I’ve worked in the collectibles industry for most of my life. When the Blue Jays won the 1992 World Series, everyday fans suddenly wanted something tangible to remember it by, and the Canadian memorabilia market expanded almost overnight. That moment showed me there was real business infrastructure behind the nostalgia. Decades later during 2025, I began building CollectorLINK, a blockchain-based platform designed to bring structure and transparency to physical collectibles.
While building it, I integrated AI wherever it made practical sense. It helped draft documentation, organize complex metadata structures, refine legal language, and accelerate operational workflows. It became a thinking assistant — a way to compress time. But it did not replace human (well.. my) judgment.
So in the middle of a suddenly loud national debate about whether artificial intelligence will wipe out jobs or transform them, I’m sharing my real-world thoughts on my last year building a Canadian tech startup with AI now embedded directly into its software operations.
I certainly didn’t start with a master plan for how AI would be used — I simply knew the tool was there. ChatGPT had already been helping me write, so that part was obvious. As I began researching blockchain tokenization, it quickly became invaluable for organizing ideas, structuring documents, and pressure-testing my business plan.
But startups don’t run on ideas alone. They need logos. They need design. So I tried AI for that too. The result was clear almost immediately (sucks): I needed a human designer. That became hire number one.
Hire number two was a developer. And despite all the noise right now about “vibe coding” and AI-built companies, no serious startup is getting off the ground without human direction. After a year of building a startup alongside AI, my conclusion is a lot simpler than the headlines would suggest. AI didn’t replace the people in my company. It helped me run the company better.
I discovered the creative work still belongs to humans. Design, storytelling, editing, strategy — those are decisions, and decisions require judgment. But the administrative work — the arithmetic behind the scenes — is exactly where I have found AI shines. Reconciling receipts, organizing financial records, producing the right forms. Those tasks don’t require taste or interpretation. They require accuracy.
There is a clear line I’ve observed over the past year. If a job requires taste, trust, negotiation or long-term relationship building, AI assists but does not replace. If a job is binary — correct or incorrect, yes or no — AI can likely manage it effectively.
Bookkeeping fell into the second category. Reconciling deposits, separating out HST tax amounts, generating summary reports — there is no such thing as “creative bookkeeping.” There is only accurate or inaccurate. AI handled those repetitive financial tasks with consistency which I felt allowed me at least right now, to delay the hiring a bookkeeper at this stage of the company’s development.
What AI changed was not the size of my workforce. It changed the shape of it.
For me, building a startup didn’t feel like participating in a technological revolution. It felt like adopting a powerful new tool. That’s the pattern I’ve seen so far. AI hasn’t eliminated jobs inside my startup. If anything, it helped me get the company off the ground faster — which led to hiring real people sooner than I otherwise might have. In other words, AI replaced the way I often “total”. It did not replace decisions.
For now, at least, the robots aren’t running the company. They’re just doing the bookkeeping.
UPDATE (June, 2026)
Four months ago I wrote above that AI replaced totals, not decisions in my company. I still do believe that. But I maybe filed that report a little early. Because here’s what I didn’t see coming.
By April, Claude AI was writing code for CollectorLINK. Not helping. Not suggesting. Writing it. Updates and improvements that used to take a week were landing in hours. The product got better, faster, in ways I hadn’t fully anticipated when I sat down to write that February piece and had a full-time human coder going on the CollectorLINK platform.
And here’s the part that’s harder to write – This meant a conversation had to happen with my developer. And it didn’t feel like a comfortable one. There seems no handbook for sitting across from someone you trust and respect, and saying — in whatever careful words you choose — that the tool is now doing a meaningful part of what YOU do. That they need to level up – manage what the AI is building, rather than just build alongside it.
What I can say is this: I haven’t given up on humans in the software loop. Not even close. I certainly don’t want to manage our tech stack alone. I genuinely believe there’s a role here that’s worth transitioning into. But believing that, and getting there smoothly? Those are two different things and right now for CollectorLINK – the jury is still kinda out.
I also said for the longest time I wouldn’t use AI in video. The CollectorLINK YouTube intro we recently released is old school, talking head, manual edit. Except for one shot – I needed b-roll that made a specific point – that the token is not the collectible, the token is a wrapper for the real thing. So I tried an AI video generator. It worked. Was it worth $45? Creatively, yes. The lesson was the same one I keep learning. I didn’t use AI because I planned to. I used it because it solved a problem nothing else could at that moment.
What AI changed, I wrote in February, was not the size of my workforce — it changed the shape of it. I think that’s still true. I’d just add now that reshaping a workforce – regardless of size isn’t always clean. Sometimes it’s awkward and unresolved and you’re figuring it out week by week. The robots still aren’t running the company. But the conversation about what humans do inside it — that one has gotten a wee bit more complicated.
– Richard Budman (June 2026)