Listen on: Apple Podcasts · Spotify · Amazon Music
Brian walks through the 2026 update to his 7-stage roadmap for human-AI collaboration — a framework he published a year ago where the predictions came true about three years ahead of schedule. He and Dave reframe the shape of the journey: it’s not a ladder, it’s a palace. Phase three (the cognitive extension / second brain) isn’t a rung you climb past. It’s the permanent foundation everything else builds on.
Then Brian argues that the dominant narrative around AI agents is wrong. The industry is telling knowledge workers to audit their jobs, identify tasks, and build automations. Brian makes the case that this is RPA thinking applied to the wrong problem. The real move is connecting the second brain into all the applications and data sources that human workers already have access to — not scripting workflows, but giving AI the same front door humans use.
Dave closes with his own story: what happened when he moved his second brain from a personal laptop into a sanctioned Citrix VDI environment. He calls what he had before “looking through the glass ceiling” — able to see the valuable work he wanted to do with AI, but unable to reach it from a personal system. The close: the data you already have is the product. AI is the platform that consumes it. That’s when you see value from AI initiatives.
Links mentioned
Brian’s second brain:
https://brianmadden.ai
Dave’s second brain:
https://davebrear.ai
Transcript
Brian Madden (00:01)
Hello and welcome to the Citrix AI Hotsheet Podcast. My name is Brian Madden, joined by my co-host and Citrix colleague Dave Brear. We’re recording this on June 19th, 2026.
When we started the show, I said it came from Dave — we had our one-on-one meetings and they were so interesting and fun that we said we should just record these. And in fact, I haven’t seen you since last month. This is literally our one-on-one that we’re recording.
Dave Brear (00:34)
This is it, yeah. I got everything I was looking for from the last episode in that regard — it’s exactly that. The conversations we were having, just capturing those. I think it was a good idea.
Brian Madden (00:49)
What have you been up to in the past month? You’ve been posting on LinkedIn quite often.
Dave Brear (00:57)
Yeah, it doesn’t seem like a long time since we last caught up, but actually it’s been a while. For the most part, all the stuff we’ve been talking about, I’ve just been using it. My system has been running, I’ve been working in this way. And we have some interesting things to talk about a little bit later about those.
Brian Madden (01:19)
His system — the Dave Brear second brain. I tell everyone: everyone knows me as the second brain guy, and I’m like, you’ve got to meet Dave. Dave’s the one who showed me this, blew my mind, and changed my perception of what the future of work looks like.
Let’s jump right into our first topic. This is episode three. Dave and I last met a month ago for episode one. Episode two was a special edition — I’d been traveling around Europe giving talks. I spoke at the EUCTech Cruise in Norway a couple weeks ago, and earlier this week I spoke at a DanofficeIT event in Denmark. In all of these I was talking about the future of end-user computing and the future of Citrix — a topic I’m calling “The Last Chapter of EUC.” I liked that talk enough that it should reach more than just the people in the room, so when I got home I sat in my office and recorded it and published it as episode two.
If you’re listening on Spotify or Apple, you can also watch us on YouTube. If you’re watching on YouTube, audio-only versions are available too.
One of the things I was talking about in my speeches this month — and the subject of a recent Citrix blog post — is a roadmap I created about a year ago called “The Seven Stages of Human-AI Collaboration.” The basic idea: as an individual worker, how do you use AI, and how does that change as your skills develop, as AI technologies mature, as the models get better? I’m not going to go through the whole thing step by step because that’s what episode two was — go back and listen if you haven’t. But I do want to call out a few highlights.
I just published the 2026 edition last week — an update to what I did a year ago. And what’s remarkable is how much has changed. That first version was directionally correct, but my timelines were off by a factor of three. I said this is how we’ll be working with AI in 2027 and beyond. By January of this year — six months ago — you and I were both already living it.
The quick version: phase one is using AI like a question-answering machine, just asking questions and getting back answers. That’s how most of us begin. Phase two is using it as a thinking partner — ideating, chewing on ideas, processing and analyzing things with it. Phase three is the cognitive extension or second brain, where AI becomes a knowledge corpus that’s a genuine extension of all your work. Phase four is where you take that knowledge corpus and start connecting it into your applications — now your second brain can connect into business systems, pull data, write data, connect into all of the things. Stop there for now — listen to last week’s episode for the full picture.
