Today is my one-year anniversary at Citrix. I wrote “This is gonna be awesome” as the closing line in my inaugural Citrix blog post, and I truly meant it. Looking back over the year, I can happily say it was true. I was given the mental space and intellectual freedom to deeply & publicly explore how AI is transforming knowledge work and what it means for our customers, driving important industry-wide conversations including:
Looking back at my 26 essays and dozens of talks & interviews from my first year, I knew then that we were building toward something bigger, but I couldn’t quite articulate what “it” was.
But now I can.
I recently built a personal AI knowledge system—a “second brain”—that has fundamentally changed the way I work. My second brain is a folder of plain text files on my laptop containing my frameworks, my positions, my research, my writing, and my evolving thinking on every topic I cover. An AI reads all of them, maintains them, and builds on them every day. Every “conversation” with my AI (whether typed or dictated) uses this full context. When I think out loud, the system captures and integrates it. When I write, it draws on everything I’ve already figured out. It compounds daily.
This has the potential to sound like a trendy “productivity hack,” but I will stake my reputation based on 32 years of enterprise experience that this is much bigger. Workers using a “second brain” (or “personal AI knowledge system” or whatever you want to call it) is a fundamental and structural change to what knowledge work is.
The key is the 80/20 split I’ve been writing about. Every AI tool on the market—Copilot, agents, chatbots—operates on the visible 20% of knowledge work: emails, docs, scheduling, formatting, administration, etc. The other 80%—the strategizing, pattern recognition, and judgment calls built on years of experience—has always been invisible. It lived and was processed in people’s heads and nowhere else.
A second brain reaches the previously untapped 80%. For the first time, the actual knowledge part of knowledge work can be digital too. And once it’s digital, everything changes.
It’s hard to capture the gravity of how a second brain works and its impact in just a few paragraphs. I wrote a longer post on LinkedIn which I recommend for a deeper look.
My second brain story is just a personal story of how I’ve transformed my own work. I didn’t invent these concepts. There are many how-to guides on YouTube & Substack (& mine on LinkedIn), so even though there are many workers starting to work this way, this is still something only used by those at the frontier of AI use. Like maybe 0.1% (?) of knowledge workers globally.
But as more workers discover this concept, (because (1) yes it really is that amazing, and (2) it will get easier over time), what happens in the enterprise as this concept spreads to 1%, 5%, 10%, 50% of workers?
My second brain is literally just text files in a bunch of folders on my laptop. Most of them are my actual “content”, and some are plain-text instructions for AI about how it should maintain, operate, and update this knowledge system.
Both the AI and I can read, write, and update any file, and since they’re just text files, it’s trivially simple to connect them to external systems. What happens when these files are stored in Git? What happens when multiple people start collaborating and sharing bits of their knowledge via GitHub? What happens when you put an MCP (Model Context Protocol) interface in front of a second brain? (Or just the part that should be shared?)
Suddenly any rank-and-file worker’s brain can get a data feed from any other brain they want. It’s like the way skills are learned in the Matrix (except this is real life and literally happening today). That other brain can be a coworker, their manager, their CEO, an influencer, or even a department-level or company-level brain. It can also be an “app” (or what we in 2027 will quaintly remember as “apps”).
This is incredibly powerful. It’s starting to happen now. And it’s a governance vacuum.
Enterprise IT and company leadership are just flat-out not ready. Even worse, many are not even having the right conversations about AI and the future of work.
I’ve had actual conversations with leaders, and listened to popular podcasts—both in the last month—where the most pressing security concern was, “What if someone pastes secrets into ChatGPT?” That is a 2023 problem. The 2026 problem is fundamentally different in scope.
Think about where a second brain gets its knowledge. Yes, it can connect to corporate apps and data via MCP. (Though that’s easy enough to block with traditional security approaches.) But personal AI can also see what’s on a worker’s display, scrape content from browsers, and read screenshots. The app stores are full of tools that silently record and transcribe everything they hear (meetings, conversations, rambling thoughts and pontifications), which are easily fed into a personal AI. Amazon sells devices that do the same thing from your desk. A worker’s second brain doesn’t just pull from governed enterprise systems. (In fact that might not even be the right place to look.) It absorbs everything a worker sees, hears, and reads throughout their workday.
Corporate data lives in thousands of places: some company-controlled, some departmental tools that never went through IT, some personal SaaS apps that don’t use corporate SSO, and some written on post-its, day planners, and notepads. Workers’ personal AI systems are connecting to all of them. From this day forward, you must assume everything any worker hears in a conversation, sees on a shared screen, or reads in a Slack channel will end up woven into their personal knowledge base within seconds.
This isn’t a breakdown of existing security. It’s an entirely new vector. A new category that the existing mental model of “data loss prevention” wasn’t built to contemplate.
