How Linker turned Lorenzo Castro into a leading voice in global AI tooling
Geoff Charles, CPO at Ramp, told his entire company they would become the most productive organization in the world. The first thing he did was to dismantle the way the team thought about internal tools. Geoff leads product, data, AI and everything that touches how 800+ people work internally. Before Ramp, he spent a decade in financial services - from advising Fortune 100 companies to scaling credit products to millions of consumers. Today he runs what are probably the most advanced AI-native operations in the tech industry. The results are staggering - 6,300% increase in AI usage, 1,500+ internal apps shipped in six weeks by 800+ builders including non-engineers. Geoff treats internal tooling as infrastructure, not overhead. When Ramp needed everyone to work with AI, his team built Glass - their own internal AI operating system, connected to every system the company uses. It's a custom workspace with persistent memory & reusable skills shared across teams. People who never wrote a line of code are now shipping directly to the production codebase, querying databases in plain language, and cutting research cycles from days to minutes. Geoff's bet is simple: an agent is only as good as the system it operates in. Build the right foundation, and the leverage becomes unlimited. Skip it, and AI becomes just another tool people use occasionally. He also said something that has stayed with me: a short shelf life for internal tools is healthy. It means the system is evolving faster than it can calcify. That's the exact opposite of how most companies think. They treat their tooling as a sunk cost and something to maintain. Ramp treats it as a living asset - one that compounds.
The CEO of Shopify spent his evenings last month optimizing 20 years of legacy code with an AI agent. That detail is more significant than it looks. Tobi Lütke is a German developer who moved to Canada in 2002 with one goal: sell snowboards online. He looked at every e-commerce solution available and found them all inadequate. So he did what any frustrated developer does when the tools fall short - he built his own, in a matter of months. Other merchants started asking to use it. The snowboard business became the side project. Shopify launched in 2006 as the accidental byproduct of a developer who refused to work around bad software. Then Shopify scaled. Thousands of employees, layers of executives, board meetings, press cycles, investor calls. Tobi became a CEO. His calendar filled up with the things CEO calendars fill up with. There was no longer a point in him being in the codebase - the company had hundreds of engineers for that. Founder mode gave way to executive mode, as it does at every company that grows past a certain size. Last month, Tobi ran 120 automated experiments on Shopify's core infrastructure - after hours, on his own time. The result was a 53% performance improvement. The CEO of a public company, back in the codebase, shipping again. AI is giving founders back something scaling took away from them: the ability to stay close to the product without it consuming their entire calendar. My opinion is that it is exciting for the quality of what these companies ship. A founder who still feels the product is a different kind of leader than one who only reads reports about it. But it's going to require real organizational adaptation. More horizontal structures and fewer layers between decisions and execution. Some of those layers exist precisely because the founder couldn't do this before. That changes now.
Discover our other client case studies
Go from invisible to essential in your market
Your expertise deserves to be visible, credible, and shared—on LinkedIn and in the media.
LinkedIn communication and Public Relations agency in Paris.