Dave Brear (05:38)
What a difference a year makes. A year ago the predictions were forward-looking and optimistic, and not wrong — but the speed at which they became true is what changed.
What solidified for me watching that episode: in 2025, I was looking at this almost like a ladder — you start at the bottom, work your way up, try to get to the summit at stage seven. What occurs to me now is that the ladder analogy is true only through steps one and two. Once you get to step three and you’re really using AI as a cognitive extension, that’s not something you ever get past. It becomes the foundation for everything else. The later phases augment what you’re doing in phase three — they automate things, they free up time so you can live more deeply in phase three. The cognition piece is where the real value is.
Brian Madden (07:05)
I like that. The shape is different — it’s not a ladder, it’s more like a palace on a hill. You climb a few steps to get there, but once you have your palace you’re expanding what happens within that space.
And that’s a big framing change from last year to this year. The first version was really about what AI could do at each phase — focused on the technology, what it could do as it improved. The 2026 version shifts to: what does this mean for you? How does this change how you work and how you live? Impact on the worker, not the technology.
As I’m saying this, I realize there’s a parallel frame for organizations — what happens when all the workers are doing this? That’s actually something we talked about a year ago, Dave.
Dave Brear (08:35)
Absolutely. In a traditional organizational hierarchy, the people at the top get paid the most because of their experience and strategic thinking. We give them executive assistants to handle scheduling and logistics so they can direct their effort toward what matters. We interact with their executive assistant rather than the person directly for a lot of the mundane stuff.
Where everyone now has these AI capabilities is that we can give everyone that same executive assistant. And the reality of that is my AI is probably going to talk to your AI, and we may not be involved in trivial but meaningful decisions — where we’re supposed to be, who’s doing what, what the commitments are. That gets taken out of our hands, and we’re freed up to focus on the things that actually move the needle.
Brian Madden (10:15)
I think you just outlined the topic for our next show. You and I should spend the next month building a similar framework as it applies to organizations. Because a company is made up of individuals who are all going to be at different phases. We can see the utopian future — imagine when every worker has a second brain, when every second brain is connected into all the IT systems. But what happens when five percent of your workers are there, twenty percent are a phase behind, and seventy-five percent are still using AI like a better Google? That’s a significant organizational challenge. We’ll dig into that next month.
One other thing — and this leads into our first main topic. As I was giving these speeches, I realized it’s one thing to be a blogger sitting in your treehouse pontificating.
Dave Brear (11:37)
Treehouse — that’s it! I’ve just got there.
Brian Madden (11:38)
I love that. It actually connects to Citrix’s roots. One of the reasons you and I have both stayed in the Citrix world for decades is the community. When we got into Citrix in the nineties, it was this strange side thing and we were all just figuring it out together, banging things together. That’s how you build a treehouse.
What I realized when I got out of the treehouse and started telling the story about the seven phases is that everyone is on this journey independently. I’m in phase three transitioning to phase four. Most of the world is still in phase one. Whatever phase you’re in, you understand it — and you can just barely see the next phase, because it’s “imagine this if you added that.” People can follow along. But as soon as you get two phases ahead, you’re talking about Mars. People check out.
Dave Brear (12:50)
I can see exactly what you mean. As someone firmly in phase three, when I look at the value of future phases, I’m thinking about time I can get back to live in my treehouse. But if you’re at phase one, those phases down the road look blurry — and what you see through the blur is: that sounds like it’s doing the work I’m supposed to be doing, and that’s a threat.
Brian Madden (13:20)
I love your characterization of AI as an executive assistant. If everyone had a really great executive assistant, that empowers them to spend more time on what matters. By the way, I do have a really great executive assistant — shout out to Una, my executive assistant at Citrix. She’s a human, and she’s wonderful.
Dave Brear (13:54)
We can’t all be vice presidents and futurists, Brian. I have to make do with what I can cobble together.
Brian Madden (14:01)
With AI, I think we all can be. Because people focus so much on tasks — if AI does these pieces of my job, what’s left for me? That is exactly what I want to talk about as my main topic today.
So to set it up: phase one is how most people use AI, prompting and getting answers. Phase two is loading in documents, using Projects, Cowork, or Notebook LM, having real conversations. Phase three is the second brain phase, which Dave and I have both written extensively on. Phase four is taking that second brain and connecting it into your applications.
