Dutch VMUG UserCon 2024 · Den Bosch, Netherlands · March 26, 2024 · solo keynote with slides — also titled “It’s Springtime for AI.”
A return to live speaking after two years running a pinball hardware and software company, using a “seasons” framework to make sense of AI, career reinvention, and why springtime — not summer or fall — is exactly the right season to jump in.
Feeling like the train already left
Everyone else seems to be doing amazing things with AI, and the overwhelming feeling is that the train left the station without you. Thirty years in IT since May 1994 — brianmadden.com for fifteen years, eight books on VDI, Citrix, and DaaS, founding BriForum, a stint at VMware in Shawn Bass’s EUC Office of the CTO — and yet standing on this stage with genuinely no idea what’s happening next.
Two years in the pinball industry
After VMware, the actual departure was to a company building software and hardware for physical pinball machines (Labyrinth runs on the platform) — documentation, diagrams, technical explainers, just like the old day job, but not about HIPAA or GDPR compliance. Then a move to Paris to join ILKI, an independent consulting and analyst firm with no vendor partnerships, which means being able to say whatever needs saying without a vendor relationship to protect.
The industry moved while you were gone
Coming back: Citrix and VMware are both owned by private equity now, and customers are nervous about both. Microsoft, meanwhile, is at a three-trillion-dollar market cap — higher than Apple by roughly one full Tesla. The three companies that used to define the whole EUC world look completely different than they did two years earlier.
The seasons framework
Borrowing loosely from “The Fourth Turning” and its generational cycles, technologies and companies (and careers) move through spring, summer, fall, and winter. VMware’s spring was Workstation, GSX, ESX — nobody believed you’d run production workloads on it. Summer was vSphere, vSAN, Horizon, Workspace ONE: bold, confident, expanding. Fall was acquiring Carbon Black, Pivotal, Nyansa — stockpiling for winter. Winter is private equity ownership, re-evaluating everything, letting some customers and employees go, and clearing space for what’s next — not a failure state, just the necessary part of the cycle. Citrix ran the identical arc: WinFrame and MetaFrame in spring, XenApp in summer, acquiring XenSource, Cloud.com, ShareFile, and Podio in fall, and now its own private-equity winter.
The same seasons, personally
The career maps onto the same cycle: spring was value-added-reseller work in Akron, Ohio, learning from nothing. Summer was brianmadden.com, BriForum, books, learning to speak on stage. Fall was the VMware CTO office — harvesting everything built up in summer, becoming a builder instead of just a commentator. Winter was the pinball company: stepping fully away from technology, hunkering down, learning GitHub and Copilot on the side, healing and refreshing. Every season is necessary to reach the next spring.
What each season actually feels like
Spring is exploration and experimentation — new tools, low stakes, room to make mistakes because there are plenty of other seeds. Summer is when you start implementing what you learned and sharing it publicly for the first time — that first forum post, that first talk, gaining confidence and visibility among peers. Fall is harvest: leading initiatives, mentoring others, getting recognized as someone who actually knows things. Winter is reflection — taking stock of what’s still relevant, deliberately letting go of what isn’t, and making room for the next spring. Every individual technology someone works with runs its own version of this cycle, often out of sync with the others running at the same time.
Applying it to AI, specifically
Thirty years of tech history — PC repair and Novell NetWare, desktop engineering, SMS, Citrix, VMware, VDI, HCI, MDM, consumerization, digital workplace — and every single one got the same all-in treatment: books, conferences, mailing lists. AI is next, except this time there’s a real problem: people who’ve genuinely been doing AI for ten or twenty years, training and tuning models, doing RAG, merging models on their own GitHub accounts, make “I’m good at ChatGPT” feel embarrassingly small by comparison.
The reassurance: it’s springtime for everyone
The macro cycle for enterprise AI is spring, for the field as a whole, not just for latecomers. A quick search for “artificial intelligence” returns hundreds of millions of results, and the count visibly grows in the thirty seconds it takes to formulate the next search. Fifty thousand books on Amazon. The old playbook — read the book, go to the conference, become the expert — doesn’t work here, because the target keeps moving faster than anyone can read. Prompt engineering went from the hottest new job posting to functionally obsolete within six months, once models started auto-correcting prompts on their own. A company that spent $10 million fine-tuning ChatGPT on proprietary data watched GPT-4 outperform that custom model out of the box on release day.
You don’t have to know how it works
The old EUC path required deep mechanical understanding — you could genuinely become a Horizon expert who understood every underlying layer. That path doesn’t exist for AI. But not understanding the mechanism doesn’t disqualify you from using something well: nobody flying commercial understands jet engines, and plenty of mechanics don’t fully either. Set a goal first — understanding what AI means for the company, for personal productivity, or for long-term employability — and treat the technology as something to use skillfully, not something to master mechanically.
It’s mostly hype, and that’s useful information
The AI vendor conference circuit runs the identical hype playbook every enterprise technology has run for decades — peak of inflated expectations, vendors making maximalist promises, audiences taking photos of slides that amount to “real-world data matters, don’t over-trust the tech, costs add up.” True, but equally true of cloud, of Microsoft licensing, of virtually everything sold to enterprises for the past thirty years. Recent weeks have already brought walk-backs — Copilot products warning users about specific use cases, reminders that an LLM is fundamentally statistical rather than deterministic. That’s the trough of disillusionment arriving on schedule, the same as every technology cycle before it. None of it means AI isn’t real. It means the gap between hype and reality is roughly 100x, and being behind on the hype isn’t the same as being behind on the substance.
Nobody actually knows what’s coming
Yuval Noah Harari’s point, made around this same time on late-night television: this may be the first moment in history where nobody has any real model of what the world looks like in twenty years. That uncertainty cuts both ways — it’s genuinely disorienting, and it’s also permission to not have it figured out, because nobody else does either.
AI as the ultimate consumerization of IT
Every company is already an AI early adopter, whether or not IT has sanctioned anything, because employees are already using it individually — the “secret cyborgs” phenomenon, where knowledge workers are quietly human-machine augmented without necessarily disclosing it. LLMs make unreliable software in the traditional procedural sense — statistical rather than deterministic — but they’re genuinely excellent for individual productivity, the same way trust in a coworker develops: you learn what to double-check, what to trust outright, where the failure modes are.
What comes next: Open Interpreter
The project actively being explored at the time of this talk: Open Interpreter, an open-source LLM interface that runs code directly on a local machine and can watch and interact with the desktop — effectively an early, hobbyist version of a computer-using agent, years before the term became mainstream. It needs a desktop running continuously to work from anywhere, which raises the obvious question: keep a desktop running in the cloud, reachable from anywhere? Could this, of all things, finally be the actual year of the desktop?
Key formulations
“I feel like the AI train has left the station and I missed it.”
“It’s springtime for AI… the stakes are low. Now’s the time to play.”
“You don’t have to know how something works in order to know it’s valuable.”
“AI is the ultimate consumerization of IT.”
“We’re not as behind as we thought.”