AI tools from Amazon and Microsoft convert old .NET apps quickly, and they are good at it. But an app that builds and starts is not yet an app that behaves the same as the one it replaces. That second part is the proof.
The dashed step is the one we could not find described in the public stories below.
What the public stories report
Thomson Reutersvendor-reported put 400+ .NET apps through AWS Transform, about 10 at a time. Each came out 50 to 90% done, and engineers fixed the rest. (AWS re:Invent 2025, written summary)
Bio-Radvendor-reported automated about 70% of the work, and people did a manual review to confirm each app is functionally equivalent. (AWS blog, 2026)
Both stories say people checked the result by hand. Neither describes how: what was compared, on which examples, and what the sign-off looked like. We looked for a public account of that method and did not find one.
Why "it builds" is not proof
A converted app can compile, start and look right while a number on a report is off, a date shows differently, or a permission no longer applies. Those are the kinds of differences that surface after go-live, when a clinic manager or a finance team notices. Proof is how you find them first.
What proof can look like
Pick what matters. The screens and reports your people rely on, plus the examples they know the answers to.
Run both. The same inputs on the old app and the new one.
Compare three things. What appears on screen, the numbers on the reports, and what changed in the database.
Write it down. A plain report of what was compared and what differed, and who signed off.
Proof of this kind covers the screens and reports you pick. It is not a promise that every possible path through the app is identical, and it should not claim to be. What it gives you is evidence on the parts that matter most, before cutover.
Using the tools, not replacing them
Copilot and AWS Transform do real work, and we use them where they fit. The proof is the part that goes around them: the same checks whichever tool did the conversion. AI Ready Studio, our own tool, is in early access, and a person reviews every step.
Questions to ask before you convert
Which screens and reports will be compared old against new, and who chose them?
What will we receive at the end that a business owner can read?