How one CEO quietly rebuilt his company with AI

What actually happens when a non-technical CEO stops delegating “the AI thing” and starts building with it directly? In this story, a footwear founder explains how he went from tinkering with chatbots to wiring his entire company into an AI-driven operating system—boosting conversion, speeding decisions, and changing how he hires.

From fast-growing brand to operational overload

The company at the center of this story is a vertically integrated footwear brand: it designs shoes, manufactures them in its own factory in Brazil, and sells mostly direct-to-consumer. In just a short time, it grew from producing around 300 pairs a day to about 2,000 pairs a day, with roughly 700 employees and close to a $100M run rate.

Growth, however, came with pain. Pre-orders piled up (at one point 9,000 pairs), the factory needed hundreds of new hires, and systems across sales, product, factory, logistics, and finance all had to keep up. The founder describes this period as one of the most intense of his career: rewarding in hindsight, but chaotic in the moment.

On top of that, tariffs hit exports from Brazil to the US, forcing a quick pivot into the Brazilian market. That move worked—Brazilian sales grew more than 10x year over year—but it added even more complexity to an already busy operation.

Why this CEO decided to build with AI himself

Like many leaders, he started by casually playing with large language models: asking questions, cleaning spreadsheets, and summarizing information. But he quickly realized that treating AI as a toy or a plug-and-play app would miss its real potential.

Instead of waiting for his tech team to “bring him an AI solution,” he dove in personally. His reasoning was simple:

• Technology complexity had been “romanticized” by engineers and vendors for years.
• Modern AI tools are fundamentally democratic—if you can think and type, you can use them.
• The person with the clearest vision of how the company should work is the CEO, not an external consultant.

So he started working late nights, experimenting with agents, connecting tools, and learning on the fly. He didn’t come from a deep software engineering background, but he had enough technical curiosity to get started—and used AI itself to fill in the gaps.

Rethinking the company as a flywheel, not silos

Before AI entered the picture, he already thought of the business as a flywheel rather than a hierarchy:

• Sales & marketing bring in customers and revenue.
• Product designs shoes based on feedback and performance.
• Factory manufactures every pair in-house.
• Storytelling & brand explain the product to the customer.
• Support functions like finance, HR, and data keep everything running.

In theory, each area should constantly feed information to the next. In practice, that meant endless meetings, manual reports, and delays. Product teams waited weeks for customer feedback. Operations waited for sales forecasts. Marketing waited for tech to fix the site.

AI became the glue that could connect all of this automatically—if he could give it the right context.

Step one: connect everything and “hack” your own company

The first phase wasn’t glamorous. It was about access and integration. The CEO describes it as “hacking” his own company:

• Connecting his inbox so AI could read and summarize email.
• Hooking into tools like Shopify, Klaviyo, Slack, and the ERP system running on a server in the factory.
• Finding who had which passwords and tokens—sometimes by having an agent scan Slack conversations to locate the right person and request credentials.
• Centralizing secrets in a secure manager and setting up ingestion pipelines so new data was automatically pulled into a central store.

Every time he connected a new system, he had agents crawl it, ingest the data, and look for links to existing data: products, orders, batches, customers, emails, and more. Over time, this created a living knowledge graph of the entire business.

Cleaning and structuring data for AI, not for humans

Once the systems were connected, the next challenge was data chaos: the same product might be labeled slightly differently in the ERP, the email platform, and the ecommerce store. Traditional integration would require painstaking field-by-field mapping.

Instead, he used AI to automate much of the heavy lifting:

• Raw data from each system landed in a central “lake house.”
• AI agents transformed it through stages (bronze → silver → gold), reconciling and validating as they went.
• Data was embedded and vectorized so AI could find connections beyond exact matches.
• An ontology layer sat on top—essentially a map of how concepts relate (product, customer, batch, email, order) rather than just column names.

He didn’t know all of this upfront. He learned it by asking AI: explaining what he wanted (“I need one clean view of a product across systems”) and then asking how to get there, step by step. When the system proposed an approach, he’d ask it to explain how and why, so he could understand and adjust.

Building an AI “Google Maps” for every employee

With the data wired up, he started designing what he calls an operating system for the company—something that feels to employees like Google Maps for their job:

• On the left: the “map” of the company—sales, marketing, product, factory, logistics, finance, HR.
• At the bottom: a chat box, powered by models like Claude or GPT, that can answer questions and take instructions.
• At the top: a mix of dashboards and AI agents that don’t just show metrics, but recommend and execute actions.

For example, a marketing operator might see:

• Current performance across campaigns.
• Flags like “These ads are stale and underperforming” or “This segment is under-emailed.”
• Buttons such as “Refresh these creatives” or “Launch a new test for this audience” that trigger agents to generate assets and push changes directly into ad platforms.

Each time a person approves or tweaks an action, the system learns. After a few repetitions, the agent can be trusted to handle similar tasks autonomously—just as you eventually stop double-checking every turn Google Maps suggests.

