In the ever-evolving landscape of artificial intelligence, the concept of an “AI agent” remains a subject of much debate. While not everyone is in unanimous agreement on its exact definition, one thing is clear - these agents are causing quite a stir. At a broad level, these so-called “agents” hold the promise of going far beyond the capabilities of a simple chatbot, enabling them to make decisions and take actions on behalf of people. Some may assist with online shopping, while others could enhance the efficiency of factories with varying degrees of autonomy.
Unleashing the Power of Dynamic Memory in Enterprise Workflows
Across AI's Mission in the Enterprise World
Across AI is emerging as a key player in this arena, coming out of stealth to develop a “dynamic memory system” specifically tailored for complex enterprise workflows. The startup is targeting chief revenue officers and sales teams, offering a platform that seamlessly connects with all their internal and external enterprise data sources. This creates a shared “agentic memory” that proves invaluable in identifying and qualifying new sales opportunities, spotting risks, and suggesting relevant questions for sales teams to ask their customers.As Steven Mih, the co-founder and CEO of Across AI, explains, “Sales teams often face challenges in obtaining and utilizing the right information at the right time. Critical knowledge gets bottlenecked among a few experts or gets lost in vast amounts of unstructured data, leading to inefficiencies and costly errors. Existing AI solutions often fall short as they lack deep integration and contextual understanding.”
Mih's previous venture, Ahana, was a Google Ventures-backed company that built commercial services on Presto. After selling Ahana to IBM last year, Mih decided to embark on a new journey with Across AI. He joined forces with renowned professors Dr. Niloufar Salehi and Dr. Afshin Nikzad from UC Berkeley and Stanford University respectively, who have been researching ways to improve the efficacy of AI systems in “high-stakes” settings.
Building the Memory: Connecting Enterprise Stack Components
Across AI is designed as a web app and chatbot that integrates with various parts of the enterprise stack, including CRM systems, communication and collaboration tools, and calendars. By doing so, it builds a comprehensive memory and develops contextual understanding. This allows it to provide just-in-time assistance wherever a user is working, without disrupting their workflow.“As the system shows up where users already are, such as in Slack or [Microsoft] Team’s app, it offers seamless support and enhances productivity,” Mih said. The memory continuously adapts, retaining only relevant information and discarding outdated data. This raises interesting questions about how it determines relevance, which Mih attributes to the development of a “deep understanding of the workflow context.”
The system actively tracks, timestamps, and monitors information updates, ensuring that data remains up-to-date and conflicts are resolved. Unlike traditional AI systems that treat all data equally, Across AI's agentic memory system prioritizes information based on its contextual importance. In cases of ambiguity, determinations are escalated to relevant personnel like sales managers or product managers.
Data Security: A Foundational Aspect
Enterprises have been cautious in adopting generative AI due to concerns about data privacy and security. Mih understands this concern and emphasizes that data security is a “foundational aspect” of Across AI's agentic memory platform.“Our memory system operates within the company's secure environment, maintaining strict access control over sensitive information and not exposing data to external models for training,” Mih said. The company plans to offer both SaaS and cloud-premises deployment options to meet the diverse security and compliance requirements of enterprises.
There are subtle synergies between Across AI and Mih's previous startup, Ahana. Ahana focused on enabling users to query vast amounts of data via Presto, handling the complexities of infrastructure setup and maintenance. Across AI is addressing the same problem from a different perspective, aiming to help users analyze large amounts of data quickly.“This experience has deepened my understanding of the challenges enterprises face in navigating complex data ecosystems,” Mih said.
