What I Wish I Knew About AI Before Building Two Companies

Six months ago, I finally got good at a tool I’d been fighting with since January.

I’d watched the tutorials on the weekends. I’d rebuilt my workflow around it twice, thrown out both versions, and started a third. I read the changelogs like they mattered. By month five, I could talk about the tool the way you talk about something you’ve earned, a little too proudly, a little too often, in meetings where nobody had asked.

Then I mentioned it, a little pleased with myself, during a product review with the Ducknowl team. One of the engineers we’d brought on a few months earlier just laughed. “Oh, I’ve used that since it launched. Want me to wire it straight into the platform?”

Six months. She’d known it cold the whole time. I hadn’t even thought to ask.

I didn’t feel embarrassed exactly. I felt something closer to recognition, like I’d been solving a problem that had already been solved, just not by me, and not anywhere near where I’d been looking for the answer. I’d spent five months of evenings on something she could have handed me in an afternoon.

What got me wasn’t just the time lost. It was how long I’d assumed the learning curve was mine alone to climb. I run a staffing company. I’ve spent fifteen years telling other people that the right hire changes everything. And I still sat there for half a year, alone with a laptop, trying to out-learn someone who already had the answer sitting on my own team. I knew the principle cold. I just hadn’t applied it to myself.

If you’re looking for AI lessons for entrepreneurs, that’s the first one, and it’s not really about AI at all. It’s about where you assume the learning has to happen.

Lesson one: AI tools are not a substitute for a person. They are a multiplier for one.

I’d been treating the tool like it was the hire. Like if I just learned it well enough myself, it would do the job of a teammate, and I could skip the harder decision of bringing someone new onto the team. It doesn’t work that way, and I should have known that fifteen years of staffing before I ever touched the tool.

A tool in the hands of someone who already knows the underlying work, the judgment calls, the client history, the shortcuts that actually matter, makes that person faster. A tool in the hands of someone still learning the underlying job just makes the confusion move at a higher speed. I was the second person in that equation. She was the first. No amount of tutorials was going to close that gap, because the gap was never about the tool.

That reframed a lot for me. The question was never “which AI tool should I learn this quarter.” It was “who already knows how to put this to work in service of the actual job, and how do I get them on my team.” One of those questions costs you six months. The other one costs you a phone call.

I think about it now the way I think about the earliest Ducknowl product demos. Showing an investor the platform never made the platform valuable on its own. What made it valuable was the person behind it who already knew exactly which edge cases would break it and which wouldn’t. The demo was never the skill. It just made the skill visible faster. I’d forgotten that for a while, sitting in front of my own screen, but it’s the same lesson wearing a different jacket.

Lesson two: the tools that save a founder the most time are the ones they stop touching entirely.

Every tool I still personally operate is a tool that hasn’t paid off yet. That sounds harsh, but it’s held up every time I’ve tested it. There’s a real cost to how many entrepreneur AI tools waste time instead of saving it, and the scheduling tool I still tinker with on Sunday nights? Marginal, at best. It saves me maybe ten minutes and costs me twenty deciding how to configure it. The reporting dashboard someone else built, tuned, and now just hands me a two-paragraph summary of every Monday? That one changed how I run my week, and I haven’t opened the underlying tool myself in months.

The pattern isn’t “which AI is the most powerful.” It’s “which one disappeared off my personal to-do list because someone else took ownership of it.” I used to measure a tool by how impressive it looked in a demo, how many features it had, how sharp the interface was. Now I measure it by how quickly I forgot it existed, because forgetting it means someone competent is running it, and I’m not needed there anymore. That’s the real difference between building a business with AI alone and building a business with an AI virtual assistant on the team. That’s not a loss. That’s the whole point of hiring well.

Lesson three: building two companies taught me that judgment can’t be automated. Only time can.

When I co-founded Ducknowl, I assumed I’d already learned this from running Simpalm for over a decade. I hadn’t. I relearned it from scratch, and it was the expensive way to learn anything twice. In the early months, I tried to be the one reviewing every candidate assessment, every product decision, every new hire’s first week on the job. I told myself it was because I cared about quality. The truer answer is that I hadn’t yet separated two very different things I was doing with my time: making actual judgment calls, and doing repetitive work that only looked like judgment because I was the one doing it.

AI and automation can clear out that second category almost entirely, the scheduling, the first-pass screening, the formatting, the follow-up emails that say the same three things every time. What’s left after you clear that out is the real job of a founder: deciding who to trust, what to build next, when to walk away from a deal that looks good on paper. No tool makes that decision for you, and I don’t think one ever will.

I remember the week it actually clicked. We had two candidate pipelines running at Ducknowl at the same time, and I’d inserted myself into both, reviewing screening notes that a system had already sorted correctly the first time. A co-founder pulled me aside and asked why I was re-doing work that had already been done well. I didn’t have a good answer. I was there out of habit, not necessity. That was the week I started asking, before touching anything, whether the task in front of me needed my judgment or just my presence. Most of the time it only needed my presence, and that’s exactly the part a good tool or a good hire can take off your plate.

But every hour a tool or a person recovers from that first category is an hour you get back for the second. That’s the entire trade. I didn’t see it clearly until I was running two companies at once with no room left to fool myself about where my hours were actually going.

I later learned I wasn’t imagining how much of my week that first category was quietly eating. Deloitte’s 2025 Global Human Capital Trends survey, which polled thousands of business leaders worldwide, found that people spend roughly 41% of their working day on tasks that don’t actually contribute to the value their organization creates.

Source: Deloitte, 2025 Global Human Capital Trends

The LATAM connection.

It’s not just this one tool, either. The professionals we’ve placed across Latin America consistently arrive already fluent in whatever AI tools a role calls for. Not because we get lucky with sourcing, but because it’s a generation of professionals who came up alongside these tools rather than retrofitting them onto older habits.

A recent AmericasMI analysis found that 65% of Latin American consumers already use AI, with Gen Z in the region leading adoption at 62%. This isn’t a handful of early adopters, it’s a generational shift.

Source: AmericasMI, Latin Americans and AI Trust Report

I’ve written before about why Brazil’s specialized virtual assistants outperform in US roles, and the structural reasons go deeper than any one story of mine. But every time I watch this pattern repeat itself, on a new team, with a new tool, it confirms the same thing. The fluency gap I spent six months closing on my own is exactly the gap LATAM virtual assistants are built to close for you, before day one.

What I’d tell myself five years ago.

Stop trying to become the tool. You’re not competing with your team to be the most technically capable person in the room, you’re supposed to be the one who knows what the room needs. Find the person who already knows the tool cold, hand them the real problem, and spend your six months on the thing only you can do.

I didn’t need to learn that platform. I needed to hire the person who already had.