Lumonic's co-founder and CEO Kevin Hsu recently joined Daniel Liechtenstein (Co-founder & CEO, Hypercore) and Marc Andrew (Founder, The Private Markets Forum) for a fireside chat on building technology for the world's leading private credit funds.

You can catch the full conversation on Vimeo.

Catch some of Kevin's top insights on building and scaling private credit technology in the episode highlights below:

Episode Highlights

Marc:

Kevin, what does Lumonic do?

Kevin:

Our goal is to become the best data ingestion engine for private markets. We just happen to start with private credit. We started about three years ago, and now we're PitchBook's data ingestion product.

Where we usually come in is post-investment. We work with direct lenders, venture debt funds, and some asset-based lenders. They use us for everything post-investment—that starts at deal close, sometimes even gathering closing checklists, collecting all that data, structuring it in our system.

The mid-point usually for us is powering things like automated portfolio reviews and valuations. Then the end of the line is payoffs, LP reporting. As you prepare for a fundraise, you can use our data to benchmark against indexes. With PitchBook now, we have the LCD index we can pull into our system, so we can show fund performance metrics as well.

Marc:

When a manager first joins, what does their experience look like over the first few weeks and months?

Kevin:

If you're starting in spreadsheets, you have to think about database concepts for the first time. I always joke—this isn't something we say in our pitch—but in a lot of ways Hypercore and Lumonic are just a Postgres database, and then we built a bunch of very custom business logic on top of it.

We actually start with the end and work backwards. A lot of our early implementation conversations are really about: where do you want to end up? What are the automated alerts? What are the insights? What are the benchmarks you want to view once we're done?

Then from there we say, "Cool, if you need those outputs, then we need these inputs."

I think about the first six weeks of our implementation like moving into a new house. We get hundreds of thousands of documents dumped into our system, we take care of the rest. We're scanning all the credit agreements, spreading historical financial statements, any amendments. We're tucking that all into our system with a very clear goal on what outputs we eventually want to power.

One learning is that 80% of firms have pretty much the same data structures across our customers, and then there's a 20% that's customization. Like moving into a new house—everyone wants a different color on their walls, maybe a custom built-in somewhere.

Marc:

What does it take to achieve buy-in from the rest of the organization?

Kevin:

This is the crux of building a successful enterprise B2B product, but also a really strong account management strategy.

We have a dedicated implementation manager and a dedicated data services team for every one of our accounts. Our implementations can be as fast as 30 days, but they can also take as long as six months for larger asset managers.

It's kind of interesting—you're almost reselling the organization for people who weren't part of the sales process. A classic example for us is whoever is responsible for spreading financials, generating a portfolio monitoring template. That's usually where we start because we're solving a lot of problems for them right off the bat.

One thing we see in our analytics is three to six months into an implementation, all of a sudden the chief credit officer or the capital markets person is logging in. We've never met them, and they're pulling data out of our system proactively.

We have this saying: "Implementation is actually never done." Once you get into a firm, you can actually power so many different workflows.

Marc:

What are the core workflows people use Lumonic for?

Kevin:

You can almost think about it like Maslow's hierarchy.

The bottom of the pyramid is collection. In the first 30-45 days, we're basically trying to cut over the old way of doing things—collecting data over email, maybe getting it from a data room like Intralinks. We're pointing that toward the Lumonic way: borrower portals with automated notifications, dashboards where you can track statuses. We're giving customers a single pane of glass where they can see who's reported data and who hasn't.

The next phase is powering reports. Almost all of our customers have a portfolio review template or portfolio monitoring template—some monthly, some quarterly. 80% of our customers, this template looks the same: standard P&L, balance sheet, deal terms, cap structure, and commentary. Instead of spreading financials manually, collecting manually, pasting into Excel and hitting print—it just shows up in the dashboard.

A big thing for us, especially in middle market direct lending, is EBITDA adjustments. There's 100 flavors of EBITDA adjustments. Every firm has their own—their own EBITDA, the company's EBITDA, the sponsor's EBITDA.

The last workflow is orchestration or workflow automation. This is where I get most excited. I like to think about Lumonic like Salesforce—Salesforce is a system of record for your customers, modeled around the customer object. Lumonic is a source of truth for your assets.

