Axess Connect podcast, with Kris Haynes and Stefan Stickley · July 1, 2025 · Part 1 of 2, ~30 minutes.
Brian’s first appearance on Axess Connect, about five months into his Citrix futurist role, joined by Stefan Stickley (Axess Systems’ internal AI specialist). Part one covers the futurist role itself, the seven-stage roadmap, and why AI’s real impact on EUC comes through existing workspaces rather than new cloud-native infrastructure.
The futurist as scout
The job is really a scout role — living way out ahead, mostly focused on what’s happening five years out. Brian doesn’t live inside Citrix’s product group; his job is to describe what end-user computing and work look like five years from now, what matters in that world, and then how Citrix fits into it.
The seven-stage roadmap
Built out of noticing everyone was describing a different phase of the same underlying progression. Laid out from when ChatGPT launched: simple questions in, answers out; then deeper use; then AI agents watching your browser (Microsoft’s Copilot for Edge being a live example); stage four is computer-using agents; stage five is those same agents running in the cloud without a human watching; then AI orchestration on top of all of it.
Tools as the unlock
LLMs were famously bad at symbolic reasoning — math, counting letters. The fix wasn’t making the model itself better at arithmetic in its own token-based “brain”; it was giving it access to Python and other tools. It’s the same arc as human evolution, from critters running around to critters with tools. The moment a model could take its own output text and use it as an instruction to actually do something — that was the real unlock. Everything since has clicked forward from there.
MCP and A2A
An LLM on its own was sandboxed — a genuinely smart conversational partner, but only a conversational partner. Two protocols changed that. A2A, agent-to-agent, lets two LLMs talk directly to each other, a more sophisticated version of the old meme of holding a Copilot phone up to a ChatGPT phone. MCP, model context protocol, is a universal API interface you can put on top of anything — a database, Box, Dropbox, GitHub — so an agent can connect to an MCP server and effectively say “hello,” and the server answers “hi, I’m a file database, what would you like to find today,” wrapped in JSON but really just plain English underneath. Models are increasingly growing like a network of experts rather than one giant model, and those experts don’t even have to live inside the same model — they can be entirely separate systems, the same way human expertise lives across whole communities rather than inside one person.
The humanoid robot analogy
Task-specific robots look like their task — a robot arm, a Roomba, a self-checkout kiosk. A general-purpose worker robot has to look roughly human, because it has to do everything a human does: “make yourself about six feet tall, give yourself some fingers, and use the tools I already have.” Applied to knowledge work: the tools are a Mac laptop, Chrome, Google Workspace, Workday, email, meetings, podcasts. AI needs to enter that existing world and use the same tools workers already use, not live behind a brand-new black-box API layer in the cloud — that path is the same 25-years-and-counting project of turning every desktop application into a SaaS app.
AI as a worker
Humans are workers. AI is becoming a worker too, and from IT’s standpoint there isn’t much real difference — a desktop or workspace with apps, identity, and context ought to be reachable by an AI worker the same way a human worker connects to it. For a career built around only ever provisioning for humans, workers now split into human workers, augmented human workers, and AI workers — just one more “any” in the any-device, any-application, anywhere, any-connection strategy EUC has run for thirty years, now extended to any worker.
What AI in EUC actually means
Every product shipping an AI feature is just the latest technology iteration, not a fundamental transformation of knowledge work on its own. The real story is different: every rank-and-file knowledge worker doing TPS reports at a cubicle is now genuinely in play to be disrupted. Any job done entirely through a computer is in play — not necessarily today, but the door is open now. And for AI to actually enter that world, it has to fit into the world as it already exists, not the other way around. Thirty-one years in, this is the most exciting stretch of the industry Brian has seen.
Key formulations
“Make yourself about six feet tall. Give yourself some fingers, and use the tools I already have.”
“From the IT standpoint, there’s not much real difference” between provisioning for a human worker and an AI worker.
“Any device, any application, anywhere, any connection — now any worker.”
“I’ve been working 31 years and this is the most exciting time in this industry for sure.”