Hum-an Stories: Andrew Eisen on AI, Private Credit, and the Transformation of Private Markets
In this Hum-an Story, we sit down with Hum Capital CEO Andrew Eisen to explore the collision of two massive boom cycles — AI and Private Markets — and the profound implications for lenders, investors, and the future of underwriting. In a candid conversation, Andrew breaks down the risks of untested AI tools, the lessons from past financial crises, and Hum’s differentiated “atomic data” strategy that positions the company at the frontier of financial technology.
Q: You mentioned that Hum sits at the intersection of two historic boom cycles — AI and Private Markets. Why is that convergence so powerful?
Andrew: Very few times in history do you get two amplification waves cresting at the same moment.
Right now we’re watching the AI boom cycle collide with the Private Markets boom cycle — and the amplification effect of those two currents coming together is enormous.
For Hum, it means we’re not just applying AI in a vacuum. We’re bringing artificial intelligence directly into one of the largest, least technologically modernized sectors of the investment management ecosystem: private credit.
Most people underestimate the complexity of the data requirements behind AI. They don’t appreciate the biases, associations, and context needed to get outputs you can trust. But in private markets, that context is everything.
Q: How does Hum’s approach to AI differ from the rest of the market?
Andrew: A lot of people talk about “AI in private credit” at a conceptual level. But very few understand that the only sustainable advantage comes from the atomic unit of data — the raw, immutable records from operating systems such as accounting, banking, and payment processing systems.
That’s the foundation of Hum’s strategy.
- Not scraping PDFs.
- Not gridding out excel sheets.
- Not building on incomplete summaries or approximations.
If you want a smarter machine with lower error rates, you go to the atomic layer first. That’s where truth lives. You build up from systems of record, not down from documents.
Our view is that the future winners in this industry will be the platforms that aggregate, normalize, and understand atomic data at scale — because that’s what makes machines reliable, dynamic, and trustworthy.
Q: Many lenders and investors are exploring AI tools but struggling to separate signals from noise. How should they think about risk?
Andrew: Every time there’s a technological leap — Web, Cloud, AI — you get a “wild west” moment. Hundreds of vendors emerge, promising huge operational savings, magical insights, dramatic returns.
But the reality is: Most of these tools have not been battle-tested.
They don’t have long operating histories. They haven’t survived market shocks. And because of that, investors don’t yet understand the real execution risk.
Financial history repeats this pattern:
- LTCM in the ’90s with programmatic trading.
- Lehman Brothers a decade later.
Models that assumed correlations would hold… until they didn’t. When machines hit conditions they’ve never seen before, they can — and do — choke. Big guesses, bad assumptions, catastrophic blind spots.
So lenders should ask:
- How many adverse environments has this tool survived?
- What happens when the world changes?
- What biases are baked into the models?
The companies that treat AI as a silver bullet will get burned. The companies that treat AI as a tool — with human oversight, guardrails, and a deep understanding of the data — will thrive.
Q: You talked about the inherent biases in both humans and machines. How should the industry think about AI bias, specifically?
Andrew: Every AI model is trained on history. And history is full of biases — success bias, selection bias, human decision bias.
For example, stock indices look like they always go up over the long term, because companies that fail get removed. If an LLM is trained on Reddit or other internet forums, you inherit those cultural biases too.
Machines don’t escape this; they amplify it. The real question for investors isn’t, “Does this tool have bias?” It’s: “Do you understand the bias well enough to manage the risk?”
You need transparency. You need data governance. And you need systems that contextualize information rather than hallucinate.
That’s why Hum invests so heavily in atomic data — it’s objective, grounded, and reduces the probability of distortion.
Q: You shared a framework of three pillars behind Hum’s product strategy. Can you walk us through it?
Andrew: Absolutely — the pillars are:
1. Atomic-Level Data Collection
Pull data directly from systems of record — the stuff that can’t lie or that is hard to fake at scale. Bank transactions. Accounting entries. Credit Card swipes. This is your machine’s foundation.
2. Human-Centric Interfaces Powered by AI
You still need thoughtful, intuitive ways for humans to interact with that data — prompts, dashboards, queries, and AI agents that accelerate understanding.
Whether you call it “prompt engineering” or just building a great UI, the goal is the same: Make complex data immediately actionable.
3. Continuous Observation & Market Awareness
Most decisions today are made at a single moment in time. But markets move. Risk shifts. Information updates.
A future-proof platform continuously observes atomic data and flags changes that matter. That’s how you build a capital-raising system that aligns every party — borrower, investor, and intermediary — around truth.
These pillars work together to create a modern, resilient infrastructure for private credit.
Q: With AI evolving so quickly, what advice would you give to investors who are just starting to integrate these tools?
Andrew: Start small, start safely, and understand the why before the what.
Ask vendors:
- How do you handle context?
- What is the source of your data?
- Do your models survive outlier scenarios?
- What biases are embedded in your outputs?
And most importantly: Keep humans in the loop.
Machines are incredible at filtering noise, but they’re not infallible. They need oversight — and they need data that actually reflects reality, not just history or assumptions.
Q: Finally — where does this all lead? What’s the long-term vision you see for AI in Private Markets?
Andrew: Over the next decade, private credit will shift from a document-based industry to a data-based one.
AI won’t replace judgment — but it will replace inefficiency. The winners will be the platforms that:
- Understand the atomic layer of data,
- Contextualize it for humans,
- And continuously observe change.
That’s where Hum is heading. We’re building the infrastructure for a future where better data, smarter tools, and transparent workflows elevate everyone in the ecosystem.