I said in last week’s episode: you have to get to phase three. You have to put in the time to really understand how this works. Even if your company doesn’t support it, you can get your own personal Claude, Gemini, or ChatGPT subscription and start using AI as a second brain. Dave’s written a lot on this — links in the show notes. Everything I’m about to say assumes you’re at the second brain. If what I’m saying sounds crazy, get to the second brain first, then come back and listen again.
The second brain phase — phase three — is still inside our head. AI is managing our knowledge corpus, our tasks, connecting ideas, pulling in transcripts, helping us think through our work. But it’s all still thinking. The transition from phase three to phase four is the transition from thinking to doing. Or, to use Dave’s treehouse analogy — we’re branching out.
Dave Brear (19:49)
[Branching out — sorry, that’s terrible. I’ll leave now.]
Brian Madden (19:56)
Phase four is where AI starts to do things in the world. The analogy: you have a brain and you have claws. Claws without a brain — what can they do?
Dave Brear (20:54)
Damage.
Brian Madden (20:55)
Yes. That’s where your AI deletes your production database and the backups in nine seconds. I’m sitting squarely in phase three right now, just starting to connect into phase four — extending my AI out into the world to do things. You’re at a similar point, Dave?
Dave Brear (21:16)
Absolutely, yeah. Very early stages for me, because it needs to be done right. I’m not yet comfortable enough with what I’m asking it to do to delegate large chunks of work.
Brian Madden (21:24)
With our decades of actual enterprise IT experience, we both want to do this slowly — especially when it starts touching things in the real world.
Which brings me to my main point. I’ve got my second brain, I’m starting to branch out. This is AI entering the world — the claws, the agents. There’s a lot of conversation online about what happens when you let AI do things: you give it workflows, you automate specific parts of your job. That is the main narrative. And I believe that main narrative is wrong.
There’s a vision — and I actually wrote exactly this a year ago, so I have to own it — that you should use AI to create workflows and automate tasks. AI as a better workflow machine. That was my perspective a year ago. That was everyone’s perspective a year ago. I’ve moved past it, Dave’s moved past it. But you’re still hearing it everywhere, because most of the world isn’t at the second brain phase yet. They’re still in phases one and two.
If you don’t have the second brain and don’t really understand how this fits together, you’ll say: AI is an agent, agents do things autonomously, so let me apply that to the model I already know. Especially in our world — we’ve had agents for 20 or 30 years. They’re called RPAs: robotic process automation. You take a repetitive process, instead of paying a human to drag a mouse and transform documents and move data between systems, you create an RPA to do it automatically.
The narrative a year ago — the narrative I gave a year ago, the narrative everyone is talking about now, and which I think is wrong — is: every knowledge worker job is just a collection of tasks. AI is getting better. So the path to AI utopia is: look at your job, figure out which tasks you can automate, automate those, automate more, automate more. Eventually you’ve automated everything and AI has changed the world.
I believe a lot of people are trying to do exactly this today, which is why the narrative is “AI ROI isn’t there, it’s not working.” Knowledge workers are not going to fire up an AI agent studio, identify tasks, and build workflows. This capability has been available for decades. It doesn’t work.
Dave Brear (25:10)
The reason RPA hasn’t replaced people is because you need to put a lot of work in upfront to define the exact parameters of the workflow you’re automating, and you need to remove judgment from that workflow entirely. The fallacy is: now we don’t need to define as much explicitly, because the AI can apply judgment. So we can get to the outcome with less effort. The reality is that without the right information and context, that judgment will be flawed. Human judgment plays a huge role in how knowledge work actually gets done well, and that isn’t easily replaceable. You can’t go from RPA to AI agents and just roll it out to everybody. It’s a flawed premise.
Brian Madden (26:20)
Agreed completely. And I want to add a point of clarification: RPA absolutely has a role in the future, and AI is going to make RPA better. You’ll be able to apply AI logic to help with routing and smarter handling in contained workflow environments. I believe RPA has a place, AI helps with RPA, and the future of RPA is bright.
But within the marketing organization at Citrix, for example, there are knowledge workers who do have tasks that could be enhanced with AI — take this campaign, put it here, do this transformation. AI can help build automations for those tasks. I’m not saying AI plus RPA and task automation for knowledge workers is wrong. I’m saying it’s not the main event.
That’s still the visible 20% of knowledge work. If AI and automations can help with that, great. But so much of knowledge work is the stuff in our brains — the thinking time, the staring-out-the-window time. That is not repeatable enough to task-automate. And AI still has enormous value there. That value is not to replace your thinking.
