Artificial intelligence is transforming the fields of investment and asset management. Finance professionals must learn how to leverage predictive analytics and machine learning in order to maintain a competitive edge in the new AI-driven financial landscape. In a webinar on February 26, 2025, two experts in the field presented a round-table discussion titled “AI Tools and Techniques: Revolutionizing Investment and Asset Management in 2025.”
They touched on issues that will be discussed more fully at the upcoming conference, CFA Institute LIVE 2025 to be held in Chicago, 4-7 May, 2025.
The panel moderator, Brian Pisaneschi, [center photo] is an affiliate to the Research and Policy Center at CFA Institute. He summarized the timeline of developments in AI pertaining to finance before asking questions coming in from the remote, live audience.
Timeline of advances in AI
In Q4 2022, ChatGPT was launched by research company OpenAI. GPT stands for Generative Pre-trained Transformer, which is a type of large language model that uses a neural network architecture to generate human-like text responses. ChatGPT proved to be helpful for assistance in writing and basic industry analysis.
By Q2 2023, OpenAI launched a code Interpreter that was applicable to exploratory data analysis and basic visualization.
In Q3 2023, the capabilities extended to Retrieval-Augmented Generation (RAG), a technique that enhances the accuracy and relevance of generative AI models by allowing them to access and reference information outside their own training data, such as an organization’s specific knowledge base.
This meant AI could do basic financial statement summaries and carry out that standby of business, SWOT analysis, which identifies the strengths, weaknesses, opportunities, and threats of any given company or project.
By Q4 2023, custom GPTs were appearing, enabling real-time financial data retrieval, and basic investment analysis using natural language processing.
By Q2 2024, multi-modality in AI, or multi-modal learning, allowed chart analysis and chart recreation. Chain-of-thought reasoning models allowed code writing for Proof of Concept (PoC) in Q3 2024. A PoC is a small, functional prototype or demonstration that validates a concept or technology before committing to full-scale development.
As of Q1 2025, OpenAI Operator (AI agents) and ChatGPT Deep Research enable web scraping and research reports.
Comments by panelists
“What AI means to us,” said Andrew Chin, [LHS, photo] Chief AI Officer at AllianceBernstein, “is that we can access untapped, unstructured data we couldn’t access before. We will use that data to make decisions.”
“Being able to interpret text means we can understand regulatory filings faster,” he said, “but these models still struggle with predicting tomorrow’s stock price.”
“There is a field between quantitative and qualitative research camps. On the fundamental side, they have more depth. And the quants have more breadth,” he continued. “It’s exciting that the fundamental teams can expand their breadth more.”
Management meetings, texts, and many more features can be incorporated into AI models.
A question arose about bias in models. Chin pointed out that people are biased, too. “I have my own modeling and assumptions too. The difference is that I know what I was putting in. And what regression I was running.” Whereas with AI, “I have less flexibility in how to fine-tune a black box, plus I don’t know what data or what models they are based on.”
The foremost question coming from the webinar attendees was: will AI replace analysts? Chin answered, “Analysts with AI skills will replace analysts without AI skills. Ultimately, the person makes the decision.”
Analysts provide something that AI cannot. “The analyst will be asked, ‘do you uphold the fiduciary duty with this investment or not?’” said Eelco Fiole, [RHS, photo] Managing Partner at Alpha Governance Partners.
“The signal-to-noise ratio is tiny,” said Chin, referring to the likelihood of gaining information from data. “Your ability to think about why investments go up or down, that is still a value-added skill.” He encouraged the audience to test to see if AI tools function how they want them to. ‘Align them with your values.’”
“All firms have organizational values,” said Fiole. “How do you translate these, for example, to ESG values? It’s important to be very aware and not too broad-brush, if we want to project trust to clients and investors.”
The question arose about potential risks. “AI is a black box. There is information asymmetry,” said Fiole. “The risk is that we will violate people’s rights or even run the risk of harming people. Can we do something to mitigate the harm? We need to build in processes.”
“The regulators will want to ensure market integrity and a level playing field,” he continued. “Asset managers have a normative role to play here.”
“Along with great power,” said Chin, “comes great responsibility. Our company vets all uses of generative AI. A critical question is, ‘What data set was the model trained on? What data will you put in it? Public data seems okay, but client data may be problematic.’”
“The more data,” said Fiole, “the higher the R-squared. You must not be overconfident in what you’re doing. I’m in favor of innovating first, then regulating. Look at who the stakeholders are. The more robust you make the model, the more you can mitigate the harm.”
The question arose about newer models that have more transparency. Does that build more trust?
“To build trustworthiness,” said Fiole, “reduce the information asymmetry.”
Chin said, “2025 is the year of Agentic AI,” referring to the class of AI that focuses on autonomous systems that can make decisions and perform tasks without human intervention. “You have one agent looking at research reports and another one that does this… and another… that. Five agents, one super-agent.”
Fiole noted that in the future we can transfer developments in the blockchain world. “The analyst must pick out where to add value.”
Pisaneschi asked, “How can we keep up with the speed of things without moving too quickly?”
In the corporate world there is flexibility in how to leverage models. “To keep up,” Chin advised, “don’t lock yourself into one particular model. Also, having a benchmark is important—and continuing to monitor it.”
The discussion continued, covering such topics as skills that the new generation of financial analysts will have to learn; what techniques can mitigate the challenges; and speculations about what might happen five years down the road. The panelists were warming up for an excellent round-table discussion to be held at the CFA conference in May. ♠️
Click here to register for the upcoming conference, CFA Institute LIVE 2025 to be held in Chicago, 4-7 May, 2025.