How to navigate the ‘rule of toos’ in GenAI and capital markets

AI has transformed capital markets by powering algorithmic trading, accelerating signal detection, and improving risk management – enabling firms to make faster, more informed decisions. But generative AI (GenAI) presents a unique set of challenges for the industry, put neatly by Mark Palmer, a recent guest on our Data in the AI era podcast, as the ‘rule of toos’.

Mark suggests that GenAI adoption is a challenge in capital markets because:

• Decisions must be made too quickly
• Data moves too fast
• Data is too structured

Let’s unpack each of these limitations and consider where GenAI could still unlock transformative value.

Decisions must be made too quickly

Capital markets operate in high-stakes environments. There’s a decision to be made at every moment – a pace no human can sustain. Scaling up human teams to keep pace is equally impossible. So you need some sort of input from an algorithm, not to replace human decision-makers, but to augment and empower them.

If you don’t make decisions quickly enough, there are opportunity costs. You end up leaving things on the table, and your competitors might make decisions faster than you. Fortunately, traditional AI and ML already helps capital markets teams meet such demands, accelerating time to insight, enhancing predictive capabilities, identifying trends and assessing risks. It can also help you make split-second decisions in high-frequency trading where speed of execution is vital.

However, such technology is based on structured data, and cannot leverage the unstructured data that could offer context and other benefits to the decisions organizations make. Right now, though, GenAI just isn’t fast enough to influence real-time decisions in such fast-moving markets. Accuracy concerns further complicate adoption.

Data moves too fast

It’s a similar story with the sheer volume of data capital markets have to deal with. The unrelenting, and growing, flood of information is an ongoing colossal challenge, and one in which even minor improvements in terms of responses can have significant impact.

The risk of falling behind the data deluge is yet more missed opportunities. If you’re constantly bombarded with data while struggling to process what you already have, you get backed up and cannot utilize the data expeditiously. Adding to the complexity is the increasing recognition of value locked away in unstructured data, which remains largely untapped.

I’d say existing technologies arguably handle structured data efficiently. Traditional AI excels at this task, scaling to meet your needs and providing the real-time insights you need to inform timely decisions. But when you need to tie in insights from unstructured data too, you start to hit a bottleneck. Extracting actionable insights requires a whole different data flow that involves identifying data of interest, chunking it, embedding it, and running it through a large language model.

This process takes a lot of time and can be resource-intensive. Latency remains an issue, and speed to insight may not be fast enough. The future holds promise, though. Multi-agent frameworks, as discussed in a previous post, could soon bridge the gap, integrating disparate data types to keep pace, and accelerating analysis and ideation.

Data is too structured

As already discussed, capital markets have traditionally relied heavily on structured data. These processes have proved effective, and many organizations are understandably hesitant to disrupt what works. And with GenAI being inherently better suited to unstructured data, it’s no surprise there’s doubt about its deeper integration into the industry.

Any shift to GenAI could be seen as overcomplicating what already works. We’ve seen experiments using LLMs for time-series (structured) data, and while they can work well, they aren’t necessarily better than traditional AI. So the question is, why use ‘expensive and heavy’ when a lightweight algorithm that has worked for years gets you comparable (or even better) results – for a fraction of the cost?

Again, I’d suggest keeping an eye on opportunity cost. Organizations should want to bring in unstructured data —like analyst reports, earnings call transcripts, breaking news headlines, and social media sentiment — due to its potential to provide greater insights and a competitive advantage over those companies not using it. There are opportunities to harness GenAI’s ability to link structured and unstructured insights, or to extract structured data from unstructured data. All of which can help capital markets make more informed decisions, faster.

So this isn’t really about overcomplicating what already works, it’s about laying the groundwork for what’s coming next. Failing to adapt as GenAI tech evolves will leave organizations at serious risk of being left behind.

Prepare for the future

Traditional AI has long worked well for many capital market needs. And for GenAI to make inroads into this demanding environment, it has to offer more:

• Superior insights regarding analysis, risks and trends, alongside the richer context afforded by incorporating unstructured data.
• Speed and accuracy that meet the stringent demands of the market, because compromising either is not acceptable.

In short, when we can do millisecond analysis and bring unstructured data in, with the suitable level of accuracy, we’ll be in a good place. But the ‘rule of toos’ suggests we’re not there yet. GenAI tech still lags behind traditional methods for much of what capital markets need. Yet we also know overcoming these challenges will change the game.

The unknown is when a tipping point will occur – and whether your organization will be ready when it does. Given how rapidly tech and markets evolve, proactive preparation is vital. At the very least, I’d recommend you:

• Explore GenAI’s potential to determine how it could enhance existing workflows and unlock new opportunities
• Consider combined approaches by way of hybrid models that leverage the strengths of traditional AI and GenAI, integrating structured and unstructured data
• Find the right balance between what’s achievable and affordable today, while laying foundations for the competitive demands of tomorrow

We’re in a sector geared toward survival of the fastest. Success will belong to organizations that are best prepared for a future that may arrive more quickly than you imagine. By embracing GenAI thoughtfully, deliberately and strategically, capital markets can transform the challenges of the ‘rule of toos’ into opportunities for growth and innovation. The rest risk being left behind in a market where speed, insight, and adaptability are everything.

For more information on what it takes to build a successful AI program, read our AI factory 101 series. Discover why KDB.AI is a crucial part of the AI factory.