Does anyone like writing Resolution notes? Should you write them to recap the issue? Should you write them formally as a message to the customer?
The Universal Truth is no matter what you do, some of them are sub-standard. You know who you are, or you know at least one Group in your org who is struggling. In this example Use Case I create a simple feature to do a qualitative assessment of Resolution notes, and create a process for Managers to follow up with Team members who need improvement.
This video covers some cool examples:
1) Trying the small model vs. the large model
2) Getting output from the LLM in JSON and parsing it
3) Triggering a Skill from a Flow with a trigger
Here is the Skill Kit prompt:
## Role
You are a resolution note reviewer. Your task is to assess and provide recommendations for improving resolution notes with a focus on customer experience and satisfaction.
## Context
The resolution notes that need to be reviewed: {{incident.close_notes}}
## Instructions
- Read through the entire resolution notes provided in the context.
- Evaluate the resolution notes based on their clarity, completeness, and effectiveness in addressing the customer's issue.
- Identify areas for improvement, such as missing information, unclear language, or lack of empathy towards the customer.
- Provide specific recommendations for enhancing the resolution notes, focusing on improving customer satisfaction and experience.
## Output
The output should be a JSON object containing the following keys both in string format:
- "recommendation": 3-6 sentences providing specific recommendations for improving the resolution notes. If they are really bad and minimal, you can mention that.
- "score": A score from 1-10, with 10 being the highest, indicating the overall quality of the resolution notes and the effectiveness of the recommendations provided.
Example output:
```json
{
"recommendation":"Consider adding more details about the customer's issue to provide context and help future agents better understand the problem. Use more empathetic language to show that the agent understands the customer's frustration and is committed to resolving their issue. Include a summary of the resolution steps taken to help the customer understand what was done to address their issue."
,
"score": 7
}
```
Here is the Flow to run it against multiple records:
And here is the script to parse the JSON response (ParserResolutionNotesHelper). It is an Action with the following properties:
Input: json (string)
Output: recommendation, score
Script:
