Goldman Sachs CEO teases “6 specific processes’ it’s attacking with AI
Solomon admits it takes “an enormous amount of work to bring people along”
Financial Services
It was a Very Good Year for Goldman Sachs – which ended its fiscal 2025 last week with a record $3.6 trillion under supervision and profits up 12%.
It closed its acquisition of a tech-centric venture capital platform, Industry Ventures, earlier in January and Goldman’s CEO David Solomo could be heard opining as the deal closed that “technology and AI-led innovation should be a catalyst for significant capital markets activity in 2026…”
On the AI front, last quarter it announced the “One Goldman Sachs 3.0”, which Solomon described as “our new operating model propelled by AI.”
Goldman Sachs 3.0: “Propelled by AI”?
Analyst Mike Mayo from Wells Fargo had a sharp question on January 15’s Q4 earnings call however – pressing Solomon on this AI-propulsion.
As he put it: “Whenever I ask about AI, it's always answers at a 10,000-foot level. Like ‘it's transformational, it's a game changer, it's a super power.’
“We all get that. But what are you hoping to achieve?”
There are, crudely, tactical and strategic opportunities, Goldman’s CEO suggested, but he was not ready to break out any meaningful details.
See also: Not the CIO’s job? Getting your organisation AI agent ready – “the promise and the peril”
What he could reveal was that “to change operating processes in the firm, we've identified six specific processes that we're attacking – [it] takes an enormous amount of work to bring people along,” he admitted to Mayo.
“We started doing this in the fall. We're making good progress. To be honest, I had hoped to give a little bit more transparency at this earnings call, but we don't have the full confidence to put information out publicly.”
Solomon added: “But we are committed to giving you more over the course of the next quarters so you can track with us the efficiency progress…”
See also: The Big Interview: Bank of England CIO Nathan Monk
Near-term, the bank is putting “these models, these tools, these applications [into the hands of its staff]... they're very good at playing with them and figuring out how on a day-to-day basis they can use these tools to make themselves more productive, to do more, to affect our clients more.”
Speaking to The Stack back in early 2024, Goldman Sachs’ CIO Marco Argenti was candid on where he was focusing when it came to AI.
He told us: “We started to look at what are the buckets of focus that we want to start with. Bucket one is sustaining our business growth; the other bucket is how you make your business more productive; developer productivity and general productivity… if you look at this as a spectrum, the low hanging fruits on the left, were the very first thing that we prioritised.
“That was a good starting point, because language models operate more effectively, in the more limited vocabulary that is computer software, rather than the entirety of English" Marco Argenti
“If we shift right a little bit, we started to look at operations; at workflow; we started to look at summarization of public documents, all the Ks and the Q's [public filings]; doing complex analysis on earnings reports; can we get sentiment analysis around inflation by looking at data sources?
Goldman's CIO told us: “Then if I move right a little bit more, now we are in the bucket of ‘how you grow your business’ [using generative AI]. This is more about the Alpha generation; that part is a little bit more difficult to quantify today.
“There are some areas that are almost on the research side; for example, how do you deal with time series [data]? I'm a believer that time series [data] is kind of its own modality: you have a modality for text, you have a modality for images, which is different, and then maybe you have a modality for time series data [which has] incredible applications.”
“[There is potential here to go] from regression type of predictions into more advanced forms of prediction, which leverage many more parameters than any traditional ML or, looking at portfolio compositions based on certain levels of risk and trying to go very deeply into all the factors that could determine how we calculate an efficient frontier…” Argenti added.
But “AI is the ultimate amplifier of your data provenance… if something is incorrect at the source, gen AI is actually going to make it plausible, but it's still going to be incorrect. So having good ground truth is so important.”