Follow Integration Steps and Best Practices
Learning Objectives
After completing this unit, you’ll be able to:
- Describe how the integration of Einstein and Amazon Bedrock can improve customer support efficiency and consistency.
- Identify the prerequisites for integrating Einstein with Amazon Bedrock.
- Describe best practices for successfully integrating a custom model with Einstein.
In this unit, you explore Get Cloudy Consulting’s use case, its key challenges in customer support and data security, and how it can use Amazon Bedrock and Einstein to solve them. First, you identify specific problems. This provides the context for why the company is pursuing this integration. Next, you define the high-level steps required to successfully integrate Einstein with Amazon Bedrock. After that, AWS offers best practices for integration.
Overview
Get Cloudy Consulting is a consulting and cloud software company that provides enterprise software solutions across different time zones and languages. It uses Salesforce as its primary customer relationship management (CRM) and support platform, managing thousands of customer interactions daily. However, Get Cloudy experiences several challenges in support and data security.
Technical inquiries often require deep product knowledge, forcing support agents to spend significant time searching through documentation that can quickly become outdated. This leads to inconsistent response quality across agents and regions, especially during peak hours or with complex queries, resulting in delays. Additionally, the company must maintain customer data privacy across global operations, meeting different regional compliance requirements. Protecting sensitive technical information, including customer-specific pricing and contract details, is essential to ensure data security.
The integration between Einstein and Amazon Bedrock addresses these challenges by combining Salesforce customer data management with Bedrock’s advanced AI capabilities while maintaining security through the Einstein Trust Layer.
Here's how the company can benefit from this integration.
- Enhance customer support
- Generate responses based on the company’s CRM system data.
- Receive AI-assisted suggestions based on similar past cases.
- Automatically categorize tickets to reduce wait times.
- Generate responses based on the company’s CRM system data.
- Maintain security and compliance
- Keep customer data within Salesforce, with zero data retention in AI models.
- Benefit from the Einstein Trust Layer for data privacy regulations and toxicity detection.
- Audit trails for AI interactions provide transparency and accountability.
- Keep customer data within Salesforce, with zero data retention in AI models.
High-Level Integration Steps
Prerequisites
Before integrating Einstein with Amazon Bedrock, Get Cloudy Consulting must ensure the following prerequisites are met.
AWS Prerequisites
- Active AWS account: Have an active AWS account with the necessary permissions to access and manage Amazon Bedrock.
- Foundation models access: Set up the model in Amazon Bedrock.
Salesforce prerequisites
- Salesforce access: Have access to Salesforce Data Cloud.
- Einstein add-on SKU: In Einstein, enable the Einstein for Sales, Service, or Platform add-on SKU for the BYO LLM feature.
For more information, see Requirements for using BYO LLM.
Integration Steps
After meeting the prerequisites, Get Cloudy Consulting can get started with the integration steps.
- Initial setup: In the AWS Identity and Access Management (IAM) console, grant the IAM user invoke permissions to access the model in Amazon Bedrock.
- Model registration: In Salesforce, register a base or custom Bedrock model in Einstein Model Builder. Provide required credentials—AWS access key and secret key—and configure model parameters.
- Prompt configuration: Create custom prompts using Salesforce Prompt Builder to ensure consistent responses.
For step-by-step instructions, see:
Follow Best Practices When You Bring Your Own Model (BYOM)
When bringing your own model to Einstein, there are several considerations to keep in mind.
- Model compatibility: You can bring in your own model by using Einstein Studio and write a prompt template in Prompt Builder, which you can then integrate into your own apps or workflows.
- Performance validation: You must thoroughly test and benchmark your custom model's performance. Make sure it meets the required accuracy and response time thresholds for your specific Salesforce use case.
- Ethical AI principles: As you build out your generative AI models, you should establish clear guidelines to ensure they are operating effectively, accurately, and without introducing bias or inappropriate content.
- Secure data practices: AI models learn from the data you provide, rather than storing it. You need to implement governance and security measures to protect sensitive information at rest and in transit.
- Transparent Documentation: Document everything—from your model architecture to the training data and the performance characteristics. Maintain clear processes for updating, retraining, and monitoring the model over time.
Follow Best Practices for Successful Integration
For seamless integration of your custom model with Einstein, consider the following best practices.
- Monitoring and feedback loops: Implement robust monitoring strategies to track your model's ongoing performance within Salesforce.
- Collaboration and knowledge sharing: Engage with the broader Einstein and Amazon Bedrock developer communities to learn from their experiences and share your own learnings.
- Continuous improvement: Treat your Einstein integration as an iterative process. Regularly review and optimize your model, data, and deployment processes to drive ongoing improvements.
Where to Go If You Get Stuck
If you face any challenges or have additional questions about integrating your custom model with Einstein, consider the following resources.
- Einstein documentation: The Einstein guide provides comprehensive documentation on model deployment, API usage, and more. For more information, see the Einstein Platform Services Developer Guide.
- Amazon Bedrock documentation: The Amazon Bedrock guide offers guidance on exporting models and integrating with other AWS services. For more information, see Amazon Bedrock Documentation.
- Salesforce developer forum: The Salesforce developer community is an excellent resource for asking questions, sharing experiences, and collaborating with other Salesforce practitioners. For more information, see Trailblazer Community.
- AWS re:Post: A community of AWS experts, including AWS Partners, customers, and employees, where you can ask questions about anything related to designing, building, deploying, and operating workloads on AWS. For more information, see AWS re:Post.
- Salesforce and AWS support channels: If you identified a high-priority issue that requires urgent technical assistance, you can consider reaching out to the support teams at Salesforce and AWS for personalized guidance. For information on how AWS Support and Salesforce can help you, see AWS Support and how Agentforce can help at Salesforce.
Wrap Up
By integrating Einstein with Amazon Bedrock, Get Cloudy Consulting can address their challenges in support and data security. The integration combines Salesforce's customer data management with Bedrock's advanced AI capabilities, enabling consistent and efficient customer support, while the Einstein Trust Layer ensures data privacy and security compliance.