AI is only as powerful as the data foundation it’s built on

There’s no shortage of excitement around AI in investment management right now, and for good reason. The ability to search across thousands of documents, summarize complex research, and find relevant information in seconds is a significant step forward for investment teams. I’m old enough to remember as a junior analyst having to physically thumb through folders of research to get an answer.  Even post digitization you had to know where to look.  Today things that previously took hours can now take minutes. But I think there’s a bigger conversation we need to be having.

For all the attention being paid to what AI can do with investment information, we’re spending considerably less time talking about the quality, structure, and context of the information we’re giving it, and in my view, that’s where much of the real opportunity lies.

Investment firms don’t have an information problem, they have a knowledge management problem. Think about how much research an institutional investment team generates over the course of a year: manager meetings, investment theses, due diligence reports, internal discussions, investment committee materials, portfolio reviews, and the list goes on. Now lets multiply that by ten or 12 years, the average term of a private markets fund investment for example.

There’s an enormous amount of valuable knowledge sitting within these organizations. The problem is that much of it is fragmented across documents, emails, spreadsheets, shared drives, Slack, and, perhaps most importantly, the heads of individual team members. Finding a document in the depths of one of these systems is one thing, while understanding what it means in the context of everything else the firm knows is quite another.

Take something as straightforward as revisiting an investment manager your team evaluated several years ago. Your team might be able to find the original due diligence report fairly quickly using a basic search, but what about the conversations that followed? The concerns raised during the investment committee meeting? The reasons the team decided not to proceed? What has changed since that original meeting? That context is often far more valuable than the document itself, yet it’s frequently buried across different systems or lost entirely when people leave the firm, responsibilities change, or information isn’t captured consistently.

Despite what some may think, AI doesn’t automatically solve that problem. One of the things I find interesting about the current wave of AI innovation is how much of it focuses on making information easier to access. That’s clearly valuable, and I’m a big believer in the potential of AI to improve how investment professionals work, but we need to be careful not to confuse faster access to information with better access to knowledge.

You can put an incredibly capable AI model in front of a collection of investment documents and get some impressive results. It can summarize reports, identify themes, and answer questions about the content. But what it can’t automatically do is recreate institutional context that was never captured in the first place. If the reasoning behind an investment decision lives in someone’s inbox, or it was a discussion that was never documented, or a piece of research hasn’t been connected to the relevant investment entities and workflows, AI has a much harder job. The bottom line is that having the information is only part of the equation. Understanding how different pieces of research connect, and the context behind them, is what makes that information truly valuable. An investment thesis from three years ago, a manager meeting from last quarter, and an internal discussion from last week might each tell part of the story. That’s why I believe the structure of institutional knowledge is going to become increasingly important as AI adoption accelerates.

Here at Bipsync, we’ve spent the past 15 years thinking about this. At that time, the challenge we were trying to solve wasn’t AI, it was helping investment teams manage the enormous volume of unstructured research and information they were producing. We saw firms investing significant time and resources into generating research, but struggling to capture and organize it in a way that could be consistently shared, accessed, and reused. The technology available at the time wasn’t particularly well suited to the way investment professionals actually worked, so we set out to build something different.

Over the years, working closely with institutional investors, we’ve learned that managing investment research is about much more than organizing documents. It’s about understanding how research is created, how investment teams collaborate, how information moves through an organization, and ultimately how decisions are made. These are complex processes that vary from firm to firm, and they’re not something you can simply solve by adding AI to a collection of documents.

Connecting research to managers, funds, securities, people, meetings, and investment processes is what allows firms to build a more complete picture of what they know. That was important long before generative AI entered the conversation. Today, it’s arguably more important than ever. The better organized and more accessible that knowledge becomes, the more effectively firms can put AI to work to support informed investment decisions.

The real opportunity here is institutional memory. I think one of the most interesting applications of AI in investment management is its potential to make institutional memory genuinely accessible. Imagine being able to ask a question about an investment manager and immediately draw on years of relevant research, meeting notes, due diligence, and internal perspectives, not just to find out what was said in the latest report, but to understand how your firm’s thinking has evolved. Why did we originally like this manager? What concerns did we have? Were those concerns addressed? How does our current assessment compare with the original investment thesis? These are questions investment professionals already ask. The difference is that answering them often involves a considerable amount of manual work, searching across systems or speaking to colleagues who may or may not have been involved at the time.

AI changes that, but only if the underlying knowledge has been captured and organized in a way that makes those connections possible. That may not mean every piece of information needs to fit neatly into a rigid database, as quite simply, investment research is inherently unstructured as opinions evolve, discussions take unexpected directions, and the most valuable insights don’t always fit into predefined fields. The challenge is finding ways to bring structure to that information without getting in the way of how investment teams actually work.

That’s something we’ve spent a long time working on at Bipsync, and it’s central to how we think about the future of the platform. For me, AI should support investment judgment, not replace it.

But, there’s another important consideration here and that is that investment management isn’t simply an exercise in finding the right information, it’s about interpreting that information, challenging assumptions, weighing different perspectives, and ultimately exercising judgment. I don’t see AI replacing that process, but what I do see today is an opportunity to remove much of the friction surrounding it. Less time searching for research or reconstructing what happened in previous meetings, and more time spent on analysis, discussion, and making informed investment decisions. Less duplication of work, and ultimately more time focused on what really matters.

Of course, for institutional investors, this also needs to happen within an environment where security, permissions, governance, and accountability are taken seriously. An AI-generated answer is only useful if you can trust the information behind it, understand where it came from, and apply your own judgment to the result. That’s why I think it’s important to look beyond what an AI tool can demonstrate in isolation and consider how it fits into the broader research and investment process.

As an industry, we’re still learning how AI will reshape institutional investment research, the the technology itself is moving remarkably fast. We see that in our own work, the AI capabilities we’re building at Bipsync today are a leap ahead of what we were developing just a year ago, and I expect that pace to continue.  As the models become more capable and more widely available, they will also become more alike. The differentiator for investment firms won’t be which AI they use, it will be the quality of the data foundation the research is built on.

Firms that have invested in capturing their research, preserving investment context, and making collective knowledge accessible will be in a much stronger position to take advantage of what’s coming, while firms that haven’t will need to address those challenges, regardless of how sophisticated their AI tools become. For me, that’s the bigger opportunity.

We’ve spent the past 15 years helping investment organizations capture, structure, and leverage their research and institutional knowledge. AI gives us an entirely new way to unlock the value of that work, and I think we’re only beginning to see what’s possible. But one thing is clear to me, the investment firms that will get the most value from AI won’t necessarily be those adopting the latest technology first. They’ll be the ones that have invested in building a strong foundation of institutional knowledge and understand how to put that knowledge to work, because ultimately AI is only as powerful as the data foundation it’s built on.

Graeme Faulds is CEO of Bipsync, an investment research and knowledge management platform helping institutional investors capture, structure, and leverage their collective knowledge.