95% of Enterprise AI applications are falling short. What are the other 5% getting right?
A recent report from MIT indicates that 95% of enterprise-wide AI applications fail. Study researchers are calling it the “GenAI Divide”, a gap between the consumer success of AI and enterprise outcomes.
At Canoe, we’ve been implementing AI and machine learning in production environments since before ChatGPT was a whisper on the internet. We’ve learned that AI does work. Unfortunately, it’s not working successfully for most business applications yet, and there is a simple reason why.
Based on our experience and countless iterations of our AI model, we’ve learned that successful deployments can’t be treated as just another API to integrate. Whether you’re building your own AI or partnering with service providers, there is one fundamental at the core of AI success. Miss this, and your AI project is destined to fail.
Why AI fails.
Research conducted by EY US reveals that 83% of IT leaders attribute stalled AI adoption to weak infrastructure,¹ but the MIT study demonstrates that infrastructure alone can’t explain why most pilots fail. When MIT surveyed frontline users at over 50 organizations, poor output quality and an inability to align with existing workflows were the biggest barriers to adoption.² In most cases, applications were unable to retain complex information, and 60% were unable to adapt to specific business contexts, making them incompatible with existing workflows.³
To understand why AI stumbles in certain environments, we have to look at the application layer—the platform on which many tools are built. According to Aditya Challapally, the lead researcher on the MIT study, generic AI platforms like ChatGPT stall in enterprise use, because they don’t learn from or adapt to workflows—⁴a predictable outcome when models are trained to complete simple repeatable tasks, such as organizing structured data or generating text.
Australia’s National Science Agency recently conducted a 6-month trial of M365 Copilot, seeking to “enhance productivity, develop AI skills, and explore safe, ethical use cases for generative AI.”⁵ They found that the platform excelled at structured tasks, such as drafting emails and summarizing meetings, but failed when applied to scientific research, where information from multiple sources must be synthesized to generate hypotheses and support discovery.⁶
The same limitations apply in alternative investments, where unstructured data and complex reporting workflows mirror the challenges seen in scientific research. AI-powered data tools built on generic platforms are not trained on the complexity of unstructured alts data or designed to work with reporting workflows that vary from investor class, jurisdiction, and fund structure. As a result, investors face an increased risk of errors and require process workarounds that reduce efficiency.
To win with AI in alternatives, investors need a platform that understands the context to properly interpret alts data and the workflows that drive the critical processes behind valuation, reporting, and compliance.
Why AI is alive and well at Canoe.
The MIT study highlighted why AI projects fail in general, but it could have easily been speaking directly to the alternatives market. Issues like poor output quality and an inability to adapt to specific workflows translate into wasted hours manually reformatting documents, while errors undermine client confidence or return compliance fines. That’s why we take a different approach at Canoe, building our own models that are reared to support alts data and the people who use it.
Unlike a lot of the new AI companies that have popped up in the last three years, Canoe AI is not just a wrapper for ChatGPT. We’ve invested in our own proprietary AI architecture over the last twelve years, tuning it specifically and only for alts investors. It’s our understanding of the problems at hand, and the unique processing layer that makes AI work competitively and successfully in alts. Tools making use of generic models may stumble over the nuances of capital calls, K1s, and quarterly statements, whereas Canoe AI has been trained on the actual documents that drive alternatives, drawing from a master database of over 44,000 funds.
Able to extract line-item details from capital calls with 97%+ accuracy before expert human validation is applied, Canoe AI automatically maps to data fields across a wide variety of documents. Institutional investors can depend on extracted data to support valuations, as well as investor and compliance reporting, easily integrating with existing workflows to enhance productivity.
Trained in real-world alternative environments and designed specifically to facilitate the way investors work in alternative markets, Canoe is trusted to reduce manual effort while ensuring accuracy. With efficiency at its core, Canoe is proving daily that AI in alternatives is alive, well, and ready to adapt to whatever happens next in the alternatives landscape.
¹ Refna Tharayil. “Data Infrastructure Challenges Slowing AI Adoption, say 83% of IT Leaders.” Tech Monitor, Dec. 11, 2024. Web.
² Aditya Challapally, et al. “The GenAI Divide: State of AI in Business in 2025.” MIT NANDA, 2025. Web.
³ Aditya Challapally, et al. “The GenAI Divide: State of AI in Business in 2025.” MIT NANDA, 2025. Web.
⁴ Aditya Challapally, et al. “The GenAI Divide: State of AI in Business in 2025.” MIT NANDA, 2025. Web.
⁵ Muneera Bano, et al. “Survey Insights on M365 Copilot Adoption.” Cornell University, Dec. 2, 2024. Web.
⁶ Muneera Bano, et al. “Survey Insights on M365 Copilot Adoption.” Cornell University, Dec. 2, 2024. Web.
###
About Canoe Intelligence
Canoe Intelligence (“Canoe”) is the platform for smarter alts management. We redefine alternative investment intelligence with AI-driven software that directly addresses the core challenges of private markets. Our technology empowers institutions, LPs, and wealth managers to future-proof their alts infrastructure, modernizing systems and providing a scalable foundation for long-term growth and compliance. By automating manual data processing with AI-native precision, Canoe helps clients reduce operational costs and risks, significantly lowering overhead and mitigating errors. Ultimately, our timely, accurate, and comprehensive data enables investment teams to drive superior investment outcomes through deeper insights and more profitable allocation strategies. With Canoe, it’s all about making Alts, smarter. Learn more at www.canoeintelligence.com.
MEDIA CONTACT:
Betsy Miller Daitch
Canoe Intelligence
+1 443-690-6200
bdaitch@canoeintelligence.com
