The AI Automation Gap: Why 76% of Finance Leaders Are Investing But On

When Savant and Engage fi released their latest analysis this week, the numbers told a story that finance executives have been reluctant to admit: the gap between AI ambition and execution has never been wider. While three-quarters of financial services leaders are pouring money into artificial intelligence and automation initiatives, a mere 6% have managed to move beyond pilots into production at scale.

“We’re seeing a fundamental disconnect between what organizations want AI to do and what they’re actually equipped to deliver. The technology isn’t the bottleneck—it’s the organizational readiness.” — Financial Services Technology Analyst

The $32 Million Bet on Fixing What’s Broken

Selector’s announcement of a $32 million funding round this week underscores just how valuable the observability problem has become. The company’s AI-driven platform, which fuses large language models with knowledge graphs and causal reasoning, has seen its annual recurring revenue nearly quadruple as Fortune 20 and Fortune 1000 organizations scramble for solutions that actually work.

The investment comes at a pivotal moment. Financial institutions are sitting on massive AI budgets but struggling with a reality that vendor demos rarely capture: connecting AI systems to real enterprise data while maintaining governance, security, and auditability is exponentially harder than running a proof of concept.

The infrastructure challenge has become the defining bottleneck. Companies like Crusoe are responding with platforms like Command Center, which combines deep observability with orchestration across managed Kubernetes and AutoClusters. The goal isn’t just to run AI workloads—it’s to maximize every GPU hour and reduce the time teams spend wrestling with black-box infrastructure instead of training and deploying models.

From Cloud-Only to Technology Value Management

The FinOps Foundation’s sixth State of FinOps survey, also released this week, reveals how dramatically the landscape has shifted. A staggering 98% of the 1,192 respondents now manage AI spend—up from just 31% two years ago. AI value management has become the top forward-looking priority.

But the survey also shows how FinOps itself has evolved. What started as cloud cost management has transformed into broad technology value management: 90% now manage SaaS spend, 64% licensing, 57% private cloud, and 48% data centers. Organizations are increasingly expected to self-fund AI investments through efficiency gains.

“The question isn’t whether to invest in AI—everyone is doing that. The question is whether you can prove the value fast enough to justify continued investment.” — FinOps Foundation Survey Respondent

The Governance Layer Everyone Forgot

Redpanda’s launch of an AI Gateway this week highlights another critical gap: governance. The platform serves as a unified access layer between applications, AI models, and MCP services, centralizing routing, policy enforcement, cost controls, and observability across all AI traffic.

Similarly, Empromptu’s expansion of its AI App Builder addresses what the company calls the “bolt-on governance” problem. By embedding Golden Pipelines for data readiness and AI Policies directly into the app generation process, the platform produces auditable, policy-aligned AI systems by default.

For financial services specifically, the stakes couldn’t be higher. The Savant/Engage fi analysis notes that just 16% of banks and credit unions have an enterprise-wide AI roadmap. The rest are left with scattered point solutions, vendor dependency, and governance nightmares when AI-driven decisions raise fair-lending, fraud, or customer-complaint issues.

The Road Ahead

Industry observers are watching closely to see which approach wins: the all-in-one platforms promising to solve every problem, or the specialized tools that integrate into existing infrastructure. Databricks’ move to make Agent Bricks Custom Agents generally available suggests the market is demanding more flexibility, not less—letting developers use their preferred models and frameworks while operating against governed enterprise data.

The coming months will reveal whether the 6% who have achieved scale can expand their advantage, or whether the tools hitting the market this quarter will finally bridge the automation gap. For now, one thing is clear: the race isn’t just about who has the best AI models. It’s about who can actually put them to work.

This article was reported by the ArtificialDaily editorial team. For more information, visit Solutions Review, FinOps Foundation, and Crusoe.