Why Executives Should Use Claude Code

The following is an edited transcript of a conversation between Matthew Hall and Sam Gaddis, founder of Runpoint Partners, recorded in Austin, TX in December 2025.

---

MC: Let's start with the provocation. You think executives—CEOs, CFOs, COOs—should be using Claude Code. Not ChatGPT. Not some enterprise AI dashboard. A command-line tool. Why?

SG: Because there's a massive gap between what executives think AI can do and what it actually does. And the gap exists because they've only ever seen the magic show version. The demo. The polished output. They've never watched the sausage get made.

Claude Code closes that gap. When you're sitting in a terminal watching an AI agent read your codebase, make decisions about what to change, run tests, and iterate—you develop an intuition for what's actually happening. And that intuition is worth more than any vendor pitch deck.

MC: But most executives can't code. Isn't this like telling a CEO to learn welding so they can better manage a construction company?

SG: That's the analogy everyone reaches for, and it's wrong. Welding is a specialized manual skill. Using Claude Code is more like reading financial statements. You don't need to be a CPA to read a balance sheet, but if you can't, you're flying blind.

The executives I work with aren't writing code. They're watching an AI agent work through problems. They're seeing how it reasons about their business logic. They're understanding, for the first time, what "context window" actually means in practice—not as an abstract concept from a slide deck.

MC: Walk me through what this looks like in practice.

SG: Last month I sat down with a CFO who was skeptical about an AI automation project his team had proposed. Budget was $400K. Timeline was six months. He couldn't evaluate whether that was reasonable because he had no mental model for what AI development actually involves.

So we opened Claude Code on his laptop. I pointed it at a simplified version of the problem—a data reconciliation workflow his team does manually. In about twenty minutes, we watched Claude Code analyze the existing process, build an automation, test it against sample data, and iterate when the first approach didn't handle edge cases.

He didn't write a line of code. But he watched the AI reason through the problem. He saw it make mistakes and recover. He saw how much of the work was handling messy real-world data versus the "AI magic" part.

By the end, he had completely recalibrated his expectations. The $400K project? He approved a $150K version with a two-month timeline because he now understood what was actually hard and what wasn't.

MC: You saved the company money by making the executive more technically literate?

SG: I saved the company from a bad decision made in ignorance. The original proposal wasn't fraudulent—the team genuinely believed they needed six months and $400K. But they were padding for uncertainty because nobody in the room could reality-check the technical assumptions.

When the decision-maker can engage with the actual technology, even at a high level, the entire planning process gets more honest. Engineers stop gold-plating because they know the CEO will ask why a simple problem needs a complex solution. Vendors stop hand-waving because the buyer can distinguish real capabilities from vaporware.

MC: What about the argument that executives should focus on strategy, not tools?

SG: Strategy without operational understanding is just wishful thinking. I've seen too many "AI strategies" that are essentially PowerPoint fiction. Three-year roadmaps that don't account for the fact that the technology landscape changes every three months.

The executives who use tools like Claude Code don't become engineers. They become better executives. They ask better questions. They make faster decisions. They can distinguish between a team that's stuck and a team that's stalling.

MC: Is there a risk of micromanagement? Executives who think they understand the technology better than they do?

SG: Absolutely. And I tell every executive I work with the same thing: the goal isn't to do your team's job. It's to understand their constraints. There's a difference between a CEO who rewrites pull requests and a CEO who understands why a migration is going to take three weeks instead of three days.

The best analogy I have is test-driving a car. You don't test-drive a car to become a mechanic. You do it to understand what you're buying. Executives need to test-drive AI to understand what they're investing in.

MC: Any advice for an executive reading this who's curious but intimidated?

SG: Start small. Pick a task you do manually—analyzing a report, processing data, drafting a document. Sit down with someone technical on your team and work through it in Claude Code together. Don't try to learn it alone. The goal isn't to become proficient. The goal is to develop intuition.

And then do it again next week with a different problem. Within a month, you'll be asking fundamentally different questions in your AI strategy meetings. That's when things start to change.

MC: Last question. What happens to the organizations where leadership refuses to engage with the technology?

SG: They become the organizations that other organizations disrupt. Not because they're stupid. Because they're making decisions about a technology they've never actually touched. And in a world where AI capability is doubling every few months, that ignorance compounds fast.

The CEOs who are using Claude Code today aren't doing it because it's fun, though it kind of is. They're doing it because they've realized that delegating all technical understanding to their team is the same as delegating all financial understanding to their CFO. You can do it, but you probably shouldn't.