Why generative AI isn't uncovering alpha yet

In the world of hedge funds, everyone's talking about AI. But recently, Ken Griffin, one of the industry's most respected voices, made a bold claim: generative AI isn't delivering alpha for hedge funds. Instead, it's mainly boosting workflows and productivity.

He's not entirely wrong but he's not telling the whole story either.

Here's the thing: the problem isn't that AI lacks the capability to find alpha. The problem is that most firms aren't giving AI the environment it needs to succeed.

Breaking down Griffin's argument

Let's start by acknowledging Griffin's point. He's correct that a lot of current AI use cases are indeed improving workflows rather than generating new market-beating insights.

Many firms use AI to automate research summaries, streamline reporting, or speed up documentation reviews. These are valuable applications, no question. But they're fundamentally different from uncovering alpha.

Griffin's observation reflects a real pattern in how firms are deploying generative AI today. They're thinking about AI first, then providing it with just enough data to complete a specific workflow. It's a narrow, task-focused approach.

And within those constraints, AI is delivering exactly what you'd expect: incremental productivity gains.

So Griffin isn't wrong. He's just incomplete.

The real root cause is…

Lack of integrated workspaces.

Here's where the narrative shifts: the reason AI isn't finding alpha isn't because AI can't find alpha.

It's because most firms aren't leaning in enough to give AI and the user what they actually need, a unified, collaborative workspace where all their data is fully accessible and integrated.

Think about where alpha traditionally comes from in hedge funds.

It emerges at the intersection of insights that no one else has connected yet. It lives in the correlations between datasets that were previously siloed. A trader might notice a pattern across private equity, public equities, and macro data that nobody has synthesized before. That's where edges come from.

That discovery process isn't meant to be a machine doing it alone in a black box. It's meant to be a conversation. An AI system that can surface patterns and ask "what if we looked at these datasets together?" while a human brings domain expertise, intuition, and judgment to interpret what those patterns actually mean for the investment thesis.

Most firms today don't have that dynamic. They have fragmented data silos, which means the human and AI can't collaborate effectively. The data lives in separate platforms, the insights stay isolated, and neither the AI nor the human can see the full picture.

It's not a collaboration at all, it's a constraint.

Now imagine if you could give an AI system access to all of that data simultaneously, in one integrated environment. Not just for the AI to analyze in isolation, but as a true workspace where the AI and human can work together.

The AI could surface patterns across asset classes, teams, and time periods that humans might miss. It could ask questions like "have we ever seen these two metrics move together?" or "what happens when these conditions align?" And then a skilled analyst or trader could jump in with context, ask follow-up questions, and develop those insights into real alpha.

That's the collaboration model that actually works.

It's not about replacing humans with AI. It's about giving both the human and the AI the environment they need to think together.

I did a small POC at the start of the year of where we are heading, and it’s so clear to me that this is the future - watch this 50 second clip.

How a unified approach unlocks alpha at the intersection

When all of a firm's data lives in one place, truly integrated and accessible, something powerful happens.

The AI and human can actually work together to surface patterns that would never emerge from siloed environments.

Consider a concrete example: a firm brings together public market data, private market data, macroeconomic indicators, and alternative data sources all in one workspace. An analyst poses a question: "What correlations emerge when we combine private equity valuations with macro conditions and public market movements?" The AI agent analyzes this unified dataset and surfaces patterns. Maybe it discovers that certain private equity valuations tend to spike precisely when specific macro conditions align with particular public market movements. That insight might not be obvious if you're looking at those datasets separately. But at their intersection? That's alpha waiting to be explored.

The same principle applies across different teams and geographies. When portfolio managers, research analysts, and risk analysts can all feed their data into a shared workspace, the AI becomes a tool for collaborative discovery. It surfaces opportunities that emerge from combinations of perspectives that were previously isolated from each other.

This is why building a unified workspace where firms can bring their data in from different asset classes, different teams, and different geographies matters so much. It's not just about convenience or consolidation. It's about creating the conditions where humans and AI can collaborate effectively to discover alpha in the intersections where it actually lives.

And here's something worth keeping in mind: the models are only getting better. They're getting smarter, handling more data, parsing more complexity. The ability to digest and understand massive, unified datasets is not a limitation of future AI, it's just a matter of time.

The firms that build these integrated workspaces today won't just get the competitive advantage from the tools available right now. They'll be perfectly positioned to take advantage of the next generation of models that can do even more with all that data.

Addressing the real-world complexities

Although I’m biased, I want to be honest. A unified workspace is a powerful enabler, but it's not a magic wand. Finding alpha still requires the right strategies, the right people, and a willingness to experiment with new combinations of data.

There's also the matter of integration itself. Bringing all these datasets into one workspace isn't trivial. It requires effort, investment, and a willingness to rethink how data flows through the organization. For many firms, that's a significant lift.

But here's the thing: the long-term payoff is worth it. By bringing disparate datasets together and allowing AI to work alongside your team, firms can unlock insights that were previously hidden.

The firms that make this investment today, that build truly unified workspaces, will be the ones discovering alpha tomorrow while others are still wrestling with siloed data infrastructure.

The path forward

The future of AI-driven alpha doesn't lie in deploying more AI to individual workflows. It lies in breaking down data silos and building integrated environments where AI can discover patterns at the intersection of datasets.

A unified workspace is a critical piece of that puzzle, but it's part of a broader toolkit that also involves having the right people, the right strategies, and the right culture of experimentation. It requires firms to rethink their data strategies from the ground up but that shift is essential.

The invitation is clear: rethink how your data is organized.

Embrace integrated approaches.

Consider how a unified workspace can transform AI from a productivity tool into a true source of competitive advantage.

Truly imagine it.

The insights, and the alpha, are waiting at the intersections.

Appendix

Alex Izydorczyk has a good post on the topic as well, that is a good read: https://magis.substack.com/p/skepticism-early-trends-and-an-early.

(Alex is also compiling a list of all AI startups in the financial space, which you can check out here)

If I were to segment companies, I would probably segment them as:

1. Data first companies - Data Extraction Tools and AI Model Providers (Alpha, Quant, & Forecasting Labs). Selling signals, or derivative is still selling data, it's just - allegedly - more valuable data because it has been processed in a certain way that makes it more valuable. Competing with Bloomberg, S&P, FactSet, ...

2. AI first companies - Research Copilots (AI “Analyst” Assistants) and Excel Copilots (Financial Modeling Aides). They start with AI first and then try to find data relevant to their customers. Competing with OpenAI, Anthropic, ...

3. Infrastructure first companies - "Terminal 2.0" Platforms (Next-Gen Market Terminals). I think examples of the competition is Claude for Financial Services with their MCPs integrations or Blackrock's Aladdin with their new copilot. However, I think that this will always be limiting to create Alpha because if everyone has access to same data and intelligence, then it's hard to find alpha.

How it started

How it’s going

The future

"If a strategy with a proven positive alpha becomes public information, its profitability is eroded through market efficiency."

Then, Data + Intelligence needs to happen within organization. And the foundational infra of each of these firms becomes the most important piece.