Goldman Sachs CIO Marco Argenti said his firm is betting on generative AI and technology that won't replace human workers but make them superhuman and prioritize the customer.
Speaking during the keynote at the Domino Data Lab Rev 4 conference, Argenti outlined a bevy of thoughts on digital transformation and staying ahead of the AI curve. Among the key takeaways from the Rev 4 keynote:
Being a fully digital company. Goldman Sachs has roughly 45,000 employees and 12,000 are in technology. "We are a fully digital company. We've determined technology is the lifeblood of the organization. It was always tech that started in the back office, but it has worked its way up, so you notice it," said Argenti (check). "Technology has been disruptive in every sector and you have to put software at the center of what you do."
The beauty of being an outsider. Argenti, who was Vice President of Technology at Amazon's AWS before joining Goldman Sachs in 2019, figured he needed some sort of financial background. After an interview with Goldman Sachs CEO David Solomon, it became clear that being an outsider was a plus. "If you want to disrupt you need an external viewpoint of some kind," he said.
AI's importance. Goldman Sachs has plenty of AI and machine learning technology and generative AI will be another wrinkle over time. "There's a realization from the very top of Goldman Sachs that you cannot be a leader in your industry without technology. It's even more so with AI. The leaders in any industry will be leaders in AI," said Argenti.
The mental model for AI management and prioritizing. Argenti said the scaling of AI can rattle humans. "It's hard to say you're an expert in anything with AI," said Argenti. "It comes down to how you think about the future."
He said his mental model revolves around using AI to augment humans by prioritizing what makes Goldman Sachs special. To Argenti technology comes down to saving money and making money. "There's a bias toward saving money you have to fight. You can over index for productivity, but being fit doesn't make you a champion," said Argenti.
Argenti's approach is to use AI to "assure the synergy between human and machine and be something superhuman." His focus areas for AI revolve around impact and quantity.
Developer productivity is a big area since productivity improvement can boost returns on high impact, high-cost workers. Bankers are another area where AI can help by making meetings more valuable for clients and the firm.
Education and training are also areas where AI can be a force multiplier. "There is a lot of knowledge in people's heads and generative AI can provide summarization and learning content," explained Argenti. "What can we do to shorten the apprenticeship?"
He also said that ultimately every major persona within Goldman Sachs will have an AI copilot.
Generative AI. "When you talk about generative AI, the first question to ask is 'what is it good at?'" he said. Indeed, Argenti said more traditional AI and machine learning are better at certain tasks. Generative AI can take large amounts of unstructured data and summarize it and it's good at connecting dots. This ability could mean generative AI can connect dots to ultimately drive investment returns.
AI architecture. Argenti said Goldman Sachs is internally trying to get consensus over the AI strategy, but it's likely there will be humans, large language learning models and a "swarm of small models to create a constellation of specialists." "Humans will focus on relationships, intuition, wisdom and instruction," he said.
Whether this effort is successful will depend on the following:
- Quality of data.
- Access to clients.
- A set of people to provide a flywheel of feedback that creates something unique to Goldman Sachs.
Data strategy. Argenti said Goldman Sachs has been investing in its data platform to have one single version of the truth, transparency to lineage and easy discovery. "The data is the foundation of any AI effort," he said.
The future of engineers. Engineers will move from focusing on how something gets done to what and why, he said. "You'll have to understand the customer and customer benefit," said Argenti. "We're all becoming prompt engineers. With generative AI it will be more about writing instructions than code. It will be all about the mental model."
Argenti likened the change in the computer science profession to when compilers were abstracted. "Engineers will have to conceptualize and understand why they are doing something," he said.
Changing a culture with data. Argenti said he wants his team to move from a vertical approach to one that's more horizontal and distributed. To change culture, you need trust, he said. Data transparency will be key. "Data creates transparency. Transparency creates trust. Trust shifts culture," said Argenti.