"Your Employees Already Use AI. You Probably Haven't Sanctioned It."

"Most leadership conversations about AI start with the wrong question. Which platform to procure. Whether to assign someone to own the AI question. When to set a corporate position before employees act. If you walked through your office right now, you would see that line was crossed some time ago. ChatGPT tabs open on screens. Prompts being pasted in from confidential documents. Drafts of client emails being rewritten by a model nobody on your IT team has approved. Notes from internal meetings summarized by services your data protection officer has never reviewed. Most of it invisible to leadership. None of it documented anywhere. You probably already suspect this is happening in your firm. The research confirms how widespread it is. A 2025 study of143 small and medium-sized enterprises in one European region found that every single one used AI in some capacity. Only 3% had a formal procedure governing how.! In the other 97%, AI use was taking place daily, employee by employee, with no procedure, no oversight, and no organizational view of what was being asked or what was being shared. This pattern is not limited to one region or one firm size. OECD data covering member countries on three continents shows enterprise AI adoption has roughly doubled every two years since 2020, reaching roughly 20% of firms last year.1The European Investment Bank's 2025 survey found rates as high as two-thirds of firms in some EU markets, against a European average closer to one-third.1 Those figures measure formal adoption. The kind that shows up in surveys, that procurement teams record, that leadership knows about. They do not measure the kind your employees are doing without telling anyone." "The reason this happens is structural, not cultural. The platforms employees reach for are the ones that solve their immediate problem at the moment they have it. A junior associate at a law firm finishing case preparation at nine in the evening, with the partner meeting at eight tomorrow morning, does not stop to consult procurement. They open ChatGPT, paste the briefing note, and get back something useful enough to move forward. The friction of the sanctioned path, if a sanctioned path exists at all, is higher than the friction of the unsanctioned one. So the unsanctioned path wins. Every time. Across teams, across grades, across departments. Multiple 2025 surveys converge on the same picture}. A majority of employees use AI tools their organization has not approved. In some surveys the figure rises above three-quarters. The"

"pattern is consistent with what employees themselves say about why they use unauthorized tools: the sanctioned alternatives, where they exist at all, are often slower than the immediate work demands. This puts your organization in a specific situation. Your AI usage, at this moment, is being shaped by the friction profiles of the tools your employees can reach. Not by any decision your leadership team has formally made. The platforms in use, the data being shared, the queries being asked, the answers being acted upon: all of it is happening, all of it is shaping decisions that affect your firm, and most of it is invisible to the people accountable for the firm's risk. The honest description of where most leadership teams stand is not ""should we adopt AI."" It is ""our organization already has. Do we know what risks come along with that?"" What follows in this series is the answer to that question. What risks. What costs. What the alternative looks like. And what specifically you can do about it."

"AiSU Strata is building the governed knowledge access layer described in this series. The longer essay on this topic, ""Your Employees Have Already Adopted AI. You Haven't,"" goes deeper on the structural reasons behind the pattern, the specific risks of ungoverned adoption, and what governed AI access actually looks like."

"1. Barbu, C. M., & Sitnikov, V. V. (2025). Study Concerning AI Usage in SME from Oltenia Region. Management & Marketing, Volume XXIII, Issue 1/2025. University of Craiova. DOI: 10.52846/MNMK.23.1.11.The study surveyed 143 small and medium-sized enterprises in Romania's Oltenia region and found that all 143 used AI in some capacity, with only 3% having a formalized procedure governing its use. .-:> 2. OECD (2025), AI Adoption by Small and Medium-Sized Enterprises, discussion paper prepared for the G7. Adoption rose from 5.6% in 2020 to 14% in 2024 across OECD member countries. OECD (2026), ""AI use by individuals surges across the OECD as adoption by firms continues to expand,"" reports adoption reaching 20.2% in 2025. .-:> 3. European Investment Bank, EIB Investment Survey 2025 - Finland. ""Finland leads Europe in generative AI adoption, new EIB survey shows,"" EIB press release, 2025. Available at: https://www.eib.org/en/press/all/2025-502-finland-leads-europe-in-generative-ai-adoption­ new-eib-survey-shows. The 66% figure refers to Finnish firms, the highest-adoption EU market per the survey. .-:> 4. Three 2025 surveys converge on the same pattern. An IDC study found 56% of employees use unauthorized AI tools at work, while only 23% use AI tools their organization provides and governs. A WalkMe (SAP) survey of 1,000 US working adults found 78% admit to using AI tools not approved by their employer. A Cybernews survey of 1,000+ US employees found that 85% of those with access to approved AI tools also use unapproved ones."