AI can evaluate help text

How helpful is your help text?

Bad data is junk food for AI.  Data quality relies on end-users entering the correct data in the correct fields. Confused users, or users confronted with a page of fields, are a recipe for disaster. A recent Change Intelligence Research report shows the alarming number of fields that users are confronted with. The opportunity object on average has 150 fields. The UX experts recommend just 7 fields to reduce user cognitive overload.

And whilst reducing your page layouts to 7 fields may be unachievable, you can improve data quality with dynamic forms, help text, and validation rules. Evaluating all the help text sounds like a huge analysis job. So why not get AI to help us? If it can’t write the help text – because it doesn’t know why the field was created – could it at least help us estimate the work? 

It can.

Prioritize your workload

Ideally, you should work out which fields are not used and can be hidden from page layouts, so you clean them up first. That will reduce your workload. There is no point in adding help for fields that are never used. We’ve created a prompt for that too. It works out the value of each field based on the dependencies and the % field population. 

Prompt

We’ve built a prompt that can evaluate how good your help text is and help you prioritize the remediation work. It looks at the most commonly used fields and also their importance, which is based on the type of field and the dependencies. For example, a field that kicks off automation is more important to get right than a standalone information field.

There are 2 prompts based on the scope.  Each prompt looks at fields and, if there is help text, evaluates it.

In the prompt below is the scope of the evaluation. The scope is XXX. For example all fields in opportunity object, or the second prompt, all fields in opportunity page layout. You use the prompt inside Elements.cloud in the metadata dictionary org copilot.

If you don’t have Elements.cloud, you are going to have to work out how to export all metadata help text into an XLS and then use this prompt in ChatGPT. You will have to modify this prompt to use in ChatGPT, because you won’t have the % population or dependency information which is the power of the Elements.cloud metadata dictionary.

Object prompt

XXX is object API name.

The scope of the evaluation is each field for XXX object in the Metadata CSV.

You are an AI trained to evaluate help text. Your task is to determine whether the metadata description explains “what the field is” or “how to use the field”.

What the field is: this refers to the content or data type of the field. For example, it describes the kind of information the field holds or its properties.

How the field should be used or completed: This refers to the purpose or reason behind the creation of the field. It explains the motivation or need that led to the field’s inclusion in the application. It often has the following words, and how they appear in the sentence:

1. Why: The reason why this field is important.
2. Shows: Field shows the Date the Licences have been issued to the customer.
3. Used to: This is used on a renewal opportunity to show if a discount had been applied.
4. For: This feature is for tracking user activity.
5. Allows: This option allows users to customize their settings.
6. Enables: This function enables the processing of large datasets.
7. Helps: This software helps in managing project timelines.
8. Designed to: This application is designed to streamline workflow.
9. Serves to: This feature serves to enhance security.
10. Aims to: This module aims to improve user experience.
11. Facilitates: This platform facilitates communication between team members.
12. Provides: This tool provides a way to automate tasks.
13. Intended for: This program is intended for data analysis.
14. Function: The function of this feature is to organize files.
15. Purpose: The purpose of this tool is to simplify scheduling.
16. Utilized for: This system is utilized for real-time monitoring.
17. Allows for: This mechanism allows for quick adjustments.

Here are the steps to follow:

For each field, read the help text.

Determine if the field help text focuses on the nature and content of the field or on the purpose and rationale for its creation. Clearly state whether the help text explains “what the field is” or “why the field was created”. The status of the field is one of the 3 following:  Why, What, No help text.

Calculate if field is standard or custom. Custom field API names end with ‘__c‘.

Calculate the dependency count for the field, which is the number of times there is an occurrence of a dependency for the field in the Dependencies CSV file, using API name in the Metadata CSV file and Source API name in the Dependencies CSV file to match, but only when the dependent type in the Dependencies CSV file is:

• Flow
• Process Builder Workflow
• Apex
• Validation Rule.

Assign each field a weight based on the field type in the Metadata CSV file and the weight in the provided list. 

Calculate the weighted score by multiplying the dependency score by the weight based on subtype in the Metadata CSV file and the weight in the provided list:

Field type, Weight

Checkbox: 1

Currency: 2

Date: 2

Date/Time: 2

Email: 2

Formula: 3

Geolocation: 2

Lookup: 2

Number: 1 

Percent: 1

Phone: 1

Picklist: 3

Picklist (Multi-Select): 3

Text: 1

Time: 1

URL: 2

Provide the output in a table with a row for every field. The columns are: Name, API Name, Status, Weight, Weighted Score, Field Type, Sub Type, Number of Dependencies, % Population, Custom / Standard, Record count, Created Date, Help text, Description.

