Firms are confronting a blunt reality: Can they productionize AI or are they simply funding expensive POCs that drift nowhere.
78% of firms are running AI POCs, but only 14% have made it to production. In financial services, that figure rises to 21%, the highest of any sector. Yet that’s still a remarkably low bar, especially when you consider that firms will spend $765 billion on AI this year, growing to $1.6 trillion by 2031.
Asset managers face a particularly acute version of this challenge.
Most have given their employees access to Microsoft Copilot but are measuring success by usage. Some have layered agents on top of existing operations but have not reimagined their underlying workflows. A handful have pushed into AI enabled client engagement but with limited understanding of the readiness of their underlying data.
Three obstacles account for most AI POC failures:
First is data quality.
Firms that can’t easily answer foundational questions like “Who are our clients?” and “What products do we sell?” are going to struggle. The shiny AI tools on the front end are only as good as the back-end data management capabilities. Firms still working on their data foundations have likely already missed the AI wave.
Second is AI Governance.
Most organizations treat AI Governance as something that is owned by Legal & Compliance or the Technology team. It needs to be fully owned by the business. AI Governance is the gritty operational layer that sits on top of an existing Data Governance framework. Few firms know how to build AI Governance, yet it’s one of the most critical components to drive production success, especially for agentic AI.
Third is institutional knowledge.
Almost every firm has hired or is recruiting an AI expert. There’s no shortage of self-declared experts. Yet AI adoption is still in its infancy, and successful deployments are rare. There can’t possibly be that many ‘experts’. AI is moving too quickly for firms to onboard self-declared ‘experts’ who then need to spend months understanding the intricacies of internal data ecosystems to execute. The AI execution expertise firms need is likely already embedded deep within their organization.
KPMG’s February 2026 AI Pulse Survey puts it plainly: “The barrier isn’t compute or capability anymore. It’s whether leadership can govern systems that operate autonomously, evolve talent faster than the technology moves, and architect trust at the scale these systems demand.”
The same report goes on to say, “Over half (53%) of Asset Management and Private Equity leaders expect their organization to achieve measurable ROI within the next 12-24 months.”
The AI hype cycle is coming to an end. There is now limited runway ahead to execute and demonstrate tangible results.
The AI race is no longer about what tools firms select. It’s about how firms go about deploying AI into production with a positive ROI.
The firms that cross the production threshold in 2026 will build a competitive advantage: faster operations, lower costs, and AI-driven client engagement at a scale existing workflows can’t match.
Beyond 2026, the AI differentiator won’t be company size or budget, it will be execution excellence.
Firms relying on off-the-shelf front-end vendor tools will have the same commoditized AI capabilities that their competitors are using. Firms that have built proprietary AI platforms, robust data pipelines from high-quality sources, small language models that generate unique insights, and governance frameworks to ensure high quality output will have built durable competitive moats.
The firms still running AI POCs in 2027 won’t be the late adopters. They will be the acquired.
Listen to a discussion on this article