When we first heard about generative AI a couple of years ago, chatbots seemed like the pinnacle of tech innovation. But things have moved fast since then. Now, we have AI assistants and agents that can actually perform tasks for us. But there's something even more advanced on the horizon: Enterprise General Intelligence, or EGI.

What Exactly is Enterprise General Intelligence?

EGI is a sophisticated AI system designed specifically for business purposes. Unlike the AI tools we currently use, which boost productivity, EGI stands out for its capabilities and consistency. These systems are built to handle the intricate demands of business environments, consistently delivering reliable results.

Silvio Savarese, who heads Salesforce AI Research, explains that EGI systems are capable of high-level reasoning. This means they can tackle complex tasks and adapt to changing environments while incorporating real-world feedback. Current AI agents are just scratching the surface, handling simpler tasks but not yet ready for deep research or complex reasoning.

The consistency of EGI is another key feature. These systems are designed to avoid what Savarese calls "jaggedness," where an AI might excel at complex tasks but stumble over simpler ones. For businesses, having a reliable system is crucial—nobody wants a tool that swings wildly between brilliance and failure.

Comparing EGI to AGI

You might have heard of Artificial General Intelligence (AGI), often depicted in sci-fi as AI with human-like intelligence. While EGI and AGI sound similar, they serve different purposes. AGI is about creating AI with human-level thinking and autonomy, which can be both exciting and a bit scary.

Savarese outlines five waves of AI development: predictive models, copilots, AI agents, robotics, and finally AGI. EGI sits somewhere between AI agents and robotics, offering deeper reasoning capabilities than current AI agents but not yet reaching the level of robotics.

The Road to EGI

While AI agents are already here, EGI isn't quite ready for prime time. Before we get there, we need new benchmarks to assess EGI's performance on specific business tasks. Salesforce is working on a CRM benchmark to measure how AI can handle tasks like summarizing sales emails and making recommendations. Although it's not a perfect solution, it's a step forward.

Savarese suggests that we might see EGI systems in action within the next six to 12 months. In contrast, AGI is still a distant goal, largely because it requires advancements in robotics. According to Savarese, AI needs to experience the world through robots to truly reach AGI.

For more insights from Savarese on this topic and how businesses can prepare, you might want to check out his recent blog post.