When ChatGPT launched last November, it revealed that artificial intelligence had gotten smarter and faster than most people had previously imagined. It became clear that AI may soon transform so much of our lives, including every facet of knowledge-based industries.
Change is coming especially fast to the financial sector, where companies had already spent years deploying increasingly sophisticated algorithms to parse ever bigger data sets. Suddenly, a wave of innovative new AI products and applications are coming onto the market that are designed to make companies and investors more efficient and profitable.
Applications of technology in finance have been advancing for centuries, from the abacus to the electronic spreadsheet to the Bloomberg Terminal. AI applications will transform the financial services industry too, but the technology is still nascent. As such, this is the ideal moment for experimentation, especially with boards and management teams already pressing their teams to figure out how firms should implement the latest AI technologies and deal with new AI-related risks.
What makes this latest burst of AI innovation so different from what came before?
The release of large language models like GPT-4 essentially allows every writer to become a coder, and vice versa. You no longer need a technical expert to turn your questions and most data queries into algorithms and translate the outputs from the algorithm back into English.
So long as you can ask concise questions, you can get most of the answers you need. In effect, English has become the new coding language.
How can we start deploying AI across our business?
Think of your AI investments as a barbell.
On one end of the barbell are visionary applications for AI – where you think to yourself,
“It would be a game changer for our business if we could do X.”
Don’t worry about whether you currently have the tools or in-house expertise to do X. Make your blue sky AI aspirations known throughout the organization and especially on your tech and data teams. You don’t want today’s technical limits to stifle the development of your most lofty ideas. Given the rate of advancement in AI, those ideas may be possible to implement in the not-so-distant future. In essence, you want to ask for magic, because magic could be available sooner than you think.
On the other end of your AI investment barbell are the immediate use cases for it in the here and now.
You should implement as many off-the-shelf solutions as you can across your business so everyone gets a more intuitive understanding of the strengths and weaknesses of AI. Most of the applications we’re all already using have AI tools or plugins, like Microsoft Office’s Copilot feature and Salesforce’s Einstein. Zoom has several AI features, including one that creates a transcript of your meeting and a bulleted summary of what was just said.
One of the most compelling new tools for the financial industry is an extension of ChatGPT called Code Interpreter, which takes input from virtually any data source including Excel documents. Code Interpreter can then clean the data, chart it, and draw different inferences.
For example, an institutional investor might take global weather data and put it into Excel along with same store sales data for individual retailers. Code Interpreter could test to see where there are correlations between the weather and sales at the retailers. Then it could present its conclusions to you in a written report that includes charts and verbal explanations.
Introducing off-the-shelf AI solutions throughout your organization is a low-cost, low-risk way to experiment with AI that will help everyone understand AI’s capabilities.
What are the most pressing risks of AI adoption?
Three stand out. The first is vendor dependence. New AI tools, technologies, and companies are being developed so quickly that today’s preferred solution could be obsolete by tomorrow. So this is a moment to maintain maximum flexibility and to avoid any AI implementation or partnership that you can’t walk away from quickly if developments require it.
Second, you have probably heard about large language models generating “hallucinations.” These are false statements that read like true things you have read before.
Hallucinations seem so credible that you believe them. And they are already getting professionals into trouble. Recently, lawyers submitted a filing in Manhattan federal court that featured fictional cases that ChatGPT had made up. The lawyers had even specifically asked ChatGPT to verify that the cases were real, and it falsely told them they were. The offending lawyers were fortunate to emerge with only a minor $5,000 fine, but it isn’t hard to imagine how a ChatGPT hallucination could lead to more catastrophic consequences for a financial firm.
Finally, and perhaps most importantly, is data governance and security. AI technology and large language models will eventually become commoditized. When that happens, your firm’s differentiator will be the proprietary information that you put into these models.
The proprietary information that you are putting in – say the corpus of research on your data for the past 20 years – is incredibly valuable. It can be used to train a large language model to write cogent insights. If someone gains access to the information that differentiates your company and uses it to compete with you, that could spell big trouble. So you will want to ensure that all your vendors perform data security audits and have effective verification provisions.
Artificial intelligence is such a powerful tool that businesses will find it imperative to introduce it to most of their teams and their workflows. If you put the right guardrails in place and think carefully about what AI can potentially do for your firm, the rewards of AI will far outweigh the risks, now and in the future.
Spenser Marshall is the Chief Data Officer of Sundial Data, a direct data sales subsidiary of M Science, a portfolio company of Leucadia Investments, a division of Jefferies Financial Group.
This is the first in a series of articles on the implications of AI for financial firms. Next up: How AI is Democratizing Access to Alternative Data.
