Getting Started

Using ai to fix data quality makes this more efficient and easier, but it is nuanced. Read this quick start guide to understand the basics of how adq works.

How it works

When a prompt is submitted on the adq assistant ui in Salesforce, the Salesforce record, object, and prompts are sent to our system.


Our system then identifies the context of your data. This means taking the prompt entered into the "what looks incorrect about this data" box and comparing it to the record the prompt was entered on. If there are references to another object (e.g., opportunities when the prompt was on a contact), then it also pulls in this data.


Using this context, the adq assistant generates a query to find all records with similar problems. The output of this query populates the "similar records" tab in the returned analysis.


Once the "similar records" are identified the system creates a query to show potential changes. These are the values that are loaded into Salesforce after a user reviews them.


All queries are tested as they are put together to ensure that there are no issues. Similarly all AI models are hosted on compliant and secure infrastructure.

Prompting

Prompting is key to use an AI system effectively. The adq assistant is no different. It's not magic, it's artificial intelligence!


To reduce errors and get the best outputs include field names and object names where possible. If you're looking for a specific format or field, say so!


Avoid vague instructions (e.g., "update the opportunity to be formatted account pipe product"). Instead specific naming is important (e.g., "update the opportunity name to be formatted account name pipe product name"). As a rule of thumb, if a person would be confused, then the adq ai data assistant will definitely be confused.


As always, the variability of prompting means that checking outputs is important! Adq is mostly accurate (and will continue getting more accurate), but it's not perfect.

Setting up

The adq Salesforce data cleaning assistant is what's called a lightning web component (lwc). Although the adq team will help setting up the connected app in your Salesforce environment it still needs to be added to the page layouts for any new views that you create.


The adq agent can be added to the page like any other field. Although the Salesforce editor shows it taking up significant space on the page, it will always be located at the bottom of your screen and can be minimized with a click.


Many organizations don't have dedicated Salesforce resources to dedicate to adding adq. If that sounds like you, we can help set you up when setting up the connected app.

Troubleshooting

The best way to deal with an error is to retry with more specific prompts. Errors are often occurring due to a gap in what adq believes it is supposed to do and the possible changes based on data in your Salesforce.


Fortunately, adq believes that your problems are our problems, and any issues our customers have are an opportunity for us to improve our product. As such, the founder is available for complaints at connor@getadq.com

Getting Started

Using ai to fix data quality makes this more efficient and easier, but it is nuanced. Read this quick start guide to understand the basics of how adq works.

How it works

When a prompt is submitted on the adq assistant ui in Salesforce, the Salesforce record, object, and prompts are sent to our system.


Our system then identifies the context of your data. This means taking the prompt entered into the "what looks incorrect about this data" box and comparing it to the record the prompt was entered on. If there are references to another object (e.g., opportunities when the prompt was on a contact), then it also pulls in this data.


Using this context, the adq assistant generates a query to find all records with similar problems. The output of this query populates the "similar records" tab in the returned analysis.


Once the "similar records" are identified the system creates a query to show potential changes. These are the values that are loaded into Salesforce after a user reviews them.


All queries are tested as they are put together to ensure that there are no issues. Similarly all AI models are hosted on compliant and secure infrastructure.

Prompting

Prompting is key to use an AI system effectively. The adq assistant is no different. It's not magic, it's artificial intelligence!


To reduce errors and get the best outputs include field names and object names where possible. If you're looking for a specific format or field, say so!


Avoid vague instructions (e.g., "update the opportunity to be formatted account pipe product"). Instead specific naming is important (e.g., "update the opportunity name to be formatted account name pipe product name"). As a rule of thumb, if a person would be confused, then the adq ai data assistant will definitely be confused.


As always, the variability of prompting means that checking outputs is important! Adq is mostly accurate (and will continue getting more accurate), but it's not perfect.

Setting up

The adq Salesforce data cleaning assistant is what's called a lightning web component (lwc). Although the adq team will help setting up the connected app in your Salesforce environment it still needs to be added to the page layouts for any new views that you create.


The adq agent can be added to the page like any other field. Although the Salesforce editor shows it taking up significant space on the page, it will always be located at the bottom of your screen and can be minimized with a click.


Many organizations don't have dedicated Salesforce resources to dedicate to adding adq. If that sounds like you, we can help set you up when setting up the connected app.

Troubleshooting

The best way to deal with an error is to retry with more specific prompts. Errors are often occurring due to a gap in what adq believes it is supposed to do and the possible changes based on data in your Salesforce.


Fortunately, adq believes that your problems are our problems, and any issues our customers have are an opportunity for us to improve our product. As such, the founder is available for complaints at connor@getadq.com