Level up your AI workflow

Imagine hiring a world-class executive assistant, then using them exclusively to sharpen pencils. Ridiculous, right?

Obviously they would only use mechanical pencils (kidding).

But, that’s how most of us interact with generative AI today.

In my own work, I’ve had to learn (and relearn) how to use evolving tools like ChatGPT and Claude.

Now, I’ve taken a completely different approach to human-AI interaction that unlocks capabilities most people never discover.

In her recent post, Allie Miller shared the distinct AI usage “modes”, and I was inspired to explore how we can progress through each like a level and find increasing value and impact.

Let’s explore each of these AI "levels" and define how you can advance to the next one.

Level 1: The Task Runner

This is where most of us start and, unfortunately, where many of us stay. We use AI like a digital assistant handling discrete, well-defined tasks:

• “Summarize this quarterly report”
• “Fix the grammar in this email”
• “Convert this table into a different format”

It’s quick, convenient, and transactional. You make a request, get a response, and the interaction ends. The goal here is high precision with minimal overhead.

The Task Runner level is perfect for contained tasks with clear boundaries. It delivers immediate value by saving you minutes on specific activities.

If this works for you, that’s great! But if this is your primary approach, you’re capturing just a small slice of AI’s potential.

Level 2: The Thinking Partner

This is where interactions become more continuous and contextual.

On this level, AI becomes a collaborative companion during extended work sessions. Unlike the brief encounters with the Task Runner, the Thinking Partner maintains context throughout a continuous conversation.

Imagine working alongside a brilliant colleague who remembers everything you’ve discussed, builds on your ideas, and helps you navigate complexity in real-time. For example:

• “Let’s draft this strategic plan together, incorporating our quarterly numbers and competitive analysis”
• “Work with me on this product roadmap, helping me prioritize features based on our customer feedback”
• “Plan my whole business expansion strategy with me”

The key distinction: you remain firmly in the driver’s seat, directing the work while AI amplifies your thinking. The back-and-forth develops a natural flow, and you’ll find yourself rarely hitting creative blocks or decision paralysis.

This level captures perhaps half of AI’s potential today, dramatically improving both the quality and speed of complex knowledge work. Personally, a majority of my work is on this level, using tools like Cursor for coding and Claude to augment my planning and strategy work.

Level 3: The Autonomous Agent

Here’s the cutting-edge, and where you’ll find a lot of today’s AI buzz.

On this level, you provide a clear goal, relevant context, and constraints — then step away while AI works independently on your behalf.

The Autonomous Agent goes beyond responding to prompts. It actively pursues objectives, makes reasonable judgment calls within defined parameters, and delivers comprehensive results. You might direct an AI agent to:

• "Research three main competitors’ marketing initiatives and prepare a comparison report"
• "Monitor our company's customer reviews across platforms and send daily summaries with action items"
• "Run multiple economic scenarios against my banking dataset"

The defining feature is that you walk away after defining the task, allowing AI to self-direct while you focus elsewhere. You should review the final output before making consequential decisions, but your active management isn’t required during the process.

Teams operating on this level frequently find they can accomplish in days what previously took weeks, capturing a whole lot of AI’s potential value. New “reasoning models” like Claude 3.7 Sonnet and advanced systems from OpenAI are making this level increasingly accessible.

How to Level Up Your AI Workflow

Most people remain on level 1, using powerful AI systems like basic search engines. Your existing mental models about delegation and control might limit your progress, but you can break free with deliberate practice.

Ready to unlock the full potential of generative AI? Here are specific actions you can take this week:

1. Track and analyze your AI usage patterns: Review your last 10 AI interactions and categorize them by level. Then identify two opportunities to upgrade:
   • Example: Instead of asking ChatGPT to “summarize this quarterly report” (Level 1), try “Let’s analyze this quarterly report together. First identify the key metrics, then help me develop three strategic recommendations based on these trends” (Level 2).
2. Delegate one complete project to AI this week: Choose tasks with clear success criteria but flexible execution paths:
   • Example: Rather than editing your own customer success stories, provide the raw interview transcript and instruct: “Create three variations of a customer case study from this interview. Format as a one-pager, include relevant quotes, and highlight measurable results” (Level 3).
3. Create a personal AI interaction playbook: Document which level works best for different task types:
   • Example: For creative work, schedule 30-minute Level 2 sessions with instructions like: “Let’s brainstorm product ideas together. I’ll provide feedback on your suggestions and we’ll iterate until we find the right direction.”

Ready to start? I'll audit your AI in 30 mins Learn more