Navigating the AI landscape can be daunting, given the sheer number of choices available. But despite all this choice, no single platform can meet everyone’s needs. That’s why we’re committed to an open model ecosystem that lets you tailor your artificial intelligence (AI) solutions precisely to your business requirements.
Building on the success of our bring-your-own-model (BYOM) approach, which gives customers greater flexibility when choosing a predictive AI model, we’re offering something similar for generative-AI large language models (LLMs).
Bring your own LLM — or BYOLLM — means you can use an out-of-the-box LLM that we provide, or use an LLM from the third-party system of your choice. Being able to use tools and technologies from various providers encourages innovation. The result: a customized platform that does exactly what you want and meets your business needs. Each component complements the others, leading to seamless collaboration and better results.
Introducing the LLM Benchmark for CRM
Choose the best use case and the right LLM for your business with a benchmark that lets you compare how LLMs fare in terms of accuracy, cost, speed, and trust.
Flexibility and customization are great, but an AI platform strategy isn’t complete without a focus on trust and privacy. Our strategy is integrated with the Einstein Trust layer, a secure AI architecture that makes sure AI-generated outputs are reliable and safe, and ensures data privacy and integrity through advanced security measures.
Let’s look more closely at how our open platform strategy for AI offers flexibility, enhanced capabilities, and trust.
Flexible out-of-the-box LLMs
For many customers, Salesforce-managed models are an excellent starting point, meeting virtually all AI use cases. You can choose from a range of preconfigured LLMs within Prompt Builder, including the latest GPT-4 models from OpenAI, Azure OpenAI models, and Claude models from Anthropic. These models also provide an option for global data residency, ensuring compliance with regional data regulations in the EU, Japan, India, Australia, and Brazil. Other foundational models will be available in the future, and we’ll soon introduce domain-specific models for select scenarios. We also integrate AI models directly into Apex and Flow.
Integrated third-party LLMs
BYOLLM makes it easier for you to integrate AI into your operations. You can connect externally hosted LLMs to Salesforce via Model Builder. Model Builder lets you choose from OpenAI, Azure OpenAI, Google, and Amazon models. Integration is seamless, with minor setup and configuration.
A fifth option, the LLM Open Connector, lets you connect any model of your choice, hosted on any provider or platform, as long as it adheres to our Open Connector specification.
Once you’ve connected an LLM, you can use it within Agentforce custom actions via Prompt Builder. Agentforce lets you rapidly build, tailor, and deploy embedded AI apps and conversational agents using low-code tools.
How agents can take smarter actions with Prompt Builder
Prompt Builder — the intelligence layer within Agentforce — helps you tailor agents to your business’ needs.
Trust and security
Trust and security are top of mind for business leaders looking to incorporate AI into their systems. We’ve addressed this by providing a secure AI architecture called the Einstein Trust Layer. Think of it as a set of features and guardrails that ensures AI-generated outputs are reliable and safe, giving you the confidence to use AI in your decision-making processes. It also ensures data privacy and integrity through advanced security measures like encryption and access control, and it’s designed to safeguard sensitive data including personally identifiable information (PII).
We also have zero data retention and data residency with our model partners, including OpenAI, Azure OpenAI, AWS, and Google, to make sure your information is secure.
Data masking for third-party models adds another layer of protection, letting you collaborate with external partners while keeping your critical information safe. Sensitive prompt data is redacted, using pattern masking and field masking, before it goes to partner LLMs outside the trust layer.
Achieve your goals with our open platform strategy for AI
Whether you’re using AI to enhance customer interactions, automate workflows, or gain deeper insights from data, our open platform strategy for AI helps you achieve your goals. It offers you flexible options and enhanced capabilities so you can take advantage of the latest advancements in AI technology safely and securely.
