Building the demo is a weekend. Running it is the product.
Most BC partners we talk to have considered building their own AI reviewer. The instinct is right. The hard part is just not where it looks like it is.
The weekend version works. That is the trap.
Prompt, diff, model, and the first demo impresses. It writes confident, plausible review comments on real pull requests. What it cannot do is tell you which of its comments are wrong:
- Fabricated compile errors read exactly like real ones. Without the AL compiler in the loop, every one costs a developer the time to disprove it.
- The model guesses at symbols. A diff does not contain the base app, system app or platform, so the reviewer invents what it cannot resolve.
- Someone owns it forever. Prompts drift, models get deprecated, false-positive complaints land on one engineer's desk, and a billable consultant becomes your internal tooling team.
What the product actually is
- It verifies before it posts. Every compile error PEQY claims is checked against the actual AL compiler before you see it. Findings that cannot be grounded do not reach your developers.
- The Business Central symbol model, resolved. Base app, system app and platform layers are loaded and queried, and the reviewer knows what your code calls into.
- Maintenance is our job. Model migrations, eval suites, false-positive tuning and new BC releases are what we do full-time, for EUR 99 per developer per month instead of an engineer's side project.
Frequently asked questions
Why not just build our own AL reviewer?
You can. The demo takes a weekend: prompt, diff, model, done. The product is everything after: knowing when it is lying to you. Compile verification, symbol resolution and trust-gated findings are what turn plausible output into findings you can act on, and they are maintained by us instead of by someone on your team.
What does PEQY do that a prompt plus a diff cannot?
It verifies. Every compile error PEQY claims is checked against the actual AL compiler before you see it, and findings are resolved against the Business Central base app, system app and platform symbols instead of guessed from the diff. A prompt cannot tell you which of its findings are wrong.
We already built an internal reviewer. Should we switch?
Run both on the same past pull requests and compare. That is the honest test, and it takes an afternoon. Then ask who on your team owns the prompts, the false-positive rate, the evals and the next model migration, and what that time costs against EUR 99 per developer per month.
What does PEQY cost?
EUR 99 per developer per month, with a 14-day free trial. Enterprise plans with custom terms are available. Pricing is listed openly on the pricing page.
The honest test takes an afternoon
Run PEQY on your past pull requests (next to your prototype, if you built one) and judge the findings yourself.