Understanding is Harder than Building
Many of our clients come to us because something in their existing setup isn’t working anymore. Sometimes it’s a third-party system that no longer fits their growing needs. Other times it’s an internal platform that became too messy to evolve with the firm.
Either way, the story is the same: the current state can’t scale.
The Old Playbook
Before AI redefined technical innovation, most hedge funds tackled this problem with the following strategy:
Gauge what part of the process was most broken (ie “We need to do a better job managing risk” or “We need a better system to collate and leverage our research”)
Find an off-the-shelf SaaS solution that better equips your team to solve that problem
Spend months onboarding the product and teaching your team to use it
Repeat the process with the next problem
And historically, that strategy worked fine. What mattered most about technology solutions was the features it provided. It didn’t matter if your risk management system was hosted in an internal database or some other company’s servers. All that mattered was that your risk manager could access the analytics she needed to do her job.
Why AI Changes Everything
Generative AI changes the equation because for the first time, where your data lives and connects is important. If you wanted to create an AI agent for your risk manager, it definitely needs a way to access data from your risk management system. But it also needs to join that data with real-time trades, incoming news, internal research, etc to generate well-rounded insights.
That’s why a scalable AI strategy starts with a scalable data strategy. Firms that build on that foundation will adopt AI faster — and better — than everyone else.
Enter the Data Audit
The first thing we do with a new client is an extensive audit where we dive into
What is in the databases
Where the data comes from
How it flows to reports and external vendors
We use proprietary AI tools to traverse the databases and produce an interactive knowledge graph. You can click into nodes and lines to learn more about a specific table or process.
At first, we considered the knowledge graph as a means to an end. Before we built “the real platform” in a client’s environment, we needed to understand what was already in there.
What we didn’t predict was just how much this knowledge graph unlocked on its own. Here are the kinds of anecdotes we hear most often:
Over the past decade, CTOs and engineers have come and gone, each wave having built something different. Nobody at the firm understands all the pieces, so it’s been impossible to strategize an improvement plan. This knowledge graph is the first time we have the information we need to chart a course forward.
This knowledge graph helped us realize that different departments were using different analytics to make critical decisions. For example, traders are looking at fund PnL on Bloomberg, ops teams are referencing our fund admin, and portfolio managers are checking the OMS. Every source has its own nuances so everyone is making decisions off different numbers.
One of our biggest problems is retaining knowledge as people leave the firm. Even our most critical reports and workflows only have one or two SMEs. When someone leaves, we’re all left scrambling. Having all our data and reports documented in a clear knowledge graph helps us hedge against key person risk and ensures operations stay resilent.
After we finish the data audit and knowledge graph, we propose a roadmap for how we would deploy our software to solve their problems. And at this point, clients can choose one of the following:
Move forward with a longer-term engagement with us and we implement the plan, or
Keep the knowledge graph, end our partnership, and chart next steps themselves.
So far, nobody has chosen (2). Which makes the knowledge graph the best upselling tool I’ve ever seen.
Not only has it been rewarding to see the positive reaction, but it’s been incredibly empowering to see just how fast our engineers can get these audits done with AI.
A couple years ago, it would have taken months or years. Now, we do it in weeks.
AI transformation doesn’t start with agents — it starts with understanding your data.
Once you can see how everything connects, building intelligent systems becomes not just possible, but inevitable.