More Employees Are Building Mobile Apps. IT Is Picking Up the Cost.

How businesses build mobile apps has changed faster than the policies meant to govern it. More employees than ever are now shipping their own internal tools, and the IT departments asked to support them are catching the consequences the rest of the company hasn’t noticed yet.

By 2023, PwC had enrolled 2,000 employees in a citizen developer program that started as a 500-person experiment. the HBR argument that anyone can become a programmer sat behind that number. The cleanup work sat behind a different one, and the people doing it have, so far, mostly been left out of the trend pieces.

The Apps That Weren’t Supposed to Exist

PwC’s program, called the Digital Accelerators, started in 2017 and now trains employees to build the kind of small process and automation tools a software team used to handle. The products and technology group that ran the launch expected 500 volunteers. It got 2,000, and 65,000 PwC employees have been promised AI skills training as part of the rollout.

The scale matters because it stops looking like a side project. AT&T, ING, Johnson & Johnson, and Deloitte have all run similar programs, and the largest low-code vendor now reports tens of millions of users a month. The apps that come out of these programs live inside the same networks as the apps IT signed off on, and IT often doesn’t know they exist until something breaks.

An organization might end up with as many systems as employees. It might become dependent on citizen-developed systems known to only a few employees, or whose developers have long since left the company. This explosion of “gray IT” across the enterprise and the cost of reworking broken technical systems that would ensue is an important issue.

That passage comes from Tom Davenport, a Babson College professor and Deloitte adviser, and Kerem Tomak, the former global chief analytics officer at ING, writing in a 2023 Harvard Business Review piece titled “We’re All Programmers Now.” The piece made the case that the trend was real and worth managing. It also made the case that somebody has to manage it, and the person who does that work is rarely the one quoted in the press release.

Where the Cleanup Crew Lives

Every internal app that doesn’t go through IT is shadow IT, and what shadow IT now costs IT departments is mostly measured in unreimbursed hours. Cisco has reported that 80% of company employees use shadow IT, and an IBM Institute for Business Value study cited by IBM found that 41% of employees acquired, modified, or created technology without their IT or information security team’s knowledge.

According to Gartner, as cited on IBM’s shadow IT topic page, 38% of technology purchases are managed, defined, and controlled by business leaders rather than IT. The IT team still gets the page when the SaaS subscription lapses, the integration breaks, or the audit comes calling. The HBR piece on citizen developers included a European telco that automated a testing loop into production and shipped free iPhones to customers until someone noticed.

The work IT departments now absorb in the shadow-IT era shows up in three places that didn’t used to exist as separate jobs:

• Discovering unsanctioned apps and figuring out which ones hold sensitive data, a task that used to be a once-a-year inventory and is now closer to continuous monitoring.
• Patching and integrating what citizen developers ship, because the apps that pass the smoke test on day one still need security review, identity integration, and updates the original builder never planned for.
• Recovering knowledge when the employee who built a tool leaves for another company, the failure mode HBR flagged as the most expensive part of the trend.

The vendor side has started packaging products for that last job. Cognizant’s Neuro AI Trust, which gives enterprises real-time governance over AI models and agents, sits in the same category as the broader market for AI guardrails now showing up in vendor catalogs. The real-time governance vendors now sell to IT are a direct response to the cleanup work, and the category did not meaningfully exist five years ago.

Low-Code Hit Its Numbers Anyway

The trend that produced the cleanup work has its own receipts. Microsoft’s Power Platform business reported 56 million monthly active users in the company’s April 30, 2025 fiscal third-quarter earnings call, up 27% from the prior year. That figure is the cleanest vendor-disclosed number on the scale of low-code building, and it comes from a primary source rather than an analyst estimate.

Rabobank, the Dutch bank, is one of the customers Microsoft has named publicly. The bank adopted Power Platform as part of a broader digitization strategy covering 40,000 employees in 38 countries, and today runs more than 2,500 Power Apps and Power Automate solutions. One of those solutions supports the company’s re-org process, and the time it takes to run went from three weeks to three minutes. Another handles customer support calls, with 40% to 50% of calls automated, a volume Microsoft pegs at 60,000 to 80,000 calls per month.

The comparison that matters for this story is not vendor versus vendor. It is the contrast in how each program arrived at scale and what scale actually looks like in a real company.

Microsoft’s own Power Platform growth numbers put the scale in a sentence the trend pieces usually skip. Sixty million monthly active users is not the size of a developer tool. It is the size of a productivity platform, and the people running it are inside Microsoft customer companies rather than inside Microsoft’s own engineering org.

AI Now Lives in the Build Pipeline

Low-code is the surface layer. Underneath it, AI assistants now sit in the build pipeline itself, and the productivity numbers on the engineer side of the house are clean enough to read without asterisks.

GitHub’s own productivity research on Copilot ran a controlled experiment with 95 professional developers writing an HTTP server in JavaScript. The group with Copilot had a higher completion rate, 78% versus 70%, and finished the task 55% faster than the group without it. The average completion time was 1 hour and 11 minutes with Copilot, and 2 hours and 41 minutes without.

GitHub’s survey of more than 2,000 developers found that 60% to 75% of users felt more fulfilled in their job, less frustrated when coding, and able to focus on more satisfying work when using Copilot. The numbers worth flagging for this story are the structural ones, because they explain why the trend keeps accelerating without anyone deciding to push it.

• 56 million monthly active users on Microsoft Power Platform, as of April 30, 2025 (Microsoft FY25 Q3 earnings call)
• 27% year-over-year growth on that base, per the same earnings disclosure
• 78% vs 70% task completion rate with versus without GitHub Copilot, on a controlled HTTP-server task
• 55% faster average task completion with Copilot in the same GitHub study

The same build pipeline now extends past professional developers. Where vibe coding has moved into consumer apps is a separate question from this story, but the direction is the same. Plain-language prompts now produce working apps for people who do not write code, and the security review at the end of that pipeline is the part most teams still have not built.

Why Every Business App Is Now a Mobile App

The reason the trend keeps spilling past IT is partly about who builds the apps and partly about where they live. Pew’s most recent smartphone ownership data puts smartphone penetration in the United States at 91% of adults as of June 18, 2025, with 98% owning a cellphone of some kind. The first survey of smartphone ownership Pew ran, in 2011, found the figure at 35%.

When the default surface for everything inside a company is also the default surface for everything outside the company, internal apps stop looking like an IT project and start looking like a customer-facing product that happens to be private. The build pipeline that produces a field-service check-in app and the one that produces a customer service portal now share the same low-code platforms and the same AI coding assistants. The governance work is the same work too, and it is the work the trend pieces keep skipping.

The Bill Comes Due for IT

The HBR authors warned that the most expensive failure in this trend is not the security breach or the duplicated subscription. It is the app that becomes critical to a department after its creator leaves for another job, the failure mode Davenport and Tomak flagged in their piece. The same paper offered the counter: companies can reward employees for the innovation rather than punish them, and put governance in place that catches the work without slowing it down.

Three years on, the data says the catching up is happening, slowly and largely without fanfare. The HBR concern about apps outliving their creators is now a procurement category. The trend kept going. The bill kept growing. The people paying it have not changed jobs to do so.

Frequently Asked Questions

What is a citizen developer?

A citizen developer is a non-IT employee who builds business applications using approved low-code or no-code tools, a term Harvard Business Review authors Tom Davenport and Kerem Tomak used in their 2023 article “We’re All Programmers Now.” The HBR piece described the role as employees “with no coding background” who “collaborate with generative AI tools to build entire applications.”

How big is the low-code market?

Microsoft’s Power Platform reported 56 million monthly active users in the company’s fiscal third-quarter earnings call on April 30, 2025, up 27% year-over-year. The figure came directly from Microsoft’s own disclosure, cited in a Microsoft Power Platform blog post in May 2025.

What is shadow IT?

IBM defines shadow IT as any software, hardware, or IT resource used on an enterprise network without the IT department’s approval, knowledge, or oversight. IBM’s topic page cites Cisco research reporting 80% of company employees use shadow IT, and an IBM Institute for Business Value study reporting 41% of employees acquired, modified, or created technology without their IT or information security team’s knowledge.

Why are IT teams worried about AI coding tools?

According to the HBR piece, the concern is apps that become business-critical and remain known to only the employee who built them, or that stop working when that employee leaves. The HBR authors wrote that organizations may end up “dependent on citizen-developed systems known to only a few employees, or whose developers have long since left the company.”

Do mobile app building trends require coding?

No. Low-code platforms handle a large share of business mobile app building now, with Microsoft reporting 56 million monthly active users on Power Platform as of April 30, 2025. Pew Research Center reported 91% of U.S. adults owned a smartphone as of June 18, 2025, making mobile the default surface and reducing the cases where building without code is not an option.