Power Platform's original bargain, faster delivery by non-developers in exchange for per-seat rent, no longer clears for greenfield internal tooling. The rent rose while the addressable surface stayed fixed, and AI-assisted development now reaches artifacts the platform cannot produce at any price. The exception, real and not small, is where Dataverse or Dynamics 365 is the system of record.
Low-code was sold on a trade: business technologists ship working applications without waiting for scarce, expensive developers, and in exchange the organization pays per-seat rent to the platform. That was defensible in 2016. The usual rebuttal argues about price. This assessment argues about surface. The first-order question is not whether the platform is cheaper, it is whether the platform can make the thing at all. Price is a second-order argument when the first-order answer is that the tool cannot build it.
Six findings, each sourced. Four things changed since 2016, and three of them moved against the platform.
Microsoft's public Power Apps pricing page now lists Premium at $20.00 per user per month and a 2,000-seat rate of $12.00. The $5 per-app plan that was the cheap way in no longer appears on it, and the metered alternative runs at roughly twice the old per-app rate.
Seeded AI Builder credits inside Power Apps Premium, Power Automate Premium and Dynamics 365 licenses are removed on 1 November 2026 for all customers, including Enterprise Agreement holders, with no conversion to Copilot Credits. Features already running in production stop unless new capacity is purchased.
Of nine products shipped by a non-developer, seven cannot be produced by Power Platform at any price and two require significant workaround. Signed desktop applications, self-hosted container services and local-only data processing sit outside the platform's addressable surface entirely.
No controlled study compares building the same internal application in a canvas designer against building it with a coding agent. The quantified claims in circulation come from research the platform vendor commissioned. When neither side has a measurement, the burden falls on the party charging rent.
Independent and vendor research finds refactoring in sharp decline and security defects in a large minority of generated code. That is a genuine liability. It is also a discipline problem, addressed by process harnesses and shared skills, not by a per-seat license.
The Agent Skills format was published as an open standard in December 2025 and donated to a Linux Foundation-backed body, with Microsoft, GitHub, OpenAI and others adopting it. Encoded process now lives outside a licensed platform in a way it did not two years ago.
Read the full 17-page assessment for the three-year cost comparison, the carve-out where the platform premium still survives, and what would move the call.
This isn't a vendor summary. Every sentence is labeled by what stands behind it: verified fact, vendor claim, third-party estimate, my assessment, hypothesis, or scenario. Sources are numbered and clickable. Forward-looking sections use scenarios with observable tripwires, not forecasts. It's the same method behind every market assessment I write.
Seventeen pages, built from public sources with no client brief and no interviews. Read it in the browser or take the PDF.
Each report here answers a real question, directed and researched against public sources and evaluated against a stated assumption, then delivered as Word and PDF. If you're weighing a platform, sizing a category, or defending a number to a board, tell me the decision behind it and I'll tell you honestly whether a report is the right tool.
Commission an assessment