Here’s how I’ve been framing it: if we’re at phase three using AI as a thinking partner, and we want to go to phase four and extend AI into the world, the next step is connecting AI into all our applications. This ties back to episode one — why I believe AI is going to enter the knowledge work world by using the same front door that human workers use, inside the same security and application delivery and governance environment.
Think about how I’m using AI as a second brain. Take an account strategist at Citrix who has to do a monthly account review. They probably spend days preparing: was that contract signed? Are there open help desk tickets? What are the current call notes across the account team? What’s in email, what’s in meeting transcripts? Multiple systems, all going through their brain one at a time.
Phase four is why this matters. I want to connect my second brain into all the applications — so I can talk to my AI and say, let’s look at customer number one. I think there’s a follow-up call to schedule, I think there might be a support ticket that’s probably been resolved. And my AI, through its skills, knows here’s your CRM, here’s your ticketing system, here’s the other account team members. It pulls data from all of those, builds the full picture, writes back when needed.
Dave Brear (33:54)
Absolutely. And as an account technology strategist, I’ve been in exactly that position. I took over an account recently where due to team changes the system wasn’t updated to show me as the ATS. An account review got scheduled that I wasn’t invited to, and I found out an hour before that I had to present.
I was able to take all the random thoughts in my vault, all the meeting transcripts from the handover, and synthesize a complete snapshot: current situation, where we are, the risks, what we’re doing now, what the customer needs from us. I showed up to the call and gave a comprehensive update with an hour’s notice, just by pulling together information that already existed across the business.
And you’re right — we need to find a way to bring these things together, and it will have unexpected results. A great example is Amazon and AWS. AWS didn’t come about from Amazon saying “let’s build a cloud service and sell it.” It started from an internal ethos: every service — storage, compute, networking — was treated as a commercial product provided to internal consumers. Those consumers had to interact with each other in a standardized, trackable, accountable way. By making all services addressable in a uniform way, they ran their business efficiently. Then someone said: we have unused capacity, we could sell this. A new use case emerged from the infrastructure.
The same will be true for data. We need to treat all our data sources as commercial products within the business — available to the right users in an appropriate way. Information locked in a silo is information not being used. You’re paying for the storage, paying for the line-of-business application, and someone is making a critical decision without the right context.
Brian Madden (33:54)
That ties right back to episode one — why AI needs access to all the applications and data sources that human workers have. Today’s AI can use APIs, connect via MCP, operate web browsers, operate desktop environments. If you want workers extending their second brain into all the systems they need, you’ve already built that infrastructure for the human workers. The takeaway: when you think about AI agents in enterprise environments, don’t think “workflows I’m going to automate.” Think: you have your second brain, your workers are using AI as a coaching and thinking partner, and that whole environment connects into everything and expands its universe. When you hear Jensen Huang talking about every enterprise needing an agent strategy, it’s not so you can automate workflows. It’s so you can extend second brains into every data source, application, and system of record that the humans need.
I feel like this tees up what you wanted to talk about, Dave.
Dave Brear (35:22)
Yeah, fantastic. What I’ve been going through for the past several months started as almost a skunk works experiment — what’s possible, working agile with the tools at hand. It started on a personal laptop with notes connecting to LLMs, figuring out the model, what works.
I have a notebook here — I write down thoughts in this notebook — and what was going into my personal AI system was essentially this, but now addressable by AI. There were huge parts of my job where the nuggets going into my second brain were returning enormous value. But in the piece I wrote about this last week, I said there’s a glass ceiling. The really valuable stuff — the account review work I was just talking about — is on the other side of that glass. What I really need is access to all the company information and all the systems. On a personal system, there’s no way to do that without violating employee conduct policy.
Brian Madden (37:05)
The glass ceiling being that barrier between your AI system and all the company data and apps, but you can’t connect those together. It’s like videos of a young kitten walking into a glass window and not understanding why it can’t go through.
Dave Brear (37:12)
Exactly. On a personal system I can have AI help with maybe 20 to 30% of my day-to-day — the stuff that’s appropriate to take from my brain, put out, and have played back. But the really useful stuff — show me all the customer support cases for this customer over the last 12 months, what are the trends, do they need a health check and if so on what subjects — all of that is sensitive information that should be secured in enterprise systems.
