How big companies convert hundreds of .NET apps with AI, and who finishes the last stretch

Published 25 October 2026 · dotNewLABS

AI tools from Amazon and Microsoft now convert old .NET apps by the hundred. The public stories from the companies using them share one shape: the tool does most of the code, and then people finish the rest. Here is what they say, and what it means for your team.

Numbers are reported by the vendor or the customer and have not been independently checked.

What the public stories say

Company	What they did	What people still did
Thomson Reuters vendor-reported AWS re:Invent, Dec 2025 (video)	Running AWS Transform on 400+ .NET apps (500 million+ lines of code), about 10 apps at a time.	Each app came out 50 to 90% done. The team described the results as "never fully successful" and had engineers fix the rest.
Bio-Rad vendor-reported AWS blog, Sep 2026	About 70% of the work automated. Each app took 2 to 3 months instead of 9 to 10.	Business-logic fixes, integration testing and manual review to confirm each app behaves the same.
Experian vendor-reported AWS customer roundup, May 2025	7 apps (687,000 lines of code), with 80% of the code converted automatically.	The remaining 20% of the code.

Microsoft's own pages for GitHub Copilot app modernization mostly carry short customer quotes, for example a dozen apps upgraded in hours. We did not find a public Microsoft story covering hundreds of apps. One trade report says developers spent many hours fixing upgrades that only partly worked.

The shape is always the same

The tool converts

Most of the code moves over fast, in batches. This is where AWS and Microsoft are strong.

People finish

A large in-house team, or a big partner, fixes the leftovers: business rules, integrations, odd corners.

Someone proves it

Each app is tested and signed off before cutover. The stories say "manual review" and do not describe how.

None of this is a criticism of the tools. A tool that gets an app most of the way in hours is a real gain. The leftover work is normal, and it is where the time and the risk now sit.

What this means for a smaller team

Thomson Reuters and Bio-Rad have engineers to spare and direct access to the vendor's programs. A mid-size clinic group, bank or credit union usually does not. Their apps are fewer, but the leftover work is just as real per app, and it competes with everything else on the roadmap.

Use the tool. For many apps, Copilot or AWS Transform is the right first move, and we use them where they fit.
Plan for the last stretch. Budget people and time for the leftovers, not just the conversion.
Ask how each app will be proven. "We reviewed it" is not evidence an auditor or a clinic manager can read.

What we do

We finish and prove it. For each app we write down what it does before changing it, run old and new side by side on your own examples, and run the cutover with a way back. AI Ready Studio, our own tool for this, is in early access, and a person reviews every step. If a vendor tool does the conversion, we take it from there.

Questions to ask before you start

What share of each app does the tool finish, and who finishes the rest?
How will we know each app behaves the same as before, and who signs off?
What happens to data, logins, reports and scheduled jobs, which sit outside the code?
Can we see this on one small piece of our own app first?

Not sure where your apps stand?

Book a free 20-minute call about one app or data move: what it runs on, what's risky, and the shortest safe path forward. You keep a short written summary whether or not you hire us.

Book a free call

Or email contact@dotnewlabs.ai