The GenAI Divide: State of AI in Business 2025 a report released by MIT’s NANDA initiative has shared major insights on how AI is being used in enterprises – particularly on what has worked and what has not. The report states that 95% of generative AI implementation in enterprises has had poor results for a variety of reasons. Nonetheless, it also states that 67% of businesses saw success from purchasing AI tools from specialized vendors and building partnerships, while building internal AI tools only succeed 1/3rd of the time.

Greater success was seen in organizations that partnered smartly with AI startups to address a main pain-point which allows for a deeper integration of the tool. This allows for the enterprise and the tool to “learn” and constantly improve its output.

A Bain & Company brief on AI Transforming Productivity further elaborates this point–without process redesign, businesses risk automating inefficiencies rather than removing them, and the introduction of AI into processes frequently results in micro-productivity because of new bottlenecks.

Both these reports echo Pavestone’s approach to its AI investments—AI tools and platforms need to be integrated at a core level to be successful.

Additionally, Pavestone’s observation has been that external AI tools and platforms are more successful as there is a holistic approach to understanding the entire workflow rather than offering band-aid solutions. These vendors tend to work with cross functional teams rather than siloed AI teams. Thus, these external platforms only offer a larger view of the workflow. Moreover, they tend to be more robust, have better built-in compliance and risk management due to their experience with multiple organizations. Most importantly, external platforms face greater scrutiny before being onboarded with extensive pilots, company diligence, and governance.

The report also mentioned that while most enterprises have invested in AI for sales and marketing, the biggest return on investment has been in back-office automation—eliminating business process outsourcing, cutting external agency costs, and streamlining operations. A Bain & Company reports says that 78% of 500 IT leaders surveyed expected at least some ERP functionality to be replaced or enhanced by agentic AI over the next three years.

We can see these trends also reflected in our portfolio companies. For instance, e42.ai ( https://e42.ai) , a pioneer in the agentic AI space, offers shrink-wrapped domain-specific solutions for CFO organizations while still having a flexible platform for enterprises to expand to other use cases. This allows for the tool to deeply integrate into the enterprise solving major pain-points while allowing for “learning” time for enterprise to expand its use case. Bizom’s (https://bizom.com) Real Intelligence Platform is focussed on redefining global Consumer Packaged Goods (CPG) distribution by blending AI and human intuition. They help brands find the right demand in the market, and place the right products at the right outlets. LivNSense’s (https://livnsense.com) GreenOps™, is a specialized “Energy Balance” AI platform that addresses the “Net Zero” Challenge. GreenOps™ is a Digital Twins Platform that uniquely addresses the energy balance by harnessing the power of Artificial Intelligence (AI) and Digital Twins technology to impact decarbonization across process value chain.

Yubi has (https://www.go-yubi.com/)powered its end-to-end debt Lifecyle platform using AI by taking a holistic approach and reimagining the workflows. The platform facilitates every stage of the flow of money from lenders to borrowers and back to lenders including – credit marketplace, risk evaluation, and collections.

NewSpace Research and Technologies works closely with the Indian Armed Forces to understand their needs and developed the world’s first swarm drones that collect and share intelligence.

CynLr works with the automotive industry to create “object intelligence” for robots, this allows them to visually differentiate between objects and manipulate them without pre-training creating efficiency in factories – this was a major pain-point in the automotive industry solved using visual intelligence.

All of Pavestone’s investments are in line with the success trends seen in the MIT report as well as the future of AI in ERP seen in the Bain & Company survey. We believe we are headed in the right direction with our approach to AI in our investments.