This article explains how to write effective prompts when using OI-Q and OI-Q²
Jump down:
Advanced Uses
OI-Q² Prompts
OI-Q Introduction
OI-Q tools analyse customer conversations and return insights based on the transcripts selected. The quality of the response depends on the quality of your prompt, so clear and structured questions produce the best results.
The transcript on screen is the source of information, so there's no need to include “read the transcript” in your prompt.
If you're unsure of how to get to this screen, please see OI-Q Overview.
To get the best results, follow these prompting principles:
Be clear and specific
Ask direct questions. The more precise your question, the more useful the answer will be.
"Did the customer appear frustrated on this call? Respond in 50 words or less.”
Provide context when it helps
Additional background may help to interpret the conversation and yield the result you want.
“This customer experienced a failure with our online app. Identify what the agent could have done to improve the experience.”
Define the format of the response
If you want a structured output, tell OI-Q how to present it.
“How could this interaction have been handled better? Provide the top 3 opportunities in bullet points in no more than 250 words.”
Extract verbatim information
If you need exact wording from the transcript, ask for it.
“Tell me the exact words the customer used when describing their product issue."
Use examples where needed
Providing reference terms can help OI-Q identify specific concepts.
“Does the customer mention the XYZ product? Answer Yes or No.”
Refine your prompt
AI responses improve with iteration. Small changes to prompts can significantly improve the result.
If the first response isn’t quite right try:
- Adjust the wording
- Add clarity
- Specify the format of the response
Use AI to help build prompts
If you're unsure how to structure a prompt, tools like Microsoft Copilot can help generate or refine one. You can paste your inital transcript question into these tools and ask them to suggest a clearer prompt for OI-Q.
Save and share effective prompts
Once you’ve developed a prompt that consistently produces useful insights:
• Save it for future use
• Share it with your team
Well-designed prompts can become repeatable analysis tools across the contact centre.
Ask one question at a time
OI-Q² processes all selected transcripts together. For the most reliable results:
• Ask one question per query
• After receiving the answer, start a new query for the next question
Keep responses focused
Shorter responses tend to produce clearer insights.
“Summarise the main customer complaints in 100 words or less.”
Advanced Uses
OI-Q can also assist with deeper analysis of conversations.
Estimate time spent on tasks
“Tell me how long the agent spent verifying the customer on this call.”
Validate calculations
“Across these 75 interactions, what percentage of customers mentioned login issues?”
Creative summaries
“Summarise this interaction as a short script between the agent and customer.”
Using OI-Q² for Summarising Multiple Interactions
OI-Q² is the second step after running an initial multi-select OI-Q prompt. It analyses the responses to your original prompt and returns a single summary or insight.
This is useful for identifying themes or patterns across a group of interactions at scale.
| Initial OI-Q Prompt | OI-Q Prompt Results | OI-Q² Prompt | OI-Q² Prompt Result |
|---|---|---|---|
| Tell me in 25 words or less the reason for the customer support contact | Transcript 1: [reason] Transcript 2: [reason] ... Transcript 100: [reason] |
Tell me in 300 words or less the top 5 reasons for the customer support contact |
Based on the provided data, the top 5 reasons for customers contacting support are:
|
💡Make sure to click the OI-Q² button before you enter your OI-Q² prompt!
Current Limitations
OI-Q currently works only with the transcript text itself.
It does not have access to metadata such as:
- Agent names
- Handle time
- Call duration
- System data
These capabilities may be introduced in future updates.
Key Takeaway
The most important rule of prompting is simple:
The output is only as good as the prompt.
Clear, structured prompts will produce the most useful insights from OI-Q and OI-Q².
Comments
0 comments
Please sign in to leave a comment.