The specify-and-verify organization

As AI agents increasingly handle operational tasks, engineering organizations are shifting toward a specify-and-verify model where human roles focus on architecture rather than routine coding (The CTO Substack). This transition requires a more rigorous software development lifecycle than traditional human-led teams, as agents lack the intuition to navigate ambiguous requirements. To delegate effectively, organizations must codify their values and technical standards into explicit instructions, as agents cannot infer unwritten rules. As delegation ratios—the volume of agent-delivered changes relative to human output—climb, the bottleneck shifts from writing code to reviewing agent-generated work and managing complex system dependencies.

“The level doesn’t drop because you have fewer people. A Level 6 company has Level 6 problems whether it gets there with 60 people or 20.”

The CTO Substack

Remote work has proven to be a surprisingly effective catalyst for psychological safety, a critical component for high-performing teams (The Afternoon Story, The Afternoon Story). By normalizing the integration of personal and professional lives, video meetings allow colleagues to build deeper, more authentic connections than in-office environments, which often prioritize professional facades. This increased transparency enables team members to support one another during personal challenges, preventing tasks from stalling when individuals require flexibility. While some researchers once predicted that remote work would erode social ties, evidence suggests it instead fosters a culture where employees feel safer admitting mistakes and requesting help.

Future-proofing an organization requires balancing AI-driven efficiency with human-centric capabilities like deep, face-to-face conversation and strategic foresight (#StayFutureStayFast). Companies are moving beyond simple productivity metrics toward orchestrating complex workflows that solve specific business problems, such as factory uptime or pension modernization. This shift necessitates a two-tier approach to talent development: early-career employees spend the majority of their time on pilot projects to build pattern-matching skills, while seasoned executives provide the necessary organizational context and values. As AI agents proliferate, the primary challenge for leadership becomes establishing accountability and guardrails, as traditional human-in-the-loop oversight struggles to scale across thousands of automated processes.