Goldman Sachs transitions to a new phase of artificial intelligence: Marco Ardenti on agents

Goldman Sachs is rethinking its artificial intelligence strategy amid rapid growth of agent tools and more mature corporate scenarios. Marco Ardenti, the bank's Chief Information Officer, is now talking not just about internal experiments, but about how a new wave of AI is changing development and everyday processes within the financial organization.

What Changed in 18 Months

Eighteen months ago, Goldman Sachs mostly talked about building its own internal AI tools and exploring how to apply models without reaching a mass audience. Since then, the market has accelerated dramatically: generative models have become more accurate, corporate use cases clearer, and agent systems have taken center stage—systems that don't just answer a query but can execute chains of actions. This is also changing the conversation within companies: from the idea of "let's try this" to the question "what part of the work can we hand over to the machine right now?"

For a bank like Goldman Sachs, this shift matters not just because AI is fashionable. If value was once measured by the quality of chat responses, now employee productivity is front and center: how much faster a developer writes code, how quickly an analyst drafts a report, how fast a team finds the information it needs within its own systems. In this sense, agent platforms like Claude Code become more than a standalone tool—they become a new interface to corporate work, where both results and control matter.

Where the Bank Is Looking

From the description of the new conversation with Ardenti, it's clear that Goldman Sachs' focus is shifting from the mere fact of implementation to the quality of application. The bank no longer thinks it's enough to simply have internal AI services. Now it's more important to understand which tasks these tools handle better than traditional software, where they actually save time, and where they require too much human verification. For a large financial organization, this is a practical approach: any automation is valuable only when it fits into existing processes, security requirements, and team accountability.

• Internal tools for employees
• Support for code development and working with code
• More complex chains of actions instead of single requests
• Assessment of real value, not just response quality

Why Agents Are Needed

The real question here is how the bank evaluates the agent approach specifically. We're no longer just talking about an assistant that summarizes documents or answers questions from a knowledge base, but systems capable of executing multi-step tasks. For engineering and product teams, this is especially noticeable: such a tool can read code, suggest changes, find dependencies, and perform part of routine work without constantly switching between windows.

But the higher the autonomy, the sharper the question of trust, action logging, and the human's right to make the final decision. That's why Goldman Sachs' experience is interesting not as a story about "a bank that's also implementing AI," but as a signal of the market's next stage. When such organizations discuss not demo pilots, but working scenarios and impact for specialists, it means the technology is passing a usefulness test.

Even if some experiments don't yield immediate results, the shift itself is important: corporate AI is gradually turning from a set of chatbots into an infrastructure layer that helps write, search, verify, and execute actions within the company.

What This Means

For the entire corporate AI sector, this is a marker of maturity. The conversation around AI is moving away from abstract promises toward the question of measurable efficiency: where agents save hours of work, how they integrate into sensitive processes, and which tasks can be entrusted to them without losing control. If this approach takes hold in banks, it will quickly become a benchmark for other large international companies as well.