APPENDIX B
The 4-Level Gen AI Ladder
The 4-Level Gen AI Ladder
A framework for knowing where you are and what’s next. You don’t need to reach Level 4. Most of the value sits in Levels 1 and 2.
Level 1 | Basic Prompting
You open an AI tool, type a request, get a response. Each conversation starts fresh. No memory, no saved preferences, just you and a blank prompt box. This is where everyone begins.
What it looks like: “Write me an executive summary for this report.” “Help me draft an email to my director.” “Summarise this policy document.” You’re using the tool as a smarter search engine or a writing assistant. One prompt, one response.
The unlock: Learn RTCO, which stands for Role, Task, Context, Output. Tell the AI who you are, what you need, why you need it, and what the output should look like. Your results improve dramatically.
Ready for Level 2? If you’re typing the same context at the start of every prompt, you’ve outgrown this level.
Level 2 | Gen AI Agents (Role-Based Assistants)
You’ve built AI agents that know their role. Gems in Gemini, Custom GPTs in ChatGPT, Projects in Claude, each one pre-loaded with your context, your format, your standards. You teach the agent once; it remembers. You stop repeating yourself.
What it looks like: A Gem that formats meeting minutes in your ministry’s standard format. A Claude Project pre-loaded with your company’s brand guidelines and past reports. A Custom GPT that generates job descriptions in your HR framework. These are your digital staff officers: smart, fast, but they only know what you’ve told them.
The unlock: You start thinking in systems. “What do I do every week that follows the same pattern?” becomes your design question. Each agent you build removes one more repetitive task from your day.
Ready for Level 3? When you catch yourself wishing two tools could talk to each other automatically.
Level 3 | Gen AI-Powered Automation
You’ve connected tools together. Data flows automatically. A spreadsheet updates and certificates generate themselves. A form submission triggers a summary email. The pieces talk to each other without you copying and pasting between them.
What it looks like: The certificate automation from Chapter 6. Names from a Google Sheet, merged into Slides templates, exported as PDFs. One click, fifty certificates. Or: meeting audio uploaded to NotebookLM, summary auto-generated, action items pushed to a shared document.
The unlock: This level requires some technical comfort, or access to someone technical. You don’t need to code it yourself. Find your Aiman. Show them what you need. Let them build the bridge.
See Appendix C for a complete walkthrough of a Level 3 automation.
Level 4 | Agentic AI
AI systems that plan, execute, and adapt. They don’t just respond to prompts; they break down complex tasks, use multiple tools, make decisions along the way, and course-correct when something doesn’t work.
What it looks like: An AI agent that monitors your inbox for procurement requests, drafts responses based on policy guidelines, flags exceptions for human review, and tracks approval status. Not a chatbot. A collaborator that handles multi-step workflows end to end.
Reality check: This level is advancing fast but is not yet practical for most knowledge workers. The technology exists. The reliability and governance frameworks are still catching up. Watch this space.
Don’t let Level 4 intimidate you into inaction at Level 1.
The officer who builds one good Gem and uses it every day has already transformed their workflow. The manager who learns RTCO and applies it consistently will save hundreds of hours a year. You don’t need to be on the frontier to escape Work Drift.