IDC’s research VP for software development said it plainly this summer: low-code and no-code platforms have become the primary way companies build and run AI agents at scale. Not a growing trend. Not an emerging pattern. The primary way.
Microsoft Power Platform now has 56 million monthly active users. That number climbed 27 percent in a single year. Gartner projects that by 2029, more than 80 percent of all new business applications will be built on AI-powered low-code and no-code platforms. In 2024 that number was 20 percent.
If you are still treating AI agent deployment as an engineering project, the rest of the market has already moved past you.
The Results Are Coming From People You Would Not Expect
The companies posting the strongest AI numbers right now share a pattern that has nothing to do with technical sophistication. Their agents were built by the people who do the work, not by the people who write the code.
Navien, a global HVAC manufacturer, saved 28,000 hours annually through AI agents and automation. Thirty-two percent of their employees now use self-service analytics to make decisions without calling IT. Nobody hired a machine learning team. The people closest to the problems built the solutions.
Then there is Grupo Bimbo — one of the largest baking companies in the world — where two AI agents got built during an internal hackathon using Microsoft Copilot Studio. Planning-phase audit time dropped 20 percent. Risk and control matrix creation went from days to seconds. Two agents, one hackathon, and the results showed up before the quarter ended.
EY tells a different story in scale but the same story in direction. They deployed Microsoft 365 Copilot to 150,000 employees, saw a 15 percent productivity gain, and are now expanding across more than 400,000 people globally. The numbers are hard to ignore: 95 percent faster lead times, 37 percent lower finance operating costs, up to 90 percent less manual work across core business processes.
These are not pilot results. These are production systems that employees built and run themselves.
The Complexity Assumption Is the Bottleneck
Most organizations I work with still assume AI agent deployment requires a dedicated technical team, a custom integration layer, and months of development time. That assumption was reasonable two years ago. It is not reasonable now.
The tooling has caught up. Copilot Studio, Power Platform, and their equivalents across the market have reduced the build-to-deploy cycle to days for many use cases. The orchestration layer that used to require custom code is now drag-and-drop. The governance that used to require a separate infrastructure investment is now built into the platform.
What has not caught up is the mental model most leaders carry. The belief that AI agents are fundamentally an engineering deliverable prevents the people who understand the workflows from building the solutions. And it pushes the work to people who understand the technology but not the process it needs to change.
That mismatch is behind most of the pilot failures that keep showing up in every enterprise AI survey. The agent works technically. It does not work operationally. Because the person who built it never did the job it was supposed to improve.
What This Looks Like on Monday
If you run a team and you have not explored what your people can build with the low-code agent platforms already available in your enterprise license, you are sitting on unused capacity.
The first step is not a strategy document. It is a test. Pick the most tedious, repetitive process your team complains about. Find the person who knows that process best. Give them access to whatever low-code agent builder your organization already pays for. Set a one-week deadline. See what they build.
Atos did a version of this at scale. They now operate 19,000 AI agents through a unified operating model built on Copilot Studio and Power Platform. They reduced incident triage time by 68 percent, cut laptop compromise investigations from 25 minutes to eight, and save 337 hours per week on investigations. Their security operations staff did not shrink. Twenty percent of them moved from incident response to governance, risk, and compliance work.
The common thread is not the platform. It is who builds. Every one of these organizations stopped treating agent deployment as a technology project and handed it to the people who run the operations.
The Spending Tells the Story
Gartner projects worldwide spending on low-code development technologies will reach $58.2 billion by 2029. That number is not driven by hype. It is driven by the fact that the model works. Companies that give simple tools to the right people get measurable results. Companies that give complex tools to the wrong people get proof-of-concept presentations.
Every quarter that gap gets wider. And the variable is not budget, not model capability, not technical talent. It is whether you trust the people doing the work to build the tools that change it.
Fifty-six million people already made that call. If your organization has not, the window is closing.