What “Error Handling” Actually Means
You’ve seen it. You ask the AI a question, it says “I don’t have enough information to answer that” and stops. Nothing else. You’re back where you started.
Error handling is the instruction you add that says: don’t stop, give me what you’ve got.
❌ What most people do
They ask the question and hope.
What coaching models work best for instructional coaches working with resistant teachers?Two things happen. Either the AI produces a generic list that reads like a 2015 textbook, with no sign it’s uncertain about any of it. Or it says it doesn’t have enough context and offers nothing.
✅ What to do instead
Add a fallback instruction at the end. Tell the AI what to do when it can’t give a full answer.
Prompt:
What coaching models work best for instructional coaches working with resistant teachers?
If you don't have complete information:
1. Provide what you do know
2. Clearly state what's missing
3. Suggest where to find the missing information
4. Give your best analysis with the data you have
Never say "I don't have enough information" without attempting a partial answer.Three things to try right now
Go back to a question you asked the AI that got a vague or dead-end response. Add the fallback block above and send it again.
For any evaluation task,a workshop plan, a session design, a curriculum map, ask the AI to include a confidence rating: “Give me your assessment and tell me your confidence from 1–10, with your reasoning.” The number forces it to show its thinking instead of hiding behind a confident-sounding answer.
Use this on any research or policy question: “If you can’t find a direct answer, tell me what you do know, flag what’s missing, and suggest where I should look next.” Works on anything where the AI would otherwise hedge and stop.
How to save it for next time
Paste this into your Claude Project instructions. Every prompt you send in that Project carries this behaviour automatically no need to add it again.
Copy this block:
When you can't complete a full request or are uncertain:
- Provide what you do know
- Clearly state what's missing
- Suggest where to find the missing information
- Give your best analysis with the available data
When uncertain about a specific claim, include a confidence level (1–10) and your reasoning.
Never say "I don't have enough information" without attempting a partial answer first.Open Claude, go to your Project instructions, and paste that block in now.
Phil