“Why do 80–90% of AI pilots in financial services never make it to production?”
That was the starting point for a recent webinar featuring Alexey Utkin (Head of Data & Analytics Lab, DataArt), Ed Simmons (Principal, Branch Brook Advisors LLC), and Thomas Mathew (Senior Director, Partner Technology for Financial Services, Microsoft).
The answer, as the panel made clear, lies less in the algorithms themselves and more in what underpins them: data infrastructure, governance, and organizational readiness.
The Pitfalls: Why AI Gets Stuck in Purgatory
According to Utkin, the industry is great at producing demos but struggles to scale them. Quick proofs of concept often mask deeper issues:
- Fragmented data and silos across legacy systems
- Poor governance leading to low-quality inputs
- Lack of monitoring and guardrails once models are in production
As a result, firms spend “100 times more effort” trying to operationalize AI than it took to build the first flashy prototype.
Mathew reinforced this, pointing to the business-technology divide. Boards push for AI transformation, yet 95% of organizations report low returns. The reason? “AI is only as good as the data,” he stressed, making unified, high-quality data platforms a necessity rather than an option.
95% of organizations are seeing a low return on AI, meaning high-end adoption but low transformation.
Modern Data Platforms: Simpler, Faster, Smarter
The good news is that technology has caught up with the challenge. Both Utkin and Mathew highlighted how cloud-based platforms like Microsoft Fabric, OneLake, and Purview are enabling firms to:
- Consolidate data without ripping and replacing legacy systems
- Ensure lineage, transparency, and compliance from day one
- Scale on a pay-per-use model, making AI adoption financially sustainable
Many modern data platforms are actually much simpler architecturally and technologically. Today, it is much easier to build, adopt, operate, and evolve a platform based on cloud technologies like those coming from Microsoft. So I think it is important for the reason for success: simplification.
Governance as an Accelerator, Not a Roadblock
One recurring theme: governance doesn’t slow innovation, it fuels it.
- Strong frameworks reduce data wrangling, freeing teams to build models faster
- Built-in compliance avoids delays caused by regulatory rework
- Transparency builds trust with stakeholders and regulators alike
“Good governance shifts the focus from fixing data to innovating with AI,” noted Mathew.
A Build-as-You-Grow Roadmap
So where should asset managers begin? The panel emphasized three layers of readiness:
- Technology Foundations – unify data, eliminate inefficiencies, adopt modern cloud platforms.
- Operating Model Transformation – embed governance, adopt data mesh/domain-driven structures, and treat data as a product.
- Business Value First – start with targeted domains like portfolio management or risk analytics, deliver quick wins, and expand iteratively.
This modular approach allows firms to capture value immediately while laying groundwork for long-term transformation.
From Risk to Resilience
The discussion also tackled risk management:
- Three tiers of AI risk – from low-risk engineering copilots to higher-risk client-facing systems.
- Model Ops and human-in-the-loop workflows as safeguards.
- AI literacy and cultural change to overcome resistance and empower employees.
AI, the speakers stressed, is an enabler, not a replacement. Success depends as much on people and processes as on technology.
Key Takeaways for Asset Managers
- Fragmented data kills AI, fix the foundation first
- Modern platforms like Azure Fabric & Purview simplify, unify, and scale
- Governance accelerates AI adoption when built in early
- Build-as-you-grow strategies reduce risk and deliver value now
- Success stories start with focused domains, not enterprise-wide migrations
Conclusion
The closing discussion made it clear that successful AI adoption is as much about people as it is about platforms. AI should be seen as a tool that empowers specialists, not something that replaces them. While it simplifies processes and drives productivity, it also changes the nature of work. That shift requires communication, education, and support to help employees adapt with confidence.
Equally important is building a culture of AI literacy and trust. Organizations that frame AI as an enabler—focused on reducing repetitive tasks, improving client services, and amplifying expertise—are the ones best positioned to gain long-term value.
It’s not that we can’t move quickly. It’s that we need to move holistically: technology, governance, and people together.
Ultimately, the panel agreed that the path forward requires a holistic approach: robust data governance, the right modern platforms, thoughtful automation, and people at the center. Success comes not from rushing pilots into production, but from aligning technology, governance, and culture into one cohesive strategy.
If your organization is exploring how to move beyond AI pilots and build systems that truly scale, you’re not alone. Many asset managers face similar challenges around fragmented data, governance, and adoption. That’s where the right data foundations make all the difference.
At DataArt’s Data & Analytics Lab , we combine deep expertise in data, analytics, and AI with over 20 years as a Microsoft-certified partner. Recognized as a Microsoft Solutions Partner with four Azure designations and an Advanced Specialization in AI Platform, we help organizations turn complexity into clarity and scale innovation with confidence.
Contact us to accelerate your transformation toward secure, scalable data systems.