Nobody’s governing any of this. Each AI platform—OpenAI, Anthropic, Google, Microsoft—is building its own governance controls, but each one only governs its own connections. An enterprise running Claude and ChatGPT and Copilot gets three separate, incompatible governance layers with no unified audit trail and no centralized policy enforcement. And that’s before you account for open-source models, personal AI systems, ambient recording devices, screen-reading tools, and the thousands of community-built MCP connectors that have no enterprise governance at all.
I realized in the past few weeks that I fell into my own trap.
Everything I wrote in year one—the 7-stage roadmap, the agent security pieces, the workspace governance arguments—assumed that work was still happening within the traditional structure. I was looking at how AI enters the existing model: here’s how workers access systems, here’s how apps deliver data, here’s how IT manages it all—and now add AI. Even when I said “AI agents will be like coworkers,” I was imagining a traditional coworker—someone who accesses a workspace, uses the apps, and follows the well-worn and current governed paths.
I fell for the faster-horse problem. I was thinking about future frameworks for a present day world. But that world is dissolving.
What second brains reveal is that work itself is breaking out of the 20% container. That’s so important but also such a fundamental shift that even though the 80/20 framing was mine, I myself didn’t even fully understand the implications until the past few weeks.
For most of my first year at Citrix, I was focused on the future of work within the visible 20%: how AI transforms the apps, the documents, the meetings, and the systems IT can see. But the second brain operates in the 80%. That part was never in any system, and the part that no framework I built last year deeply considered.
Once you see that, a lot of what I spent the year thinking about needs to be reframed. Of course, managing apps and workspaces will still be necessary, but it’s not where the action is anymore. The action is in the data flows: the MCP connections, ambient inputs, and places where a worker’s AI touches corporate knowledge regardless of whether an “app” is involved. The governance question isn’t, “Which apps are workers using?” It’s, “What data sources is a worker’s AI connecting to, what is it absorbing through screens and microphones, and where is that knowledge flowing?”
That’s a fundamentally different question than anything enterprise IT is set up to answer today.
I’m truly optimistic about this future. The productivity and cognitive gains from second brains are real. I’m living them. The ability to connect knowledge across people, teams, and organizations is going to create value we can barely imagine right now. I don’t think companies should try to stop this. (They can’t anyway, and the ones that try will lose their best people to companies that don’t.) Companies will need to embrace this, because finally they’ll be able to get some tech leverage to the “knowledge” part of “knowledge work.” Modern companies have always been massive knowledge machines, and for the first time in history we have technology that can broadly dig into that previously untapped goldmine. Oh, if Alfred P. Sloan could experience this!
But governance has to catch up.
Luckily we have some precedent for how this could happen, as the consumerization of IT in the early 2010s played out the same way. Back then, workers wanted iPhones and Dropbox. IT said no and tried to block them. That didn’t work, and workers found workarounds because the gap between what the company provided and what workers could get on their own was too wide to ignore. The ultimate (and positive for everyone) resolution was providing workers with a corporate-sanctioned experience that was as good or better than what they could get on their own, delivered within a managed and secure environment. EMM, enterprise file sync, BYOD programs… our whole industry figured out that everyone wins by enabling, not restricting.
Personal AI systems will follow the same pattern, but at a much larger scale. Workers are building these today because the capability gap is massive (and is compounding daily). The answer in 2026 isn’t to ban second brains any more than the answer in 2012 was to ban iPhones. The answer is to provide workers with AI-powered work environments that are just as powerful, within infrastructure that gives organizations the visibility, security, and governance they need.
That requires rethinking of the full end-user computing stack. This is so much more than just managing apps and securing endpoints. The governance that matters now spans the entire work stream: from the device, through the connection, into the working environment, and across every data source a worker’s AI can reach, with an intelligence layer woven throughout that can see what’s happening, flag what’s risky, and enable what’s productive. That’s the architecture this future demands.
For decades, the invisible core of knowledge work—the strategizing, judgment, and pattern recognition—lived safely in people’s heads because there was nowhere else for it to go. Now there is. It’s becoming files, connections, and data flows. It can be shared, built upon, and compounded. It can also be leaked, exfiltrated, and exploited.
We’re not just watching AI enter the workplace. We’re watching the fundamental nature of knowledge work change in ways that break every assumption our industry was built on, including some of mine. That’s what makes this moment so important, and honestly, it’s what makes this job so exciting.
At Citrix, we’ve been securing and delivering work for over 35 years, and this is exactly the kind of moment we were built for. If you thought year one was interesting, wait until you see what’s coming in year two.
Read more & connect
Join the conversation and discuss this post on LinkedIn. You can find all my posts on my author page (or via RSS).
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。