The Agentic Way of Working

Working with AI agents requires a fundamentally different mindset than working with traditional tools. This lesson covers the shift from doing everything yourself to designing systems that do the work for you.

The Old Workflow vs. The Agentic Workflow

Traditional workflow:

1. You receive a task
2. You open the right tools
3. You do the work step by step
4. You check and verify
5. You deliver the result

Agentic workflow:

1. You design the system (tools, rules, triggers)
2. You define what success looks like
3. You deploy the agent
4. The agent does the work
5. You review outcomes and refine

The bottleneck shifts. In the old way, your time and energy are the bottleneck. In the agentic way, system design is the bottleneck. The better your system design, the more the agent can accomplish without you.

The Orchestrator Mindset

When you work with agents, your role changes from executor to orchestrator. You are no longer the one sending emails, formatting reports, or searching for information. Instead, you are the one who:

• Defines the goal: What should the end result look like?
• Selects the tools: Which services does the agent need access to?
• Sets the boundaries: What is the agent allowed to do autonomously? What requires your approval?
• Monitors quality: Are the agent's outputs meeting your standards?
• Iterates on the system: How can the process be improved?

This is similar to managing a team. You do not write every line of code yourself — you set the direction, provide resources, review output, and course-correct when needed.

Practical Example: Morning Briefing

Consider a daily morning briefing. The manual approach:

1. Open news sites, scan headlines (10 min)
2. Check email, flag important items (15 min)
3. Review calendar, identify conflicts (5 min)
4. Check project management tool for blockers (10 min)
5. Write summary for yourself (10 min)

Total: ~50 minutes every morning.

The agentic approach:

1. Design once: Configure an agent to scan RSS feeds, email, calendar, and project boards
2. Set schedule: Run every morning at 6:00 AM
3. Define format: Deliver a structured briefing via Telegram
4. Deploy: The agent runs automatically

Now your 50-minute routine takes 2 minutes to review what the agent prepared.

What Changes — and What Does Not

Changes	Does Not Change
Who executes the work	Your responsibility for the outcome
How fast tasks complete	The need for quality checks
How many tasks run in parallel	Your domain expertise requirements
When tasks can run (24/7)	The importance of clear instructions

Agents amplify your capability, but they do not replace your judgment. You still need to know what good output looks like, because you are the one defining "good" for the system.

Key takeaway: The agentic way of working shifts your role from executor to designer. Your job is no longer to do the work — it is to design the system that does the work. The better your system design, the more autonomous the agent becomes.

Next: How to evaluate current agent frameworks realistically — what works today and what does not. :::