What happened
Zeroset, an AI startup founded by 18-year-old Akshat Kannan and William Zhang, has emerged from stealth with $5.2 million in pre-seed funding. The company is developing a product called Nebula, which tracks workplace software usage to help AI agents understand specific company workflows and task sequences rather than just retrieving documents. The funding round was co-led by Gradient Ventures and 2048 Ventures, with participation from Leblon Capital. Zeroset plans to use the capital to hire researchers and engineers, support early business deployments, and cover the significant costs of training its AI models, with a single training run reportedly costing approximately $100,000. The startup currently has five employees and its product is in a closed research preview, with access expected to expand gradually. Zeroset aims to compete with companies like Mem0 and Zep by building models that understand organizational differences in processes and approval systems, enabling AI agents to complete multi-step tasks with less human involvement.
Zeroset, an AI startup founded by Akshat Kannan and William Zhang, has emerged from stealth with $5.2 million in pre-seed funding. Kannan, 18, left Stanford University in 2026 to lead the company, while Zhang left the University of Texas in 2025. The funding round was co-led by Gradient Ventures and 2048 Ventures, with participation from Leblon Capital.
The company’s first product, Nebula, is designed to help AI agents understand company workflows. It collects information from workplace software and tracks how tasks, decisions, and activities change over time. By analyzing the sequence of events, Nebula aims to provide AI agents with context on how similar tasks were handled previously, rather than relying solely on document search.
Zeroset plans to use the funding to hire researchers and engineers, support early deployments with businesses, and train its AI models. According to Business Insider, one training run cost approximately $100,000 over two weeks. The startup currently has five employees and its product is in a closed research preview, with access expected to expand gradually rather than through a single public launch.
The company aims to help AI agents complete multi-step tasks with less human involvement by understanding organizational differences in processes and approval systems. Zeroset plans to combine license fees with usage-based charges linked to data processed and activity generated by AI agents. It enters a competitive field that includes companies such as Mem0 and Zep, which are also developing technology to help AI systems retain and retrieve information.
Source details: m.economictimes.com ↗
Why it matters
This funding highlights a growing trend in the AI industry toward developing infrastructure that allows agents to navigate complex, organization-specific business processes. While large language models are powerful, they often lack context on how specific companies operate, leading to errors in multi-step automation. Zeroset’s approach of tracking workflow sequences addresses this gap, potentially making AI agents more reliable for enterprise automation. The significant cost of model training, as noted by the founders, underscores the financial barriers and technical challenges involved in building specialized AI systems. This development is relevant for businesses looking to automate complex workflows and for investors tracking the evolution of capabilities beyond simple chat interfaces.
The development of tools like Nebula addresses a critical limitation in current deployments: the lack of context regarding specific organizational workflows. By tracking how work progresses within a company, Zeroset aims to make AI agents more reliable for complex, multi-step automation tasks.
The significant cost of model training, with a single run costing around $100,000, highlights the financial intensity of building specialized AI systems. This cost structure may influence the pricing and accessibility of such tools for smaller businesses.
Zeroset’s entry into the market adds to the growing ecosystem of AI infrastructure companies focused on memory and . Its approach of tracking workflow sequences differentiates it from competitors that may focus more on static document retrieval, potentially offering a more dynamic understanding of business operations.
The startup’s gradual expansion of its closed research preview suggests a cautious approach to market entry, allowing the company to refine its product based on early feedback. This strategy may help ensure that the tool is robust and effective before broader public release.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?
What to watch next
Monitor the expansion of Zeroset’s closed research preview to see when and how the product becomes available to a broader audience. Watch for details on the company’s pricing model, which combines license fees with usage-based charges, as this will impact adoption costs for early customers. Observe how Zeroset differentiates itself from established competitors like Mem0 and Zep in the AI memory and space. Keep an eye on the hiring plans for researchers and engineers, as the team’s growth will determine the pace of product development and model improvement. Finally, track any public case studies or results from early business deployments to assess the practical effectiveness of Nebula in real-world scenarios.
The timing and scope of the expansion of Nebula’s closed research preview will be a key indicator of the product’s readiness for broader adoption. Details on how access will be granted and to whom will provide insight into the company’s go-to-market strategy.
The implementation of Zeroset’s pricing model, which combines license fees with usage-based charges, will be important for potential customers to evaluate the total cost of ownership. Transparency on these costs will be crucial for adoption.
Competitive dynamics with established players like Mem0 and Zep will be significant. Zeroset’s ability to differentiate its workflow-tracking approach from existing memory and solutions will determine its market position.
The hiring of researchers and engineers will impact the pace of product development and model improvement. The quality and expertise of the new team members will be a factor in the startup’s long-term success.
Public case studies or results from early business deployments will provide concrete evidence of Nebula’s effectiveness in real-world scenarios. These results will be important for validating the company’s claims and attracting further investment and customers.