During my recent trip to China, I noticed something striking among the programmers I met who work at large tech companies: they're all paying for Cursor accounts out of their own pockets.
Not one was seeking reimbursement from their employer.
This wasn't an isolated case. As I spoke with more developers, the pattern became clear. These professionals—employed by companies with strict data security policies and generous expense accounts—were willingly paying monthly subscriptions from their personal finances for AI-powered development tools.
Large tech companies usually don't allow employees to upload internal code to external tools, but the productivity gain from Cursor is so big that people are willing to pay for it themselves.
"Getting approval would mean convincing IT security, legal, and finance departments," explained one senior developer at a tech company. "Even if I eventually got approval, the process would take months. And I need to ship code now."
What we're witnessing is the emergence of what I call "Shadow AI"—the grassroots adoption of AI tools by individual knowledge workers who find the productivity benefits so compelling that they're willing to sidestep official channels and pay from their own pockets.
This phenomenon extends far beyond developers. Marketers are quietly using ChatGPT and Claude to draft copy and brainstorm campaigns. Designers are leveraging Midjourney and DALL-E to generate visual concepts. Product managers are using AI to draft specs and analyze feedback.
The implications of this trend are fascinating:
1. The B2B vs B2C distinction is blurring for productivity tools. What looks like enterprise software on the surface is actually being purchased through consumer channels.
2. Bottom-up adoption is becoming more powerful than top-down implementation. By the time companies formalize an AI strategy, their employees may already be proficient with tools the organization hasn't even evaluated.
3. The value proposition of these tools is so clear and immediate that professionals are making the rational economic decision to invest in their own productivity, even when their employers should theoretically be covering these costs.
4. We're seeing a new kind of "bring your own device" culture emerge—except now it's "bring your own AI."
As one developer put it to me: "I could wait for my company to figure out their AI policy, or I could ship twice as much code this quarter and position myself for promotion. It's not even a hard choice."
For AI startups, the lesson is clear: sometimes the path to enterprise adoption isn't through the CIO's office, but through the individual practitioners who will champion your tool after experiencing its benefits firsthand.
And for enterprises? The message is equally clear: your talent is already using AI. The question is whether you'll acknowledge this reality and create policies that harness this enthusiasm while addressing legitimate security concerns—or continue pretending it isn't happening while your most innovative employees find workarounds.
Shadow AI isn't just a temporary phenomenon; it's the leading indicator of how the next generation of productivity tools will enter and transform the workplace. The developers I met in China are just the beginning.
What AI tools are you paying for out of pocket? I'd love to hear your experiences in the comments.
Great point about bottom-up adoption becoming more powerful than top-down implementation. I've read a few articles recently that point to a lack of enterprise adoption as supporting evidence for generative AI's failure to produce any real economic value. Clearly, the writers are missing this perspective (among others).
Yep, also seeing this too. Great newsletter.