Resources · Research
Research and Evaluation
The benchmark, the finance assessment and the document test. Three ways to see where you stand before you buy anything.
What this section is for
Evidence before a sales conversation
Every tool here returns something you can show your own team the same day: a score, a breakdown or a set of extracted fields.
Most automation evaluations start with a vendor demo on the vendor's sample documents. That tells you the product works on paperwork chosen to make it look good. The tools on this page work the other way round. They start from how your work runs today and from a document you actually received, so that when you do ask for a demo, you already know which process, which documents and which controls you want to see.
There are two kinds of tool, and it helps to keep them apart. The Automation Benchmark is a method: seven dimensions, weighted to add up to 100, with four bands. The finance workflow assessment is where you take it for payables; for any other process, book a call. The DocPro document test is different again. It asks nothing about your process and reads one file instead.
None of them replaces a pilot on a month of your own documents. They tell you whether a pilot is worth setting up, and what it should measure.
The section
Seven pages, four of them tools
The first four give you something back. The last three are reading, and two of them are honest about being empty for now.
Automation Benchmark
The PTAS AI 2026 Automation Benchmark: what it measures, how the seven weights add up to 100, and what each of the four bands means. Read this to understand the score.
Assess Your Finance Workflow
The finance track. The same scorer pointed at accounts payable: intake, matching, approvals, exceptions, vendor data, GST and posting.
Test Your Document
Upload an invoice or a UPI payment receipt and see the fields DocPro reads from it, beside the page. No account and no contact details.
Insights
The blog, sorted by the questions buyers bring: agents, payables, documents, GST and control.
Whitepapers
Nothing published yet. The page says what is in preparation and why we have not padded it.
Webinars
No recorded sessions yet. A private working session on your own documents is available now.
Which tool, when
Start from the question you have
Each tool answers one question well and the others badly. Pick by the question, not by the tool that sounds most thorough.
Four ways to evaluate, and what each one hands back
The benchmark is the method behind both assessments. Read it once, then take whichever track fits the process you own.
Answers and documents
Why the assessment and the document test are both worth doing
They look at different things. A process can score well and still run on paperwork nobody can read, and the reverse happens just as often.
Two kinds of evidence · Illustrative example
Illustrative data. A blank total is a useful result: the agent did not settle the value, and it did not invent one.
The assessment hears how work moves: who keys what, how clean the master data is, how a result would reach the ERP. It cannot see a single document. The document test sees one document closely and knows nothing about your process. Run both and you can tell a process problem from a paperwork problem, which decides whether the first project is a DocPro deployment, a FinSuite workflow, or a clean-up of the vendor master before either.
The public test reads invoices and UPI payment receipts only, one page at a time, and does not display confidence scores or run validation. Those depend on thresholds and masters set for your process, so they are shown in a demo. Bank statements, purchase orders and KYC documents have their own agents and are shown the same way.
How we handle research
What we label, what we hold back, and what we keep apart
On every page that scores
The method, labelled as ours
The benchmark page and both assessments say the same thing in so many words: PTAS AI methodology, not an industry-standard benchmark, and not scientifically validated.
Not on this site yet
Aggregate results
We do not publish averages, peer comparisons or industry figures from assessment responses. If we publish benchmark research later, it will state its sample and method.
Kept separate
Research and marketing consent
The assessment asks for each separately and both are optional. Only responses with the research box ticked may be included in aggregated research.
Keep reading
Elsewhere in Resources
Guides
Six long reads on agentic AI versus RPA, AP automation, IDP, GST, human review and readiness.
Read more →FAQ
Short answers on the products, control, integrations, deployment and how to evaluate.
Read more →Glossary
The terms the tools and guides use, from approval band to three-way match.
Read more →Resources
Everything a buyer can read, test or evaluate, in one place.
Read more →Start with the score
Get one number to argue about, in seven minutes
The assessment scores seven dimensions and names the two places we would start. Take it before the document test, and both before a demo.