An alternative to the neural network architectures powering AI models like OpenAI's is gaining traction—symbolic AI. This approach uses specific rules for tasks, such as rewriting text, to address larger problems effectively.

Symbolic AI excels in areas where neural networks often struggle, and recent research indicates that it can be scalable, addressing a historical limitation of symbolic architectures in terms of computational efficiency. This scalability has sparked a surge of startups leveraging symbolic AI across various sectors, including Orby, TekTonic (which focus on enterprise automation), Symbolica, and Unlikely AI (founded by Alexa co-creator William Tunstall-Pedoe). One of the newest entrants is Augmented Intelligence, which has secured $44 million in backing from investors, including former IBM President Jim Whitehurst.

Augmented Intelligence aims to build conversational AI that is more predictable and “agentic”—a trending buzzword in the industry—compared to typical neural network systems. For instance, rather than simply providing information about booking a flight to Mexico, Augmented Intelligence's AI can present fare options and book the flight directly, according to CEO Ohad Elhelo.

When asked if ChatGPT can already perform such tasks, Elhelo acknowledges that it can, but claims it requires more setup and manual integration than his company's technology. He emphasizes the distinction between chatbots designed primarily for user interaction and conversational agents that can take actions on behalf of companies. “Once you connect the AI to tools—either to retrieve information or to act—the model no longer relies on its training data, which significantly diminishes the quality of intelligence,” he explained.

Elhelo co-founded Augmented Intelligence with Ori Cohen in 2017, initially under the name Delegate. The company aimed to provide an AI-powered app for delegating tasks to low-paid gig workers. However, the venture faced challenges, leading to a pivot toward AI solutions.

The AI developed by Augmented Intelligence can power chatbots that respond to a variety of inquiries, such as pricing policies, by integrating with a company’s existing APIs and workflows. Elhelo claims the AI was trained on conversational data sourced from tens of thousands of human customer service agents.

When questioned about the origins of this data, Elhelo did not provide details, nor did he clarify whether agents were compensated for their contributions or informed about the data usage.

Despite mixed opinions regarding brand chatbots, Elhelo argues that businesses may prefer Augmented Intelligence over other AI vendors because its AI is designed to utilize tools to gather information from external sources for task completion. While other AI platforms can also leverage tools, Elhelo asserts that his company’s solution performs better than those based on neural networks.

Moreover, he claims the AI is more transparent, offering logs that explain how it generated responses, allowing companies to refine its performance. Additionally, Elhelo notes that Augmented Intelligence does not require training on a company’s internal data, using only authorized resources for specific contexts. This aspect is particularly appealing to businesses concerned about potential data leaks, as demonstrated by Apple’s reported ban on OpenAI tools for employees due to confidentiality worries.

Although some of Elhelo’s claims—such as the assertion that Augmented Intelligence can "eliminate hallucinations"—are questionable, the 40-employee company has been securing business, including a recent strategic partnership with Google Cloud to integrate its models into the platform.

While Elhelo did not disclose revenue figures, he shared that Augmented Intelligence's last $10 million funding round, led by New Era Capital Partners, valued the company at $350 million—a noteworthy figure for a startup that has only recently launched its product.

“Traditional models excel at pattern recognition and language generation,” Elhelo said. “However, these architectures fall short in scenarios requiring actions, decisions, or interactions with tools. Apollo’s neuro-symbolic architecture addresses these challenges and offers new opportunities for companies.”