Please follow this structure to calculate and order the fields, and provide the output for all the fields in a CSV.

Give the distribution of statuses for the fields, by field type split by standard and custom. The distribution % should be to 2 decimal places.

Page layout prompt

XXX is page layout API name.

The scope is a list of fields. Find all the fields in Dependencies CSV file where field name is the Dependent API name, the Dependent type=Field, the Source type= Page Layout, and the Source API name is XXX.

You are an AI trained to evaluate help text. Your task is to determine whether the metadata description explains “what the field is” or “how to use the field”.

What the field is: this refers to the content or data type of the field. For example, it describes the kind of information the field holds or its properties.

How the field should be used or completed: This refers to the purpose or reason behind the creation of the field. It explains the motivation or need that led to the field’s inclusion in the application. It often has the following words, and how they appear in the sentence:

1. Why: The reason why this field is important.
2. Shows: Field shows the Date the Licences have been issued to the customer.
3. Used to: This is used on a renewal opportunity to show if a discount had been applied.
4. For: This feature is for tracking user activity.
5. Allows: This option allows users to customize their settings.
6. Enables: This function enables the processing of large datasets.
7. Helps: This software helps in managing project timelines.
8. Designed to: This application is designed to streamline workflow.
9. Serves to: This feature serves to enhance security.
10. Aims to: This module aims to improve user experience.
11. Facilitates: This platform facilitates communication between team members.
12. Provides: This tool provides a way to automate tasks.
13. Intended for: This program is intended for data analysis.
14. Function: The function of this feature is to organize files.
15. Purpose: The purpose of this tool is to simplify scheduling.
16. Utilized for: This system is utilized for real-time monitoring.
17. Allows for: This mechanism allows for quick adjustments.

Here are the steps to follow:

For each field read the help text.

Determine if the field help text focuses on the nature and content of the field or on the purpose and rationale for its creation. Clearly state whether the help text explains “what the field is” or “why the field was created”. The status of the field is one of the 3 following: Why, What, No help text.

Calculate if field is standard or custom. Custom field API names end with ‘__c‘.

Calculate the dependency count for the field, which is the number of times there is an occurrence of a dependency for the field in the Dependencies CSV file, using API name in the Metadata CSV file and Source API name in the Dependencies CSV file to match, but only when the dependent type in the Dependencies CSV file is:

• Flow
• Process Builder Workflow
• Apex
• Validation Rule.

Assign each field a weight based on the field type in the Metadata CSV file and the weight in the provided list. 

Calculate the weighted score by multiplying the dependency score by the weight based on subtype in the Metadata CSV file and the weight in the provided list.

Field type, Weight

Checkbox: 1

Currency: 2

Date: 2

Date/Time: 2

Email: 2

Formula: 3

Geolocation: 2

Lookup: 2

Number: 1 

Percent: 1

Phone: 1

Picklist: 3

Picklist (Multi-Select): 3

Text: 1

Time: 1

URL: 2

Provide the output in a table with a row for every field. The columns are: Name, API Name, Status, Weight, Weighted Score, Field Type, Sub Type, Number of Dependencies, % Population, Custom / Standard, Record count, Created Date, Help text, Description.

Please follow this structure to calculate and order the fields, and provide the output for all the fields in a CSV.

Give the distribution of statuses for the fields, by field type split by standard and custom. The distribution % should be to 2 decimal places.

Enhanced help text

The Salesfore field help text field is plain text and it is a max of 510 characters, and this may not be enough – especially for picklist dropdowns. Elements.cloud has solved that. We take over the field help icon and turn it into a dropdown so you can add help in a rich text file, URL link to any webpage or external help system, or you can link to a process diagram. The original help text is converted to hint text for the help icon. From the image below you see that we also add a help icon at object level. The help text dynamically changes based on record type.  

BTW Also, you can embed process diagrams in page layouts.

Here is a detailed support article on setting up in-app help.

Final word

Clean data is even more important now that AI is reading it. Cleaning up page layouts by removing unused fields and using dynamic forms is a critical activity that reduces user confusion. Then you can ensure that the remaining fields have easily understood and relevant help text so that end users enter the correct data. AI can help prioritize the work to improve help text.

As we highlighted in our recent Change Intelligence Research Series, simplifying Salesforce interfaces can increase user adoption by over 40%. With ElementsGPT, evaluate the quality and importance of help text. Connect with our team today, to discover how our GPT add-on, can help improve your data quality.