According to Gerard Francis, JPMorganChase’s chief product officer for AI and data, all the excitement around AI is meaningless without a structured, enterprise-wide data strategy. Speaking during a customer spotlight at Snowflake Summit 2025 — where clients shared insights on building and scaling AI systems — Francis stressed that “without a solid data, AI, and governance platform, every AI experiment becomes non-replicable.”

As companies race to implement AI, few have successfully transitioned from pilot projects to full-scale deployment. A key reason, Francis explained, is that many organizations lack the foundational data infrastructure required to scale beyond early-stage trials.

From AI Experiments To Enterprise Impact

The pattern of AI adoption among Fortune 500 firms follows a familiar cycle: an initial proof of concept, followed by a public announcement, and eventually stagnation. While most companies have experimented with GenAI, only a few have moved beyond limited trials to full-scale, repeatable implementations that generate enterprise-level value.

“When I’m asked what enterprise AI really means at JPMorganChase,” Francis began, “it’s all about the data.” He pointed out that success isn’t about impressive models, but about solving real-world challenges in banking, asset management, fraud detection, and more — and doing so at scale. “How do you identify the most impactful use cases and scale them to deliver real business results?”

This is where most AI initiatives stumble. Analyst firm Gartner estimates that at least 30% of generative AI pilots will be discontinued after proof of concept by the end of 2025. According to Francis, the issue lies not in the power of AI models, but in infrastructure and governance readiness. “Without a clear infrastructure strategy and preparedness,” he said, “no level of AI investment will yield long-term value.”

This understanding led JPMorganChase to develop a unified platform that integrates data, AI, and governance into real-time workflows and reusable insights. The company’s internal GenAI chat tool, “LLM Suite,” enables employees to interact securely with large language models, protected by access controls and data usage rules. Early deployments focused on tasks like document creation, workflow optimization, and internal communication — areas where benefits are clear and risks are manageable.

“We had the proper governance and safeguards in place to protect data,” Francis said. “The idea was straightforward: deploy AI where it can deliver immediate value while maintaining safety.”

The Data Discipline Behind AI Readiness

What does it mean for a global financial institution like JPMorganChase to be AI-ready?

For Francis, it starts with data accessibility and permissions. “Is your data easily discoverable?” he asked. “Does it have the appropriate access controls so users only see what they’re allowed to see?” These aren’t just technical concerns but critical compliance requirements for a company operating under multiple regulatory frameworks and client classifications.

Next comes the challenge of unstructured data — documents, notes, spreadsheets, contracts, and more. These were historically difficult to process, but with retrieval-augmented generation (RAG) and other GenAI tools, they’re becoming valuable assets. Still, data quality remains essential. “Avoid duplicates,” Francis advised. “Ensure proper version control so users receive accurate responses.”

Structured data — often scattered across numerous internal systems — tends to be addressed last, yet can be the most powerful when properly integrated. That’s why JPMorganChase created Fusion — an external data platform for clients that functions as a “data factory,” orchestrating pipelines, standardizing formats, and preparing datasets for AI use.

JPMorganChase’s architecture spans multiple vendors and platforms, including Snowflake, which supports the company’s broader efforts to unify enterprise data for AI readiness. “Think of us as a data factory that operates at scale,” Francis remarked.

Governance As A Backbone

Talk to any enterprise AI leader, and governance will inevitably come up. At JPMorganChase, however, it’s not an afterthought — it’s built into the strategy from the beginning.

“In a regulated environment,” Francis explained, “you must ensure that the data used for any purpose is approved for that use.” This requires aligning AI applications with internal policies, regional laws, contractual agreements, and customer privacy terms.

Managing these controls manually would be impractical. JPMorganChase is working toward a system where “governance moves from being a human-driven process supported by technology to a fully technology-driven one.” Until then, scalability and compliance depend on how effectively governance is embedded into the AI development lifecycle.

AI Agents Are The Next Frontier

Imagine an AI system that doesn’t just summarize a document but also reconciles data, files reports, schedules meetings, and updates compliance records. These are early signs of what’s next in enterprise AI: autonomous agents capable of acting on behalf of users with minimal oversight.

While most current AI deployments still focus on text summarization or content generation, the shift toward agentic AI is already underway — and JPMorganChase is watching closely.

These self-directed systems, capable of reasoning and making decisions, offer major potential, especially in complex processes like software development, research, or operations. But Francis is cautious. “Agentic AI brings incredible value,” he said, “but also introduces new risks. That’s an area we need to deeply understand.”

He approaches it strategically. Rather than aiming to replace jobs, the goal is augmentation — helping teams work more efficiently with better insights and smarter tools. As Francis put it, “It’s not about whether you use agents, but whether you can truly solve a business need effectively. If you do, you either reduce costs or unlock new revenue opportunities.”

For a company as large and complex as JPMorganChase, AI only makes sense when it drives tangible value. “We identify the areas where we can create the most value, and that guides how we prioritize AI,” Francis said.

That value-focused approach also applies to ROI. While he didn’t share exact figures, Francis noted that the firm publicly reports on AI-driven value creation — with much of that still coming from traditional machine learning, though GenAI is rapidly catching up.

“If we can significantly reduce the cost of any use case, then ROI becomes easier to justify and easier to scale.”

The Long Game Of AI Leadership

Looking ahead, Francis hopes JPMorganChase will be recognized for tackling one of the toughest challenges in enterprise AI: creating a platform that unifies data, AI, and governance across multiple technology ecosystems.

“It’s often easy to do this with a single vendor,” he said. “But making it work across vendors is extremely challenging.”

Still, the rewards are clear. If platforms like Fusion can bridge the gap between experimentation and production, AI will shift from being a novelty to a standard enterprise tool.

In that future, the companies that succeed won’t be those with the most advanced models, but those with the strongest commitment to data discipline.

The above is the detailed content of AI Without Data Discipline Is Just Hype, Says JPMorganChase's CPO For Data And AI. For more information, please follow other related articles on the PHP Chinese website!