Across AI Raises $5.75 Million to Help Businesses Turn Scattered Data into Actionable Intelligence

Businesses generate enormous amounts of information every day, from customer conversations and sales records to internal documents and market research. Yet, finding the right information at the right moment remains a challenge for many teams. A new artificial intelligence startup, Across AI, is developing a system designed to help businesses connect that information and make better decisions.

The company was launched by Steven Mih, a technology entrepreneur who previously co-founded Ahana, a data infrastructure startup acquired by IBM in 2023. After spending 14 months at IBM following the acquisition, Mih left in July 2024 to pursue a new opportunity: helping companies make their growing volumes of data more useful through AI.

Across AI is building what it calls an “agentic memory” system. Rather than functioning like a conventional chatbot that responds to individual prompts, the platform is designed to build an evolving understanding of a company’s information, workflows and priorities. It aims to help teams identify potential sales opportunities, recognise emerging risks and determine which questions they should be asking customers.

The idea addresses a familiar problem in large organisations. Important knowledge is often scattered across customer relationship management systems, emails, calendars, collaboration platforms and internal documents. Some information exists only in the experience of a few employees, while other details become difficult to find as data accumulates. This fragmentation can slow decisions, create inefficiencies and cause teams to miss opportunities.

Across AI intends to bring these sources together into a shared system that can provide relevant information within the tools employees already use, including platforms such as Slack and Microsoft Teams. By connecting information across these systems, the startup hopes to reduce the time employees spend searching for answers and help them act on insights more quickly.

A key feature of the platform is its ability to update its understanding as information changes. The system is designed to track new developments, identify outdated or conflicting information and prioritise details according to their relevance to a particular task. Where information remains ambiguous, the company says the system can escalate the issue to a relevant human decision-maker rather than making an unsupported determination.

Mih is building the company alongside Dr. Niloufar Salehi of the University of California, Berkeley, and Dr. Afshin Nikzad of Stanford University. Both researchers have worked on improving the reliability of AI systems in high-stakes settings, bringing research experience to a product aimed at enterprise users.

In December 2024, Across AI announced a $5.75 million seed funding round co-led by Cota Capital and Village Global. At the time, the startup was refining its product with selected business partners ahead of a planned commercial launch in 2025.

Data security is another central consideration. Businesses may be reluctant to connect sensitive customer and operational information to AI tools if doing so exposes proprietary data to external systems. Across AI said its platform was designed to operate within a company’s secure environment, maintain access controls and avoid exposing company data to external models for training. It also planned to offer both software-as-a-service and cloud-premises deployment options.

For Mih, the new venture builds on lessons from Ahana, which helped organisations analyse large datasets using the open-source Presto query engine. Across AI approaches a related challenge from a different direction: instead of focusing primarily on accessing data, it seeks to help people understand the information, retain relevant context and use it in their daily work.

The broader opportunity lies in the transition from AI tools that simply answer questions to systems that can support more complex business processes. However, the effectiveness of agentic memory will depend on how accurately it interprets context, manages changing information and protects sensitive data.

Across AI’s proposition is that businesses do not necessarily need more information; they need a better way to connect what they already know with the decisions they must make. If the company can deliver on that promise, it could help sales teams and other enterprise users spend less time searching for answers and more time acting on them.

Source: TechCrunch