The moment everything changed: tripling website conversion

One of the clearest wins came from applying AI directly to the ecommerce funnel. At one point, the website’s conversion rate had dropped to around 0.4%—less than one sale per 200 visitors. The problem was buried in a mix of:

• Slow page loads.
• Poorly ordered scripts and apps on Shopify.
• Friction in the checkout steps.

Instead of waiting weeks for the tech team to investigate, he asked the AI system to:

• Analyze the code and the full funnel.
• Identify where users were dropping off.
• Recommend specific fixes.

Then he did something most CEOs wouldn’t: he let the system implement the fixes. Within about an hour, conversion jumped from 0.4% to 1.7%—more than a 4x improvement. For any brand spending on digital ads, that’s the difference between barely breaking even and printing money.

He argues that any Shopify merchant should be doing something similar: backing up their code, letting AI analyze and optimize it, and treating the whole stack like a time machine they can roll back if needed.

What one month of focused AI work can build

After a particularly intense month of building, he asked the system to estimate what had been created. The answer: the equivalent of 10–12 senior engineers working for 18 months.

In other words, roughly three years of traditional engineering effort compressed into one month of focused work by a single, highly motivated founder using AI as a force multiplier.

This echoes what many power users discover when they go all-in with AI tools—similar to experiments like spending heavily on ChatGPT-driven campaigns or even replacing parts of daily life with AI companions. Once you push past the surface, the leverage curve gets steep very quickly.

Security, compliance, and vendor risk in an AI world

What about security and compliance? The company handles sensitive customer data, payment flows, and internal systems. The CEO’s view is that, used correctly, AI can actually improve security rather than weaken it.

He uses AI agents to:

• Scan code and systems for exposed tokens or credentials.
• Enforce policies like “never expose this type of secret” across repositories.
• Audit for potential compliance issues faster and more thoroughly than a manual review.

He still values human oversight—especially from people who understand cybersecurity—but believes their job is now to define policies and review AI findings, not to manually comb through every file.

On the vendor side, he’s blunt: many SaaS companies are in trouble. Tools that simply sit on top of your data and charge high subscription fees, without adding real intelligence or automation, are easy to replicate with AI.

He’s already seen some vendors respond by:

• Trying to lock customers into aggressive auto-renewals or price hikes.
• Or, more positively, adding genuinely useful AI-powered workflows and interfaces that justify their fees.

His warning to software vendors: if your customers can rebuild your product with AI faster than you can evolve it, you’re at risk. Don’t let your customers become your competitors.

How AI is changing who he hires and promotes

AI hasn’t just changed how the company operates—it’s changing who thrives inside it. When he thinks about hiring now, he focuses less on hard skills and more on:

• Ethics and trustworthiness.
• Will—the grit to learn new tools and push through discomfort.
• Curiosity and a growth mindset.
• Good judgment and taste, especially in ambiguous situations.

Technical skills are still useful, but AI makes them far easier to acquire on the job. What’s harder to teach is the willingness to ask better questions instead of clinging to the identity of being “the person with the answers.”

In his words, people who excel now are those who:

• Use AI constantly as a thinking partner.
• Document what they learn so the system can compound that knowledge.
• Share wins so adoption spreads socially across the company.

He’s seen that once someone has a real “aha” moment with AI—where it solves a painful problem or unlocks a new capability—they rarely go back. They want to do more, sleep less, and keep building.

Preparing for a world of AI agents and autonomous shopping

Looking ahead, he’s already thinking about AI agents not just inside the company, but on the customer side too. Imagine a future where:

• Customers have personal shopping agents that know their style, size, and preferences.
• Those agents talk directly to brand agents that know real-time inventory, production capacity, and delivery windows.
• Most of the discovery and selection happens in a chat interface, not on a traditional homepage.

To prepare, he’s making sure the company’s data and APIs are structured so agents can easily understand products, availability, and fulfillment. That means building vectorized, AI-friendly databases rather than only classic SQL schemas.

At the same time, he believes the fundamental advantage for brands won’t change: you still need to make great products and get them to people reliably. AI can generate beautiful images of shoes, but it can’t stitch them or ship them. Owning manufacturing and direct customer relationships remains critical.

Why every CEO should get hands-on with AI

His closing message to other leaders is direct: stop outsourcing AI strategy.

He argues that:

• If you’re in footwear, you should understand how shoes are made.
• If you’re in fishing gear, you should understand rods.
• And if you’re running any modern company, you should understand how AI can program your business.

Hiring a big-name consultant to “do AI for you” is, in his view, a mistake. Those teams don’t live your business every day, and the new generation of tools is designed to be driven by domain experts, not just engineers.

Instead, he recommends that CEOs:

• Spend real time inside tools like Claude, ChatGPT, or others.
• Start with their own workflows: email, reporting, decision-making.
• Then gradually wire in systems, data, and teams.

It won’t be plug-and-play. There will be a stretch period where it feels confusing and uncomfortable. But the payoff—faster decisions, fewer meetings, less waste, and a smarter company—is hard to ignore.

If you’re still skeptical, it may be worth reflecting on why so many leaders who go all-in with AI end up saying the same thing: they don’t regret the time invested—they regret not starting sooner. For more on why many CEOs are still missing this shift, see our piece on the AI delusion in the executive suite.