For example, if you use Hypercore and Lumonic, there's a specific pricing grid update that needs to happen if a covenant gets tripped. We can set up a workflow: if a covenant is breached, notify the Hypercore API to update pricing dynamically.

The dream for Daniel and me—especially with AI workflows and systems becoming agentic—is that our customers don't even log in. They just get alerts on what they need to pay attention to.

Marc:

What performance benefits are your clients seeing?

Kevin:

I do this secret shopper thing where I reach out directly to someone who's been using our system for six months. I don't tell my team. I'm just trying to get a pulse on immediate benefits.

One thing I knew roughly from the sales process is that junior team members have a huge efficiency gain right off the bat. An analyst recently told me they used to spend 70% on portfolio monitoring, 30% on doing deals. You talk to them—they're two years out of college and they're like, "I want to work on deals."

We've actually inverted that for this specific analyst. Now they're spending 70% on deals, 30% on monitoring. We're taking six hours of work they would have done at the end of each quarter—pulling their hair out—and compressing it down to 30 minutes.

Marc:

How is Lumonic thinking about AI?

Kevin:

As a founder, I go day-to-day from being terrified to really excited.

I had the craziest conversation with one of our engineers last week. We were working on a bug and they said, "Man, I have to write code now." I'm like, "What do you mean?" They said, "Well, we use Claude Code, and it writes most of our code."

That was wild to me—if you're a junior or senior engineer now, you have this superpower of generating 10X the output.

For Lumonic, we use AI in very specific ways. We have a degree of separation from us and the large language models. We use AI in a way that's very deterministic—even if it hallucinates, we can make edits, we show its work clearly. It's not a black box.

You can use natural language to ask our agent, Lou, "Show me weighted average EBITDA margin across the portfolio." We'll generate a chart, show the source files it's pulling from.

Part of our thesis is that what's going to be defensible for companies like Hypercore and Lumonic is having the structured proprietary dataset. Structuring data is something we think AI won't be very good at.

Today when you use Claude or ChatGPT, there's no database. We've all had that interaction where you load in a bunch of data into a chat, and a week later you're like, "Where's that thing I made?" It's not backing up into a database.

At our core, we have this dataset we think is very important. We're structuring data in a central place, we've built workflows and UI to make it easy to structure, and then we've layered AI on top.

One thing Daniel and I are keeping a close eye on is what AI companies are doing within Excel. The bear case is if these horizontal AI companies can build agents that work inside Excel and understand the domain. There's a lot of products trying to do AI agents inside Excel—Ramp recently launched something. If you could tell a customer, "You don't have to leave Excel, the agents are just native inside"—I think we're a ways away from it, but that's something we watch.

Marc:

Are you seeing managers make use of AI themselves?

Kevin:

We see a ton of activity on pre-investment workflows using AI—screening, automated deal sourcing. The vision for a lot of these companies is: today you look at 1,000 deals a year, what if you could look at 10,000?

One thing that stood out: we came across one asset manager that paid seven figures for Palantir. Palantir coming into private markets is pretty interesting.

Almost all of our customers are signing up for some kind of enterprise LLM subscription. I always tell the senior folks: "Whether you like it or not, your junior team members are using AI. They're going to, so you might as well sign up for these enterprise subscriptions."

For anyone claiming AI can replace all these things, do a side-by-side comparison with something off the shelf. Pay for the expensive ChatGPT—$200 a month—run it side by side with the tool and compare results. You'll be surprised how powerful a lot of these models already are.

Marc:

What inning are we in for high-tech credit operations?

Kevin:

We're very early innings—second or third.

Three years ago, 80% of our pipeline was spreadsheets. It's probably down to 65-70% now. There's more competition, but we're very early. I'm still occasionally convincing a decision-maker that their analyst time is better spent somewhere other than spreading numbers.

This is why Daniel and I are so excited about this market. It's so clear that this was the last bastion because it was the most complex, most sophisticated asset class.

What does it look like in the future? I think there's going to be a ton of interesting fund structures and capital structures for liquidity in private credit, making it more tradable.

What we're really talking about now is the picks and shovels, the underlying plumbing. Today, a lot of people struggle with basic questions: what's the next interest payment, what's the current IRR on this deal? Daniel and I are solving those problems every single day. We're laying the groundwork.

Once you have that infrastructure, it's really interesting to think about what it enables. Liquidity is something I keep a really close eye on.

For the full conversation, watch the webinar recording.