AI Auto-Fill

What is AI Auto-Fill?

Goldie Conversation Capture logs emails and meeting transcripts in an AI ready format in Salesforce. We refer to the complete log of emails and meetings belonging to an Account as a conversation log.

Goldie's AI Auto-Fill is a feature that can automatically populate Salesforce fields on the Account and Opportunity object - based on the account's conversation log. The field values are generated by an LLM, grounded in:

1. the conversation log

2. the field label

3. the field help text - the LLM understands the semantics of the field by reading the help text.

AI Auto-Fill re-evaluates field values each time an email is captured or a meeting transcript is logged by Goldie's Conversation Capture. It can also manually be triggered through the UI (see AI Auto-Fill on Demand below). If the AI can't infer a proper field value, the field is left blank.

Configuring AI Auto-Fill is as easy as adding the ✨ symbol to the field label and adding instructions for the LLM in the field help text.

How is AI Auto-Fill Useful?

AI Auto-Fill is beneficial in use cases where Salesforce fields should be populated according to what happens in conversations (emails and meetings). This makes sense because a good amount of CRM data is based on what's said in meetings and written in emails. Examples are Salesforce fields you created for deal qualification (e.g. BANT), conversation summaries, and calculating metrics fields - anything that can be inferred from emails and meeting transcripts. AI Auto-Fill gives you these benefits:

1. Users save time since they don't have to remember to update fields after every meeting or email.

2. Qualification and summary fields are reliably populated with unbiased content.

3. You can avoid writing code in order to calculate activity metrics.

Examples

Example 1: BANT Deal Qualification

BANT is a lightweight deal qualification framework that helps determine whether a potential customer is a good fit for your product or service. In this example, we're creating four AI Auto-Fill fields on the opportunity object that Goldie re-populates each time a new email or a meeting transcript is logged under an account and its respective opportunity:

Field Name: Budget ✨
Field Type: Text (600 characters)
Field Help Text: BANT qualification "Budget": According to our conversations, does the customer have the budget? What's their budget? Can they allocate money for the purchase? How much money do they currently spend on this issue? Have they tried any other options and how much did they cost? What ROI do they expect?

Field Name: Authority ✨
Field Type: Text (600 characters)
Field Help Text: BANT qualification "Authority": According to our conversations, do the people we are talking to have the authority to make a purchasing decision? Who are the decision makers and who has sign-off authority? Who will make use of our offering? What procedures must be followed in order to approve such a purchase? How did the last/similar decision-making process go? Do we need to include other people with sign-off authority?

Field Name: Need ✨
Field Type: Text (600 characters)
Field Help Text: BANT qualification "Need": According to our conversations, does your prospect have a need for your product or service? What are the pain points they're experiencing and would a purchase address the pain points? What difficulties are they now experiencing? What actions did they take to address this issue?  How critical is solving this issue? What will happen if they don't find a solution to this issue?

Field Name: Timing ✨
Field Type: Text (600 characters)
Field Help Text: BANT qualification "Timing": According to our conversations, does the prospect have a timeline for making a decision? When do they need to make a decision by? What is the urgency of this issue? What are the consequences of not solving this issue in time? Are there any external factors that could affect the timing of this decision? Are there any compelling events that could accelerate the decision-making process?

Example 2: Technical Summary

In this example, we are creating a field for sales engineers and support staff. We are creating an AI auto-fill field on the account object that summarizes all technical issues the account encountered each time a new email or meeting transcript is logged under an account.

Field Name: Technical Issues ✨
Field Type: Text (600 characters)
Field Help Text: According to our conversations, this is a list of summaries of all technical issues we encountered with this customer. 

Note: If the user has further questions on technical issues, they can use Goldie Conversation Q&A.

Example 3: Activity Metrics

In this example, we are creating two AI Auto-Fill fields and one formula field. The formula field gives us an effort/engagement metric by dividing the number of inbound emails by the number of outbound emails - a metric that helps determine "how engaged" a customer is. Counting emails would usually require us to write code, but we let Goldie count emails for us.

Field Name: Inbound Emails ✨
Field Type: Number
Field Help Text: This is the number of inbound emails - i.e. emails we received from the customer.

