Burst of light illustration
Getty Images/Cobalt88
When generative AI (Gen AI) burst into public consciousness two years ago, chatbots were seen as the cutting edge of technology. However, developments have evolved rapidly in the space. Now, more capable AI assistants and agents have evolved to execute actions on your behalf.
Also: ChatGPT's subscribers and revenue soar in 2025 - here's why
Beyond those AI agents lies Enterprise General Intelligence (EGI), which, as the name implies, refers to more advanced AI solutions tailored to businesses' domain-specific needs. Although much less is known about EGI than artificial general intelligence (AGI), a computer system that can solve problems as well as, or better than, a human being, some experts believe EGI could transform business operations.
To learn more about EGI, ZDNET spoke to Silvio Savarese, head of Salesforce AI Research, who has just released a blog post explaining the findings from his pioneering research into the concept. Keep reading to learn more about EGI, how it compares to AGI, and how far away it is.
What is Enterprise General Intelligence (EGI)?
An EGI is a highly capable AI system that handles business applications reliably. However, to understand the concept, it is important to understand what differentiates EGI from other business AI solutions that already offer big productivity gains, such as AI agents.
EGI's standout features are performance across the dimensions of capability and consistency. EGI systems' high capability means they can navigate the complex needs of business environments with predictable results on the Capability-Consistency matrix.
On the capability front, Savarese said EGI systems will have higher levels of reasoning that enable them to perform complex and operational tasks, such as deep research, in rapidly changing environments while implementing real-world human feedback.
"Agents we are deploying are a bit of a beginning of this trajectory; they can do simple things, but definitely not this kind of deep research, they cannot do long horizon tasks, they cannot do complex reasoning yet," said Saverse.
Also: ChatGPT's subscribers and revenue soar in 2025 - here's why
The consistency axis refers to EGI systems that can deliver predictable, reliable, and accurate results. Specifically, Saverse said that consistency helps these systems avoid "jaggedness", where AI models excel at complex tasks but get more simple ones wrong. This consistency makes EGI systems fit for enterprise use cases because an inconsistent system would be useless, even if it excels at peak performance.
"You don't want to go from stellar to crap; you want something that is very, very, reliable, trusted," said Saverse. "If customers use this tool, they need to know that this is pretty much guaranteed to work well."
How does EGI compare to AGI?
Because AGI refers to AI with human-like intelligence and autonomy, such as a system in a sci-fi movie where AI takes over the world, there is usually some hesitation surrounding the tech. As a result, you may be relieved to hear that, despite having similar names, EGI and AGI differ in function and rollout.
To better understand this relationship, it's helpful to consider the broader AI landscape. Savarese divides AI development into five waves: predictive, copilots, AI agents, robotics, and, lastly, AGI. Even though EGI is not officially listed as a wave, it lives between agents and robotics because, as described above, it takes AI one step further from agents via deeper reasoning.
Waves of AI
Salesforce
Also: Adobe brings four highly-requested Premiere Pro AI features out of beta
Even though AI agents are available now, we are not yet at the point where EGI is feasible. Before that can happen, new benchmarks must be developed that look specifically at the EGI systems' performance on AI tasks, plus stress-testing environments to push these tools to their limits.
"If you look at how these LLMs are evaluated, they're evaluated on the AGI benchmark, not the EGI benchmark. They are evaluated on those tasks that are all over the place, but they're not really focusing on the enterprise tasks," said Savarese.
To help combat this issue, Salesforce has been working on a CRM benchmark that measures proficiency in performing tasks, such as how AI can summarize sales emails and transcripts, make commerce recommendations, and more. Although this benchmark is not the perfect solution, it is a step in the right direction. Savarese suggested EGI systems could emerge soon in six to 12 months.
Also: Microsoft's new AI agents aim to help security pros combat the latest threats
AGI is a concept that could be realized further in the distance, so far that researchers can't quantify when that point will be reached. However, to contextualize the distance, if you look at the chart above, AGI is wave five, which follows the deployment of robots. Although there has been a lot of progress in robotics, the hardware is still not at a point where it is commercialized easily. Savarese said that robot development is pivotal to the emergence of AGI.
"AI understands how the world works through just literature, through books, and it's not the same, right? It's not the same as experiencing the world. So, by the time AI will start experiencing the world, which is through robots, that's when we open the door to AGI," he said.
Visit his blog post to read more about Savarese's findings, including steps businesses can take to prepare.
Artificial Intelligence