Part-time locker. Part-time philosopher. Full-time high-tech boss. Meet Marco Argenti, executive at the helm of Goldman's push to artificial intelligence.
Argenti is the chief information officer of Goldman Sachs, but his interests go beyond finances. Before joining Goldman in 2019, he was a top Amazon executive and helped build a cloud computing business. He also plays in a rock band in Seattle and has long been fascinated by the way philosophy, particularly the way technology shapes people and their decisions.
That perspective helped shape his AI's views. At Goldman, Argenti has overseen an engineering team of 12,000 people (one of the largest engineering teams on Wall Street) and guided the bank's push to artificial intelligence, including the deployment of GSAI assistants for internally generated AI platforms. Argenti sees AI as a tool that not only increases productivity, but also redefines your career path and saves your life.
In a recent interview with Business Insider, Argenti shared his views on the potential of AI that impacts the world with Wall Street, an engineer.
Below is a conversation with Argenti, edited for length and clarity.
Goldman has a workforce of around 12,000 coders to report to you. What do the different groups do and how will their work be affected by AI?
There are several ways to slice it.
The developer sits in a business unit. Developers working on trading stacks. Developers working on a private wealth advisory website. Or, for example, developers working on marquee platforms for in-house and corporate clients.
Then there is the central group (over 3,000 people) sitting in what we call core engineering. These people are creating common platforms that are used across the enterprise, including data centers and cloud, networks, storage, middleware, and higher-order services such as identity and access management.
We were one of the first companies to establish an interesting profession within the engineering industry known as Strats. They are similar to Quant Developers, but they are also coders, writing code specific to risk and pricing models. Of these, it exceeds 2,000. Those people are unique as they are still increasingly changing into data science today.
Do you think AI can replace some coders? Or is it about how developer roles evolve?
I think we will redefine the profession through exchanges at the task level rather than at the work level. The metric I want to see is the output of the organization. What can you generate for your business?
Most CEOs say, “There's an almost endless supply of great ideas we have to wait for, and it's not just within budget.”
With AI, you can break the output of its paradox and scale, whether or not you add people. I see it as a multiplier of force.
Related Stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
Does Goldman expect to have more, fewer, or roughly the same number of coders in five years as a result of AI?
I think there's more output. For example, there are expectations of changes to the workforce composition, such as more senior engineers than juniors. When you ask me the net number, up and down – it's hard to say at this point. It really depends on how production is growing and how it translates into growth.
We are in the stage of growth. On the business side, there are many inputs to do more and grow faster. So, it's good to have options, but the reality is that it depends on your appetite for reinvestment.
How has AI changed its own workflow? What do you personally use?
I had a huge backlog of projects like my charity and my music (my charity and my music, etc.) like a backlog of ten years. I can literally do them this weekend.
At work, we have a GS AI assistant. This is a very powerful AI tool. It uses the latest models and is connected to corporate data. That's my first question.
There is a lot of fear in Agent AI, or systems that act autonomously to complete tasks. What are your concerns?
The risk was our best heart. Autonomous operation of agent systems is powerful. You need to be very careful about reducing risk rather than increasing it.
One of the important things for us is how agents decide what they can do. Put rails around agents to limit their qualifications.
Think of it like autonomous driving. The agent's job is to disable access to the highway, prohibit outside the route, and prohibit speeds beyond the limit when moving from point A to B to deliver packages on the surface. Even if you come up with unpredictable output, there are only a few actions you can actually take.
Does AI change the meaning of building a long-term career in finance? What skills will determine your future success?
We can help people accelerate our paths. Becoming a manager from an individual contributor by starting management before AI agents manage people. They learn to explain, explain and break down tasks. They learn to oversee, trust, but validate. This technology accelerates leverage and delegation skills.
Historically, you tend to do a lot of work yourself, then you get leverage and resources that you can delegate. I think you have your own expansion team now, so you accelerate your seniority in managing resources. Also, AI is a board that resonates with ideas, so it opens up to the big picture.
Do you think AI can free employees to achieve a better work-life balance, or is it more likely to raise expectations?
Note Returns will be one of the biggest currencies in AI. For example, all developers have a protection shield for the agent, allowing them to focus on what creates the maximum value to remove noise.
It creates the possibility of a better balance, but whether or not it happens depends on how the company decides to use productivity.
What keeps you at night? Are you afraid of certain vulnerabilities that could cause your entire system to collapse?
One of the worries is that people may rely on AIS and create self-satisfaction.
If you have high skills and high delegation, it is a very good quadrant. If your skills are low and your delegation is high, it is not a good quadrant.
My biggest fear is overreliance without the ability to understand and supervise.
Imagine that autopilot has gotten so much better and at some point you decide you don't need a captain on the plane. That's not a good idea. You will always need a captain.
Looking at 10 or 20 years from now, what is the biggest way for AI to change how people live and work?
There is always a risk that there may be even greater digital disparities between people who are really good at leveraging AI and those who are not. Or perhaps among people who can afford different levels of access.
Today you have a $20 subscription and there is a $200 subscription that will give you more. What if I had an incredibly good $2,000 or $20,000 subscription?
I personally don't think it will happen because that could be the case with computers. At one point, perhaps it was, but the PC and the Internet created counter actions. Still, that's a potential problem.
The other is energy consumption. Energy distribution issues do not arise as AI wants not only to be more powerful but also to be more efficient.
And, like cancer charity, they applied technology to causes outside of Wall Street. How do you think AI is contributing to the progress of health, science, or society as a whole?
We are in the age of medical software. It's like DNA mapping, mutation sequences, pathway understanding, immune system reprogramming – like programming code.
A few years ago, I lost my best friend who had pancreatic cancer. Soon after that, I joined the advisory committee of the Fred Hatch Cancer Center in Seattle.
Our mission is to give patients 1 million years of survival in five years. I think that can be done by increasing standards of care and discovering new combinations. The early signs are strong.
At the heart of that, we ask questions, can we cure cancer?
