Key Takeaways

  • Barney Hussey-Yeo, CEO of Cleo, now uses AI agents to bypass traditional reporting structures, gaining direct, unfiltered insights into his engineering team's output.
  • These AI agents synthesize information from tools like Slack, Notion, and GitHub pull requests, creating a dynamic knowledge base of what's being shipped and by whom.
  • This direct visibility allows Hussey-Yeo to know “who is the highest velocity, highest code quality” among his 500-person team, often before his engineering managers do.
  • The result is a push towards flatter organizational structures, with the potential to significantly reduce the need for middle managers whose primary role was information aggregation and reporting.
  • This shift promises a “more visceral clear way” of understanding organizational activity, leading to higher innovation velocity and a direct link between executive decision-making and operational output.

The CEO with X-Ray Vision

Barney Hussey-Yeo, founder of the AI financial assistant Cleo, sees his organization differently than most CEOs. Instead of relying on layers of reporting, he taps into a direct, unfiltered data stream, thanks to AI agents. He explained that before LLMs, managing a company of “500 people and you'd have all these division leaders and you talk to the division leaders and you'd go farm for information, give them your opinions.” That's the old way.

Now, Hussey-Yeo's day-to-day is dramatically altered. He has AI agents constantly scanning and synthesizing activity across his team's communication and development tools. These bots go over “every PR and it summarizes this is what that PR does and it grades it and it tells me what's going on that PR,” he said. They also crawl Slack messages and Notion documents, building a live, comprehensive knowledge base of everything happening across the company. This means Hussey-Yeo knows “what we're shipping in a really visceral clear way which I didn't know before LM.”

This isn't just about reading more data; it's about seeing the ground truth directly. He isn't relying on summaries filtered through multiple managerial layers. He gets an unvarnished view of progress, bottlenecks, and individual contributions. It gives him a level of operational clarity most founders can only dream of.

The End of Middle Management (Or So He Hopes)

This newfound visibility has a profound impact on organizational design. When a CEO can see who's doing what and how well they're doing it, the traditional role of middle management begins to shift. Hussey-Yeo isn't shy about the consequences for his team: “My engineering managers hate me. Like honestly, like I because I I know I know who's shipping. I know like who is the highest velocity, highest code quality. I know which teams because it's a collection of engineers are actually doing great work.”

He can identify top performers, assess code quality, and understand team velocity directly from the source data. This bypasses the need for managers whose primary function was aggregating information, translating directives, and monitoring individual performance. The AI agent becomes the ultimate performance reviewer and progress tracker, leaving managers to focus on coaching, mentorship, and strategic alignment rather than simply reporting up.

Ultimately, Hussey-Yeo believes this leads to flatter structures. “I really hope you're going to have less middle managers, as and I think orgs are going to definitely become flatter,” he shared. For ambitious founders, this isn't just a prediction; it's a blueprint for a leaner, faster, and more transparent organization where innovation velocity isn't hampered by information silos.

What to Do With This

Stop waiting for reports. Pick one critical area in your business—say, engineering velocity or marketing content output—and identify the raw data sources: GitHub, Jira, Slack channels, Notion documents. Start small: experiment with an open-source LLM or a simple API call to summarize activity from these sources into a daily digest for just one project or team. Your goal is to see the raw output, not the curated summary. Identify what information you're currently receiving through a manager that could be synthesized directly by an AI, and then challenge the necessity of that intermediary layer.