- 01 Nov 2024
- 2 Minutes to read
- Print
- DarkLight
19.2.1. How to make use of AI Assist
- Updated on 01 Nov 2024
- 2 Minutes to read
- Print
- DarkLight
With AI Assist you’re able to employ the power of generative AI to achieve a series of goals such as :
Auto Wrap-up
Save time by summarizing key points and outcomes at the end of call
Auto-Update Fields
Ensure accurate record keeping with fields updated automatically based on call details
Suggest Next Actions
Get proactive guidance on what steps to take next to keep deals or service resolutions on track.
Generate Follow-up Content
Quickly generate tailored follow-up emails to and tasks to maintain customer engagement after a call.
In this article we’ll walk through the example of Auto-Updating a field whereby you can save your agents the time of having to manually complete certain admin at the end of the call.
Please see the prompt library for more details on the other use cases that can be achieved with AI Assist.
Defining your Fields
Firstly you need to identify the fields that need to be updated as part of the user’s call process, the service works best if each element of the process has it’s own dedicated custom field in Salesforce.
For example :
Support ticket
Product
Issue summary
Once identified and defined you’ll need to create a prompt per field.
Creating the prompts
Next up you will need to create the prompts for each of the fields for the AI populate.
Lets take ‘Support ticket’ as our example, the prompt could look like :
Please analyze the provided phone call transcript between the agent and the customer. Scan the transcript and extract any ticket numbers. You can identify ticket numbers in the transcript as they will start with a XYZ followed by further digits. <OUTPUT> -If a ticket number was mentioned and it began with XYZ, output the ticket number only. -If the ticket number was not present, return "Ticket not given" only. </OUTPUT> |
---|
Note for the above example, you will just need to amend the XYZ with your expected Support ticket format for the AI to be able to extract the tickets.
Follow this guide here for a step by step on prompt creation.
When creating these prompts just be sure to set the Type to ‘Free Text’ as this will allow any kind of query to be made to the AI.
Configure a Salesforce Flow
For the final step you need to configure a Salesforce Flow to take the data created after a call and populate it within your target object (in this example a Salesforce Task).
So the flow itself will trigger when the prompt responses appear within the Insight records and then in this example it will populate the associated custom fields on the Task object in Salesforce.
To configure a Flow like this see here.
End Results
Now as part of your agent’s calls a part of their admin should be automatically handled after hanging up.
This should be visible from the Task or Voice Call records depending on how your organisation is configured.
FAQ
Is there a limit on the amount of data that AI Assist can return back?
Yes currently there’s a limit of 10500 characters which translates to roughly 2,200.
If you find that your prompt is exceeding this amount, it’s recommended you look to optimising the prompt and potentially breaking any large tasks out into separate prompts, see here for more.
How long before a call is automatically wrapped up?
It can take roughly up to 15 minutes to retrieve the results back from the AI.