Barndoor.ai Finds That 50% of Knowledge Workers Are Giving Unsanctioned AI Access to Work Systems

New research commissioned by Barndoor.ai reveals 9 in 10 workers are already using AI on the job, and more than half have granted it access to work systems

Venn.ai, recently launched by Barndoor.ai, connects Claude, ChatGPT, Cursor, and VS Code to your business apps, so your AI can take real actions with accuracy and security

NEW YORK, March 18, 2026 /PRNewswire/ -- Artificial intelligence has rapidly evolved from a simple writing assistant into a powerful operational tool, yet there are still no guardrails in place. According to new research commissioned by Barndoor.ai, 91% of enterprise employees are now using AI on the job, and more and more are using it to take real action inside the business systems that run their companies.

The findings paint a portrait of a workforce that has embraced AI far ahead of their employers' policies, with 28% already using agentic AI tools that take actions on their behalf and perform multi-step tasks. And they're not slowing down: 60% use AI daily or more, with 35% using it multiple times a day.

"People want AI to do tasks across the apps they use every day, but without proper controls, that's risky," said Oren Michels, co-founder and CEO of Barndoor.ai. "Workers aren't going to stop, this data is more proof they're already in motion. In fact, the data suggests a widening gap between how fast workers are moving and how prepared companies are to support them safely. The question is whether companies will catch up before something goes wrong."

AI has the keys, whether companies know it or not

The most striking finding: 50% of participants have already granted AI access to work-related applications. While email and calendar access are the most commonly connected tools, workers have also handed AI the keys to customer data, financial systems, API keys, and internal databases.

Nearly half (48.4%) have used non-approved AI tools at work, either intentionally or without knowing what their company had sanctioned. Of those, 40% cite pressure to get work done, and 42% say non-approved tools simply produce better results. Meanwhile, nearly 20% of employees don't even know what their company's AI policy is.

From writing tool to workflow engine

Workers are using AI for far more than just drafting emails, they are deploying AI for complex, multi-step workflows: automating invoicing from project completion through client delivery, summarizing support tickets and drafting team responses, managing GitHub builds end-to-end, compiling daily trading activity reports, and running sales pipeline analysis that previously required hours of manual work.

Real-world examples from the research include: a field support manager who uses AI to summarize tickets, create training notes, and schedule follow-ups for a team of 14; a finance worker whose AI agent pulls prior-day reporting and summarizes it across multiple dimensions each morning; and a freelancer who has automated their entire invoicing workflow from project close to client payment.

Tool loyalty is low. 58% of workers use two to three different AI tools, with 7% using four to six, and workers say they want fluidity rather than to be locked-in to one provider. ChatGPT and Microsoft Copilot lead adoption, followed by Gemini, Claude, and Perplexity.

A workforce that feels unsupported

Despite the pace of adoption, over 40% of employees (42%) do not feel well-supported in using AI at work. Only 46% work at companies with a defined AI policy, and of those using company-authorized tools, ease of use (45%), capability (43%), and data privacy (38%) are the top drivers of compliance, not security concerns.

Venn.ai: Closing the gap between AI ambition and AI safety

Barndoor.ai recently launched Venn.ai, a platform that gives AI tools like Claude and ChatGPT controlled, permissioned access to everyday workplace applications, including Salesforce, Jira, Google, Notion, and more.

Venn.ai is designed for the workers the research describes: people already using AI at work who want to do more, but need guardrails that protect them and their organizations. Unlike standalone AI platforms that require users to abandon their existing tools, Venn.ai works inside the AI environments people already use, with human confirmation built into any action before it executes.

"Venn.ai gives everyone, not just engineering and IT teams, the ability to use agentic AI safely," Michels said. "We're putting security controls directly into users' hands."

Venn.ai is available today with free early access at Venn.ai. Paid plans will launch at a later date.

Read the full report here

Research methodology

This research and virtual focus group was conducted between January-February 2026 among 155 professionals at mid-sized to enterprise companies with Remesh, a hybrid insights platform that allows you to speak with your audience as one whether live, asynchronous, or via video. It combines qualitative and quant style questions, speed, and scale powered by Remesh's proprietary artificial intelligence to present and analyze audience responses simultaneously.

About Barndoor.ai

Founded in 2024, Barndoor.ai is the AI governance platform built for the agentic enterprise. Barndoor.ai's centralized control plane gives IT and security teams the visibility and fine-grained access control they need to govern AI agents across every tool, team, and platform. Venn.ai provides knowledge workers secure, governed access to agentic AI across the workplace apps they already use. Together, Barndoor.ai and Venn.ai address the AI governance gap from both the enterprise end, as well as individual empowerment. Learn more at Barndoor.ai and Venn.ai.

SOURCE Barndoor AI