Beyond the Hype: Why Data 'Plumbing' is the New Frontier for AI Investment

The conversation surrounding Artificial Intelligence has officially hit an inflection point. For the past two years, the market has been obsessed with model capabilities—the "who has the smartest chatbot" race. But according to Marco Argenti, Chief Information Officer at Goldman Sachs, that era is ending.

In a recent episode of the Goldman Sachs Exchanges podcast, Argenti argued that the primary hurdle for AI adoption has shifted from the models themselves to the "plumbing"—the complex infrastructure and data integration required to turn experimental AI into a core business utility.

For investors, this shift is critical. It suggests that the companies poised to win in the next phase of the AI cycle aren't necessarily the ones with the most "advanced" models, but those with the most robust data strategies.

The Shift from 'Build vs. Buy' to 'Build and Orchestrate'

For years, corporate IT strategy was defined by the "build vs. buy" debate: do we develop our own software, or do we purchase an off-the-shelf solution? Argenti suggests that in the age of AI, that framework is obsolete.

Instead, the new mandate is "build and orchestrate."

"You can have the most sophisticated model in the world, but if it’s not grounded in your firm’s proprietary data and integrated into your existing workflows, it’s just a toy," Argenti noted.

The strategy now involves leveraging high-end foundation models—the "buy" component—but wrapping them in proprietary security, compliance, and data layers—the "build" component. This is where the real value is created. Companies that can successfully integrate their unique, internal data sets with these foundation models are creating a moat that generic AI adopters simply cannot replicate.

Productivity vs. Decision Support

While much of the current market focus is on productivity gains—automating emails, summarizing documents, and speeding up routine tasks—Argenti views these as mere "short-term" benefits.

The long-term competitive advantage, he argues, lies in AI-driven decision support.

"Imagine an environment where AI isn't just summarizing a document, but is actively helping you stress-test a hypothesis or identify a risk you hadn't considered," Argenti said. This is the transition from AI as a tool for efficiency to AI as a partner in strategic judgment.

For the investor, this means looking beyond companies that are simply using AI to cut headcount. Instead, look for firms that are using AI to enhance the quality of their decision-making, risk management, and hypothesis testing. These are the organizations that will likely see the highest ROI on their AI capital expenditures.

The 'Human-in-the-Loop' Imperative

A major risk often cited by skeptics is the "black box" nature of AI—the fear that companies will trust algorithms to make high-stakes decisions without oversight. Argenti addressed this by emphasizing a "human-in-the-loop" design.

At Goldman Sachs, the philosophy is clear: AI is a "force multiplier," not a replacement for human capital. By keeping humans in the loop, the firm ensures that accountability, nuance, and professional judgment remain at the center of the business.

This cultural integration is perhaps the most difficult part of the AI transition. Companies that fail to bridge the gap between AI output and human trust risk treating the technology as a "toy" rather than a core business utility. Investors should be wary of firms that treat AI as a "set it and forget it" solution; the winners will be those who treat it as an augmentation of their existing talent.

Key Takeaways for Investors

• Follow the 'Plumbing': Look for companies investing heavily in data infrastructure and integration. The ability to connect proprietary data to foundation models is the new primary indicator of AI success.
• Evaluate the CIO’s Role: The CIO is no longer just an IT manager; they are a strategic partner. Companies where the CIO is driving business model transformation are likely to be more agile and better positioned for long-term growth.
• Look for 'Force Multipliers': Distinguish between companies using AI to simply replace labor (which may face cultural resistance and operational friction) and those using it to enhance human decision-making and risk assessment.
• Beware of 'Toy' AI: If a company’s AI strategy is disconnected from its core workflows or lacks a clear path to proprietary data integration, it is likely an experimental project rather than a competitive advantage.

As we move deeper into 2025, the market will likely begin to reward companies that have moved past the "what is AI" phase and into the "how do we scale this" phase. The winners will be the ones who treat AI not as a magic wand, but as a complex piece of infrastructure that requires careful, human-led orchestration.