Platform Awareness and Task Delegation
Know what today's AI tools do well and badly, then divide the work between you and AI with a simple decision check you can reuse on any task.
Platform Awareness: Know Your Collaborator
You wouldn’t ask a talented graphic designer to audit your taxes. Platform awareness means knowing the capabilities and limits of the AI tools available to you, so you match the task to the tool.
Specific features change quickly, so focus on patterns that have held up across tools and model generations:
| Generally strong | Generally weak or risky |
|---|---|
| Drafting, rewriting, and changing tone | Exact facts, figures, quotes, and citations without a source to check against |
| Summarizing text you provide | Very recent events, unless the tool can search the web (and even then, verify) |
| Brainstorming many options quickly | Knowing your organization’s unwritten context, politics, or history |
| Explaining concepts at different levels | Precise arithmetic in long calculations, unless the tool runs code |
| Restructuring information (tables, outlines, formats) | Judgments that need lived experience or accountability |
| Giving critique from a stated perspective | Consistency across very long documents without careful guidance |
Platform awareness also means knowing your specific tool’s setup:
- Can it search the web, read files, or run code? These change what it can verify on its own.
- What does it remember? Some tools keep memory across chats; others start fresh each time.
- What are the data terms? A personal free account and your organization’s approved tool may handle your data very differently.
- What’s approved where you work or study? That’s a diligence question too, and it constrains your choice.
Common mistake: deciding “AI can’t do X” based on a disappointing attempt months ago, or assuming a tool can do something because a different tool could.
Capabilities shift with every model update, and they vary between products. When a task matters, run a quick test with the specific tool you plan to use rather than relying on memory or reputation.
Task Delegation: Dividing the Work
With problem awareness and platform awareness in hand, you can make the actual call. For each piece of work, run this delegation check:
- Fit: Is this something AI tends to do well (from the table above)?
- Checkability: Can I verify the result quickly and reliably?
- Stakes: What happens if it’s subtly wrong and I miss it?
- Ownership: Do I need to do this myself to learn, decide, or stand behind it personally?
- Permission: Am I allowed to use AI here, with this data?
| Fit | Checkable | Stakes | Typical call |
|---|---|---|---|
| High | Easily | Low | Automate. Spot-check the result. |
| High | With effort | Medium or high | Augment. AI drafts, you verify line by line. |
| Low | — | Any | Do it yourself, maybe with AI as a sounding board. |
| Any | — | Any, but you need to own it | Do it yourself. Use AI only for feedback afterward. |
| Any | — | Any, but not permitted | Don’t use AI for this, or find an approved route. |
The check takes 30 seconds once it’s a habit, and it prevents most of the painful AI mistakes people make at work.
- Choose one piece of work from your task map in the last lesson that you marked Automate or Augment.
- Run the five-question delegation check on it. Write a one-line answer for each.
- Now test platform awareness. Ask your AI assistant:
I'm planning to use you for this task: [describe it]. Before starting, tell me honestly: which parts are you likely to do well, which parts are you likely to get wrong, and what should I double-check? Also tell me whether you can search the web or read files in this chat. - Compare its self-assessment with your own check. Treat its answer as a hint, not proof: AI can be wrong about its own abilities too.
- Do a small trial run. Did the result match what you both predicted?
1. Which task is riskiest to delegate to a general AI chat without any source documents?
Exact facts and figures about specific events are where AI is most likely to produce confident errors, especially without a source to check against.
2. Why does 'checkability' matter in the delegation check?
Delegation is safest when you can confirm the output is right. When checking is hard, shift toward augmentation, with you verifying closely, or doing it yourself.
3. Your AI assistant tells you it's very good at citing legal cases. What's the fluent response?
AI self-assessments can be wrong. Platform awareness comes from testing the specific tool and verifying high-stakes details, not from taking its word for it.
Up Next
Once you’ve decided what AI should do, you need to explain it well. Module 3 covers Description, which goes beyond prompt templates to describing the product, the process, and the way you want AI to behave.