Bankers are flooded with market signals, filings, and global events, yet lack tools to filter out what actually matters for each client, creating information overload that slows decision-making and weakens the impact of advisory. Research from the Boston College Center for Retirement Research shows that as financial complexity rises, investors reduce processing effort, and individuals with lower financial knowledge are more likely to become overwhelmed and default to simpler choices.
At the same time, client expectations for personalized advice are rising, while relationship bankers are expected to manage 400–600 clients each, according to an MIT Sloan Management Review case study, making consistent, high-touch engagement increasingly difficult amid fragmented data.
The consequences extend beyond inefficiency. The CFP Board reports that 57% of Americans have made regrettable financial decisions due to misleading or excessive online financial information, with younger clients aged 25-45 most affected. As markets move faster and information volumes grow, the gap between available data and actionable, client-relevant insight continues to widen.
Emerj Editorial Director Matthew DeMello sat down with Marco Argenti, CIO, Goldman Sachs, to discuss how AI can help relationship-driven financial professionals filter information, deliver highly relevant insights, and strengthen client relationships.
This article analyzes two core insights from their conversation:
- Optimizing knowledge work by maximizing return on attention: Boosting return on attention with GenAI by highlighting the clauses, concepts, or code that truly shape outcomes.
- Leveraging AI to filter and prioritize information: Using AI to help relationship managers sift through overwhelming data—market events, filings, and client activity—and identify what’s most relevant for each client.
Guest: Marco Argenti, CIO, Goldman Sachs
Expertise: Serverless computing, Internet of Things, and augmented/virtual reality
Brief Recognition: Marco Argenti is Goldman Sachs’s Chief Information Officer. Previously, he served as Vice President of Technology at Amazon Web Services from 2013 to 2019, helping drive the development of cloud technology platforms. He is also a Board Member Emeritus of the Pancreatic Cancer Action Network, reflecting his ongoing involvement in nonprofit leadership focused on health.
Optimizing Knowledge Work by Maximizing Return on Attention
Marco opens the podcast by sharing that one of the most important things genAI does well is to bridge deep knowledge gaps between experts and non-experts. He points to professions like law and software development, where expertise is built over decades. Lawyers learn to identify the most critical clauses in legal documents, just as developers learn to spot the most essential or risky parts of code.
The challenge arises when people without that deep domain experience, such as developers working with legal documents, are asked to operate in highly specialized domains. Marco argues that in such situations, AI can act as an adapter, translating expertise across different levels of specialization.
He explains that these systems are trained to identify what actually matters inside documents or bodies of knowledge, pinpointing the concepts, terms, and passages that are truly salient. Training AI systems based on saliency, he says, aligns with how modern AI models work at a foundational level.
Marco also describes this as a problem of attention. Knowledge, he notes, is not flat; every word or sentence does not carry equal weight. An experienced professional knows instinctively where to focus. AI models, particularly those built on transformer architectures, are designed to do something similar: assign more importance to what matters most.
He then frames the challenge as a question of “return on attention.”
Marco sees this as potentially transformative. In legal documents, AI can guide readers toward the terms that define the structure and risk of a deal. In software development, he notes, developers already use AI to highlight areas of code they may not fully understand, flag potential edge cases, or identify sections where bugs are likely to emerge.
“The sentence I like to describe this is “how do you measure your ‘return on attention’?” As in, if you have a certain amount of time, where do you actually focus your attention? Modern AI systems, particularly those built on transformer architectures, are designed around this principle — recognizing that not every word is created equal, not every word has the same meaning or importance.
By pinpointing the specific terms, expressions, or comments that are fundamental to how a deal is structured, these systems can guide readers toward what truly matters, representing a potentially transformative capability.”
–Marco Argenti, CIO at Goldman Sachs
The ability to summarize, focus, and pinpoint key areas, Marco says, makes knowledge workers far more effective. Ultimately, experience is about learning where to focus attention—and AI has the potential to compress that learning curve.
With an AI companion that highlights what deserves attention, professionals could reach higher levels of effectiveness much faster, fundamentally changing how expertise is developed and applied.
Leveraging AI to Filter and Prioritize Information
Marco frames financial services as a business built on relationships and advisory, not just transactions. Much of the value financial institutions provide, he says, comes from helping clients shape strategy — whether that’s around investments, mergers and acquisitions, capital raising, or integrating an acquisition after the deal is done.
He points out that people in relationship-driven roles, such as bankers, operate in an environment flooded with information. They are constantly exposed to macroeconomic shifts, microeconomic signals, public filings, corporate actions, and market events happening across geographies and time zones. The challenge is not access to information, but the ability to filter it and map relevance to specific clients.
“Imagine a future where a banker starts the day and an AI system surfaces an important development that occurred overnight, maybe in another market.
The AI system could then email this person to inform them that something has occurred that may have repercussions for specific positions, portfolios, or investment decisions. The system can automatically identify which clients may be affected, and suggest relevant talking points for follow-up conversations.”
-Marco Argenti, CIO at Goldman Sachs
He compares this shift to the evolution from mass advertising to targeted advertising. Where people once saw irrelevant ads, they are more likely to see offers aligned with their interests today. That relevance, he says, increases the “return on attention” for the customer.
Marco argues the same dynamic will reshape human and banking relationships. Client expectations around relevance will rise, and relationship managers will be empowered with AI systems that act as targeting, filtering, and augmentation engines — putting a complex world of information at their fingertips.
As human and banking relationships change, he says, so will the nature of professional relationships by raising the bar for how business is conducted. As service expectations rise, those who can leverage AI to deliver highly relevant, timely, and informed interactions will ultimately be the winners.