Field Name: Outbound Emails ✨
Field Type: Number
Field Help Text: This is the number of outbound emails - i.e. emails we sent to the customer.

Field Name: Email Engagement
Field Type: Formula (Percent)
Formula:  IF(Outbound_Emails__c>=Inbound_Emails__c,MIN(Inbound_Emails__c/ Outbound_Emails__c,1),1)

The field Email Engagement gives us a percentage between 0% and 100% where 0% means customer never replied to our emails and 100% means the customer sent us at least as many emails as we sent to them.

Activating AI Auto-Fill

A sys-admin can configure any fields on the Account and Opportunity object to be automatically populated - based on the account's conversation log.

Note: AI Auto-Fill doesn't use the standard activity history to generate the conversation log. Instead it is using Goldie Conversation Capture which is based on custom objects. If you were using a different activity capture solution (Einstein Activity Capture, Groove, etc.) AI Auto-Fill won't be able to infer field values. If you are interested in migrating your data contact us.

 Activating and configuring AI Auto-Fill takes three steps:

Step 1. Enable Change Data Capture

1. In Setup find "Change Data Capture".

2. Find the object "Act (absf__Act__c)" in the list of Available Entities.

3. Click the Add button to add the object to the list of Selected Entities.

4. Click Save.

Note: This step is necessary so Goldie is triggered to query the LLM for new field value suggestions, each time a new email or a meeting transcript is logged in Salesforce (in the custom object called absf__Act__c).

Step 2. Configure which fields you want auto-filled

1. In Setup, go to the tab "Object Manager".

2. Select the object "Account" or "Opportunity" (only fields on these two objects can be auto-filled).

3. Click on "Fields & Relationships".

4. Create or Select the field you want automatically populated with these constraints:

   • Only fields that relate to customer conversations should be selected

   • Only these field types are supported: Text, Text Area, Text Area (Long), Checkbox, Date, Number, Percent

5. Append the field label by copying and pasting the AI symbol: ✨

   • For standard fields, add the AI symbol by renaming the label in Setup > Rename Tabs and Labels 

6. The ✨ symbol in the field label indicates:

   • to Goldie that you want this field to be automatically populated and

   • to your users that the field will be automatically populated. We recommend to make the field read-only.

7. Important: In the field help text, put instructions for the LLM on how the field should be populated. The field help text and the field label will be part of the prompt that is sent to the LLM.

   • Note: If the field help text is blank, the field won't be auto-filled.

8. Save your field settings.

Note: While Goldie might log an email or meeting under multiple associated accounts and opportunities, it will only auto-fill fields on the "best-fit account record" and the opportunity under that account with the nearest recent close date - preferably (but not necessarily) in an open stage.

Step 3. Configure and select an LLM

1. In the App Launcher, find the item "Goldie Setup"

2. In the Goldie setup page select an LLM model:

   1. Use gpt-4o-mini or gemini-2.0-flash if you only want to populate summary fields with no need for "thinking". These models are fast and cheap.

   2. Use gemini-2.5-flash if your fields require counting and calculations or a higher degree of reasoning. This model has fair performance and is quite inexpensive.

   3. Use gemini-2.5-pro if gemini-2.5-flash didn't give you the performance you expected. This model is slower and much more expensive.

3. Depending on your choice of an LLM provider, follow the instructions for adding an API key when using the OpenAI LLM or using the Google Gemini API.

While Goldie auto-fills the configured fields each time a new email or meeting is logged, there might be cases where you want to trigger AI Auto-Fill on demand. You can add an action to the account page's highlights panel by following these instructions:

1. On any account page, open the Setup menu and select "Edit Page". The Lightning App Builder allows you to edit the page layout.

2. Click on the Highlights Panel at the top and find the button "Add Action" in the left panel.

3. Click the button "Add Action" and select the item "Goldie, Auto-Populate Fields!"

4. Arrange the order of your selected actions and click Save to save the page layout.

You can now trigger AI Auto-Fill using an action in the account page.

Note: Depending on the model you selected and the size of the conversation log, AI Auto-Fill can take up to 120 seconds before fields are updated.