Your Employees Hired an AI Assistant. Nobody Told IT.
Key takeaways
- In Reco's State of Agent Security 2026 data, about 80% of AI tools ran without IT oversight, and small and midsize companies averaged 414 unsanctioned AI tools per 1,000 employees.
- ESET found 73% of SMBs are integrating AI, 70% say it introduces new risks, and 40% have no policy restricting AI use outside approved processes.
- Today's AI tools don't just answer questions. Many can read email, open files and act on your behalf.
- The fix isn't banning AI. It's knowing what's connected, deciding what's allowed, and giving people a safe option.
Meet your newest employee
Your office has a new assistant. It takes meeting notes, summarizes email threads, and drafts replies before anyone finishes their coffee. Everyone loves it.
Nobody hired it.
Someone on your team signed up for a free AI tool over the weekend, clicked "Allow" on a permissions screen, and gave it access to their Outlook, their calendar, and the shared drive. That assistant now has more access to company information than some of your actual staff. And nobody wrote it down.
That's shadow AI: AI tools your people adopt on their own, without IT knowing or approving them. It usually starts with good intentions. It usually ends with nobody knowing where your data went.
Chatbots answer questions. AI agents take action.
A couple of years ago, most AI tools were a box you typed into. You pasted text in, you got text out. The risk was mostly "someone pasted something they shouldn't have."
Today's tools, often called AI agents, are different. They connect to your email, files, calendar, and other apps, then read, write, and send things on your behalf. Every "Allow" click hands over a key.
That's not automatically bad. It's how these tools become useful. But it means three questions matter more than ever:
- What AI tools are connected to company accounts right now?
- What can each one see and do?
- What happens to that access when the employee leaves?
If the honest answer to all three is "no idea," you're in good company. That's also the problem.
The numbers behind shadow AI
This isn't a hypothetical. In Reco's State of Agent Security 2026 report, about 80% of AI tools in its data ran without IT oversight, and small and midsize companies averaged 414 unsanctioned AI tools per 1,000 employees.
Business owners aren't oblivious, either. ESET's 2026 SMB Cyber Readiness Index found 73% of surveyed businesses are integrating AI and 70% acknowledge it introduces new risks. Yet 40% have no policy restricting AI use outside approved processes. Most owners know there's a risk. They just haven't had time to do anything about it.
And it costs real money when things go wrong. IBM's 2025 Cost of a Data Breach report found that organizations with high levels of shadow AI paid about $670,000 more per breach than those with little or none, and one in five organizations in IBM's study reported a breach due to shadow AI.
What shadow AI looks like in your office
| If you're in... | The AI shortcut that creates risk |
|---|---|
| Real estate and property management | An AI email assistant reading closing details, wire instructions, and client financial documents |
| Law firms | Privileged documents pasted into a free AI tool to "just summarize this quickly" |
| Medical and dental practices | An AI note-taker or message summarizer handling patient information |
| Accounting and financial services | Client tax returns uploaded to an AI tool for faster analysis |
| Construction and trades | An AI tool connected to the shared drive full of bids, contracts, and subcontractor pricing |
Small business AI governance: five steps to take this month
- Take inventory, no blame attached. Ask your team which AI tools they use and for what. Make it an amnesty, not an interrogation, or you'll get the polite version of the truth.
- Check what's already connected. Microsoft 365 and Google Workspace both show which third-party apps have been granted access to company accounts. Review that list and remove anything nobody recognizes.
- Require approval for new app connections. Change your settings so employees can't grant outside apps access to company data without an admin saying yes.
- Give people a safe option. If you ban AI without offering an approved alternative, people will just use their phones. Pick business-grade tools with proper data protections and make them easy to get.
- Write a one-page AI use policy. Spell out which tools are approved, what data never goes into AI (client financials, patient information, privileged documents), and who to ask. Then remove AI app access as part of every employee departure.
Related reading: Your Email Security Might Have a Door Propped Open · Is Your Small Business Ready for AI? A 5-Question AI Readiness Checklist · Managed IT
Frequently asked questions
What is shadow AI?
Shadow AI is any AI tool employees use for work without IT knowing about or approving it. It includes chatbots, note-takers, browser extensions, and AI agents connected to company email, files, or apps.
Should small businesses ban AI tools?
Usually not. Bans tend to push AI use onto personal devices where you have even less visibility. A better approach is approving a few safe tools, setting clear rules about sensitive data, and controlling which apps can connect to company accounts.
How do I find out which AI tools can access my company's data?
Start with your Microsoft 365 or Google Workspace admin console, which lists third-party apps that have been granted access. Combine that with a no-blame survey of your team. An AI readiness and governance assessment covers both and flags which connections are risky.
What's the difference between an AI chatbot and an AI agent?
A chatbot responds to what you type. An AI agent connects to your systems and takes actions, like reading email, opening files, scheduling meetings, or sending messages. That access is what makes agents useful and what makes them a security concern.
How Hawkins Consulting can help
We're not the AI police. We want your team using AI, just not in ways that leak client data.
Our AI Readiness & Governance Assessment finds the AI tools already in use and the apps connected to your company accounts, identifies where sensitive information is exposed, and gives you a short list of approved tools, a plain-English AI use policy, and the settings changes to lock it all in. We'll also point out where AI could genuinely save your team time, because that's the fun part.
You don't need to know every AI tool on the market. You just need to know which ones have the keys to your business.
Schedule your AI Readiness & Governance Assessment
Sources
- Reco, State of Agent Security 2026 (2026), as reported by The Next Web, Shadow AI is already inside your company. Here's how to get control of it, September 7, 2026.
- ESET, 2026 SMB Cyber Readiness Index, summarized in Useful but risky: A complicated relationship between SMBs and AI tools, June 8, 2026.
- IBM, Cost of a Data Breach Report 2025, press release, July 30, 2025.
More from the Learning Center.
Your Email Security Might Have a Door Propped Open
Microsoft detected 7.3 billion phishing threats last quarter, and some got through because businesses had quietly switched off part of their own protection. Here's what to check.
Read moreIs Your Small Business Ready for AI? A 5-Question AI Readiness Checklist
Five questions to answer before you buy another AI tool, a simple way to score your readiness, and the two projects that tend to pay off fastest.
Read moreThe Complete Small-Business Cybersecurity Checklist for 2026
Ransomware, phishing, credential theft, and supply-chain attacks all target small businesses. Here are fifteen practical controls for 2026, starting with the five that matter most.
Read more