So we’ve been working on an initiative to bring these experiments into the business. Right now that looks like a secure execution environment: VDI desktops running in Citrix’s systems where I can run a copy of this brain. And it has been transformative. That example I gave earlier — turning up to a meeting prepared on an hour’s notice — happened in week two of working this way. Just having all that information in a sanctioned environment, I can sleep at night and I can smash through the glass ceiling.
But even then it’s still a progression. Right now, while I have a secure place for this information, it’s still largely manual input — I find something interesting, I put it in. Over time that compounds. The next stage is getting the enterprise to treat all those data sources as products that can be subscribed to in a context-appropriate way. How do I connect to my email, to what’s in the Microsoft tenant through something like Work IQ? How do I connect to our CRM through MCP connections? So I can ask questions about customer trends without importing the data myself.
Now that I’m in here with the security clearance upgrade, I’m more eager than ever: who do I need to talk to to make this happen? These are the conversations all enterprises will be having once the first users move into these secure tenants. The first thing they’re going to say is: wow, what a world has opened up to me — what can I bring in now?
Brian Madden (40:41)
You hit the nail on the head with “who do I talk to?” You can climb from phase one to phase three on your own as an individual. But once you want to connect your second brain to all the business systems and go to phase four, that’s above our pay grade as individuals. You have to talk to the CIO, COO, CISO. Even with Citrix secure workspace in place and access to web apps, modern applications, and desktop applications — you still have to think about: what agent is running, what is it seeing, what policies are set up? We probably want separate user accounts for the agent so it doesn’t have the same permissions I have. Read access to start. I don’t necessarily want to trust it to write back into systems yet.
This expands beyond what an individual can do and requires IT — and in many cases requires real leadership sponsorship. This is a CEO, CHRO, CISO level conversation. The Citrix admins can’t make this happen on their own.
Dave Brear (42:10)
The top-down message that needs to come from the C-level is: data is the lifeblood of our organization, and data locked in a silo is data not being used to its full potential. It’s data that isn’t being thought with — and what’s the point in even having it? We need to solve this problem. We need to make data available appropriately. Not give everybody access to everything, but put the mechanisms in place so it can be consumed in a consumer-like way by everyone who needs it. That needs to be the number one priority on CIO agendas.
Brian Madden (42:51)
And I think it’s a number one priority for everyone who cares about the evolution of knowledge work — whether you’re an end user, in IT or EUC, a partner, or a Citrix colleague listening to this. This story is a lot bigger than us. But there’s a lot of work for IT to do, and it’s good work.
This whole notion of AI coming in and replacing IT jobs — so much of what we’ve always done still applies. Securely delivering workspaces, disk image layering, application layering, application virtualization, policy management, profile management. All of that still applies — it just evolves. Instead of merging application layers, we’re merging knowledge layers. I’ve got my department-level context vault, my company-level vault, my peer-level vault, my personal one. Agents are hooking into these. What are they running, what are they seeing, what access do they have, how do I audit and manage that? This is our world. EUC administrators, architects, and engineers: your skills don’t disappear. The application evolves. And there’s a lot for IT to do.
Dave Brear (44:44)
Absolutely. Coming full circle to where we started — the misconceptions about AI failing and not delivering value. To sum it up in one sentence: the misconception is that AI is the product that needs to be implemented. It isn’t. The data you already have is the product that needs to be fixed so that the platform of AI can consume it in a meaningful way. That’s when you start to see value from AI initiatives.
Brian Madden (45:24)
That is a perfect note to end on.
Thank you all for giving us your time. Shout out to Dave for teaching me about the second brain. And a reminder: both Dave and I have published public versions of our second brains — links in the show notes. Dave is at davebrear.ai. Mine is at brianmadden.ai.
What that means: you can take your AI system — even if you’re just at phase one using it for prompting — click the connect button, connect to an MCP server, type in either of our URLs, and it brings our entire corpus of knowledge into your AI so you can ideate on your projects with our perspectives built in. These second brains are updated continuously — by the time you hear this, this episode will already be in there.
My last takeaway: you have to get to stage three. You have to start using AI as a cognitive extension, as a second brain. If your company doesn’t support this, that’s fine. Don’t break any rules — but get your own personal subscription and start putting your ideas, thoughts, papers you’re reading, things you’re thinking about. Once you understand how powerful that can be, you’ll be equipped to advocate within your company for connecting it into your existing environment.
Thank you so much. Dave, as always, thank you for your time. Thanks everyone for listening. See you next month.
Dave Brear (47:03)
Thank you.