the filer uses a product on this site unless the filing names it.

Dates

Each page shows the day it was first published here; that date is recorded once and never moved back. Each excerpt shows the day its page was read. A daily routine re-reads a rotating set of vendor pages and proposes an update when an excerpt no longer appears on its page.

Corrections

If a record is wrong, out of date or quotes a page that has changed, write to corrections@bestaiforcfos.com with the page address. Vendors may point us to a page on their own domain; we quote it if it says what the record needs. Corrections are made on the page and dated.

About the site. Products A to Z.