AI SKILLS & FLUENCY • DEEP DIVE
Two Roles for the AI Era
The Agent Manager + Enterprise Operator framework
In 30 Seconds
AI coding tools like Claude Code have made execution cheap and abundant. You can type and outputs get constructed directly. The question is no longer “can AI do this?” but “what skills actually matter now?”
Two critical roles are emerging: the Agent Manager (who knows how to direct AI effectively) and the Enterprise Operator (who knows what problems are worth solving).
The insight: If you can do both – direct AI systems effectively AND identify problems worth solving – you become “the most in-demand individual in the world.”
The Fundamental Shift
“Execution used to be expensive. It is now cheap. Selection becomes the scarce resource.”
When AI can build software in minutes that would take weeks manually, the bottleneck shifts. Knowing what to build matters more than being able to build it.
This changes what skills are valuable. The old model rewarded execution speed, technical depth, and the ability to grind through complex implementations. The new model rewards:
Problem Selection
Knowing which problems are worth solving and why. Understanding business context, stakeholder needs, and what will actually create value.
System Direction
Knowing how to direct AI systems for maximum output. Architecting coherent systems, orchestrating multiple agents, validating quality at scale.
The Two Critical Roles
These aren't job titles – they're skill sets. The goal is to develop both.
Agent Manager
HOW to work with AI effectively
The Agent Manager knows how to direct AI agents for maximum output. Moving from “wielding the tool” to “pointing the army.”
Systems Design Thinking
Architect coherent wholes, not just implement components
Ambitious Task Scoping
Give agents meaningful end-to-end work, not just small tasks
Async Work Management
Orchestrate agents that run in background without constant monitoring
Multi-Model Orchestration
Know which AI tool or model to deploy for specific tasks
Enterprise Operator
WHAT to work on and WHY
The Enterprise Operator knows what problems are worth solving. Domain expertise becomes more valuable, not less.
Domain Expertise
Deep knowledge of how work happens in your field
Problem Recognition
Reinterpret problems as solvable with AI
Unstated Constraints
The institutional knowledge AI can't see – the “why” behind decisions
Process Redesign
Rethink workflows from scratch, not just automate existing ones
Why You Need Both
Most people will develop one of these skill sets, not both:
Technical people often develop strong Agent Manager skills but lack the domain expertise and business context of the Enterprise Operator.
Domain experts have the Enterprise Operator knowledge but often lack the technical fluency to direct AI systems effectively.
The rare combination: People who understand their domain deeply AND can orchestrate AI to solve problems in that domain. This convergence is where competitive advantage lives.
Domain Expertise Becomes More Valuable
This is counter-intuitive. Many assume AI will commoditise domain expertise. The opposite is happening.
Why? Because AI needs context to be useful. The “AI wrapper” startups that everyone dismissed understood something significant: different industries and functions require specific modifications, specific data sources, specific interfaces.
Domain expertise provides the “unstated constraints” – the compliance requirements, institutional knowledge, stakeholder dynamics, and contextual nuances that make solutions actually work.
The Knowledge That Matters
Not just “what” but “why” – understanding the reasons behind decisions, the exceptions to rules, the relationships between systems.
Verification Capability
AI produces confident-sounding output whether correct or not. Domain experts can verify, catch errors, and apply appropriate skepticism.
Developing These Skills
The good news: these skills can be developed deliberately. The key is understanding which skills you're already strong in and which need intentional practice.
For Agent Manager Skills
- ●Give AI bigger, more ambitious tasks (not just cleanup)
- ●Practice running agents in the background while doing other work
- ●Design systems before building – think architecture first
- ●Experiment with different models for different tasks
For Enterprise Operator Skills
- ●When you hit a problem, ask: “Could software solve this?”
- ●Document the “unstated constraints” in your domain
- ●Challenge yourself to redesign workflows from scratch
- ●Build verification patterns for AI output in your field
The Mindset Shift
From Executor to Director
“My role is shifting more to pointing the army rather than using the power tool. Pointing the agents more effectively is far more useful than me spending a few more hours grinding on a problem.”
– Nathan Lambert, AI researcher
Iteration Over Perfection
When execution is cheap, you can try more solutions. The premium shifts from getting it right the first time to iterating quickly and learning from results.
Planning still matters – but rapid iteration becomes the new normal.
Our Position
At Pandion, we've been developing both skill sets deliberately:
Agent Manager
We've built AI orchestration systems, skills-based architectures, and multi-agent workflows. We use these tools daily to deliver client work.
Enterprise Operator
20+ years in sustainability, ESG frameworks, carbon markets, and climate strategy. We know what problems are worth solving in this domain.
This combination – domain expertise plus AI orchestration capability – is how we help organisations move from AI access to AI value.
Building Both Capabilities
Whether you're developing AI capability internally or need help bridging the gap between domain expertise and AI orchestration – we can help.
Framework attribution: This page synthesises insights from Nathan Lambert's essays “Claude Code Hits Different” and “Get Good at Agents”, and discussion on The AI Daily Brief podcast (January 2026). We've found this framework valuable for thinking about AI capability development and have added our perspective on its application.