SugarCRM’s Approach to AI

Overview

At Sugar, we put a lot of thought and care into how we leverage AI. It's important to us that our AI tools deliver real value, your data remains protected, and we build an infrastructure that can grow as the technology evolves.

AI in Sugar Today

Currently, Sugar offers AI tools with predictive and generative capabilities.

How Your Data is Protected

Your data is essential to your business, and it should always be protected. To keep it safe, we have multiple layers of security in place.

• Your data is never used to train external models. All external models used by our AI tools are contractually obligated to never use your data for training purposes.
• Your data is anonymized before it is shared with external models. We replace personally identifiable information with analogous but generic data before it is shared.
  

How We Deliver Value

We don’t create AI tools just for the sake of it. We create them to deliver real value, helping you get more out of your data and time.

• No prior experience needed. AI responses are generated automatically with no prompting or prior AI experience required.
• Improved responses with grounding. Grounding aligns our AI tools with your specific data, helping responses reflect your terminology, use cases, and business realities.
• Every response is moderated. Every response from our AI tools goes through a final moderation check to improve accuracy and prevent harmful content.

How We Are Planning for the Future

• A model agnostic system. Our tools are designed to be model-agnostic, ensuring seamless integration with most AI models and enabling continual improvements as the underlying models evolve.
• Collecting and acting on user feedback. We're working closely with our users as we build new tools and improve the ones we already have. In the future, we plan to add easier and more focused ways for users to share feedback.

Our Approach in Action

To help tie everything together, here’s an example of your account data interacting with the Account Intelligence tool.

You’re leading a sales team for a new region and need to quickly get up to speed on the accounts they manage. You open the first account record and start reviewing the Account Intelligence dashlet. Behind the scenes, here’s how we generated the response you’re reading:

	First, we ground the prompt that we send to the external large language model (LLM) with your unique data. This helps the model understand and use your language and recognize what is important to your business.
	Next, we collect all of the text-based data from the account record and specific related modules. To protect your data, we mask it by anonymizing any personally identifiable information before sharing it with the LLM.
	We then share your data with the LLM, which is contractually obligated to never use it for training purposes.
	The LLM reviews your data and sends back a response, which we then moderate to improve accuracy and prevent harmful content.
	Finally, the Account Intelligence dashlet delivers the response to you. We continuously audit response quality to improve the tool over time.