Qvery Podcast Episode #3: Becoming AI-Fluent

Hey, Vlad here. My guest this episode is Warren Leow, founder of AI Training 2 You, a corporate training company that has trained 1,300+ professionals across 300+ organizations on AI automation, agents, and orchestration.

He is also the founder or co-founder of SuperHomes, SuperJobs, SuperCFO, and SuperAgentK, which gives you a sense of how Warren spends his weekends.

The reason I wanted Warren on the podcast is simple. He has seen what AI rollouts look like inside real companies, from one-person operations to enterprises with tens of thousands of employees. Most of the public conversation about AI adoption is theory written by people who have never sat in a room and tried to get a finance team to use Claude. Warren has done it 300 times.

What we ended up talking about is not really about tools. It is about what becoming AI-fluent looks like, for an organization, for an individual marketer, and for the next five years of white-collar work.

Why Founder-Led Companies Adopt AI 10x Faster

The first thing I asked Warren was which kinds of organizations move fastest when AI shows up at the door. His answer came down to one variable: who sits at the top.

Warren explained that the more entrepreneurial the organization, the faster it moves, especially when it is founder-driven. The reason is mechanical, not vibey. Founders carry the full P&L math in their heads, so when something promises a 5x productivity gain, they do not need a committee to approve trying it. They open the tool, build something, and decide on the spot.

Every operator-CEO I have ever worked with has adopted AI faster than the equivalent enterprise client with a board, a VP of Innovation, and an annual planning cycle. The board company runs a pilot. The operator-CEO migrates an entire workflow in a weekend.

Warren's own moment of conversion came when he started building things himself, which is the most common path I see for non-technical founders:

"My own light bulb moment came about last year when I was using Cursor for the first time and I could build things very quickly. That became the inspiration moment to really say, things are changing."

Cursor for a CEO who had always outsourced engineering work is the same moment Claude Code was for me a few months later. The instant you build something yourself that used to require a contractor, a spec, two rounds of back-and-forth, and a Slack thread, you stop treating AI as a productivity boost. You start treating it as a permission slip to do work you used to delegate.

What Makes an AI Rollout Stick

When I asked Warren what separates rollouts that work from rollouts that stall, he had a clean three-factor framework. He has watched 300 companies try, so the pattern is not a guess.

The first factor is senior leadership alignment. The Slack-memo version doesn't work. The CEO has to use the tools personally, week to week. Leadership only understands what AI can do when they are hands-on with it, and the executives who use AI themselves are the ones who hold their teams accountable. The ones who delegate AI down never push the rollout past Level 1.

The second factor is a critical mass of trained operators. This is where most companies get it wrong. They train the engineering team and call it done. Warren was specific that the trained operators have to span finance, operations, customer care, sales, marketing, and legal, because those are the people who own the actual processes. If only IT can build automations, you have an IT bottleneck, not an AI rollout.

The third factor is IT preparation and governance. Clarity on infrastructure, model choice, data pipelines, and the rules that keep customer data inside the building. Not glamorous work, but if you skip it, you create the kind of compliance exposure that ends a rollout overnight.

What Warren emphasized, and what most training companies get wrong, is that training itself is not the deliverable. The whole notion behind AI Training 2 You, he told me, is never to sell training in the first place.

The goal is change management, and his real KPI is how successfully the organizations roll out AI adoption after the training is done.

Training without accountability is performance art. Most large companies have already done plenty of it.

The Four Levels of AI Nativeness

Warren walked me through the maturity curve he uses with clients. It is the most useful frame I have heard for talking to executives who keep asking how AI-native they really are.

Level 1 is ChatGPT for drafting. Emails, summaries, basic research, the table-stakes use cases that became normal three years ago. Most knowledge workers live here, whether their company has paid for a license or not.

Level 2 is deeper research and simple agents. Tools like Perplexity for synthesis, plus narrow agents that handle one task end to end: OCR for invoices, reconciliation for the finance team, lead scoring for sales, and ticket triage for customer care. Each one is a single-purpose worker.

Level 3 is orchestration. This is the level where things change. Warren framed it this way:

It is possible for AI to be your partner and not just a tool. When I say a partner, it is not about you instructing it. It is turning around to be an advisor, or an executor, or even an instructor going forward.

Level 3 is where one person manages multiple agents that coordinate with each other, where the AI starts pushing back on your strategy, and where you stop thinking of AI as software you operate. You start treating it as a colleague you collaborate with.

Level 4 is approaching the point where AI runs the function. Warren thinks this is around five years out for most companies. CEOs who can take a vacation and let the agents run the org. I think the timeline is shorter for individual functions like marketing operations or finance reporting, where the work is highly repeatable. Slower for everything that requires human judgment about other humans.

The 80/15/5 Rule for Marketers

When I asked Warren what this means specifically for marketers, he reframed it as a distribution problem. In any profession, roughly 80% of people are average, 15% are good, and 5% are exceptional. The AI shift is not hitting these groups equally.

Warren's take: the average tier is getting devalued first. AI can already produce the median version of a campaign brief, a landing page, a competitive analysis, or a paid ad. When the median output is free, the median producer has a pricing problem.

The good tier, the 15%, holds on longer, because they bring craft and judgment. But the timeline keeps compressing. The 5%, the genuine strategists and creative directors with original perspective, are the ones whose work AI still cannot replicate. They are the last to be disrupted, and Warren was specific that this is where individual marketers should aim:

Having the strategies to bring it all together is still going to be the main constraint. And that is still going to be, I think, probably the last thing that AI will be able to disrupt.

I have watched this pattern play out with content writers over the last three years. In 2023, every think piece predicted content writing was dead within 18 months. In 2026, content writing is more alive than ever, but the role has shifted. The writers who survived stopped competing with AI on volume and started competing on perspective, taste, and orchestration. They became content directors with AI underneath them.

The same arc is about to play out across every marketing function. The strategist survives. The executor gets absorbed into the strategist's stack.

There Is No Perfect AI Stack Right Now

I asked Warren the question every marketing team is currently arguing about: what is the right AI stack? His answer was the most refreshing thing in the conversation.

There is no right stack right now. The incumbents are getting upended too fast for one to exist. Warren made the case that Adobe, Canva, Figma, HubSpot, and even tools like Midjourney that felt category-defining 18 months ago are all being eaten from above by general-purpose models that can do their core jobs natively. Your stack today gets rewritten every six months whether you like it or not.

Warren put it bluntly: six months ago everyone was on ChatGPT. Today everyone is on Claude. Six months from now it might be a Chinese open-source model running on-device. The stack is downstream of the model layer, and the model layer is the most volatile part of the tech industry right now.

This is exactly why we built Qvery as an AI agent company, not a piece of marketing software. Tools track. Agents work. Agents survive model changes because the strategy lives in the orchestration, not the underlying foundation model.

If you are picking marketing tools right now and optimizing for best-in-class of any single feature, you are picking the slowest-moving target on the table.

Becoming AI-fluent is not a destination. It is a spectrum. Warren's four-level framework gives you the rungs to climb, and the faster you climb, the further ahead you get of the 80% who will not make it past Level 2. T

The full conversation is here on Spotify.