CHICAGO — (December 18, 2025)
Authored by Corey Coscioni, Co-founder Lamina
This post is the first in a two-part series reflecting on last week's CLO Summit panel discussion and the evolving role of AI in the CLO Market.
PART 1: Trust, Explainability, and Guardrails in a Regulated CLO Environment
Last week, I had the opportunity to join a strong group of industry leaders at OPAL's CLO Summit in Dana Point, California for a panel discussion: How AI and Machine Learning are Transforming the CLO Market. The conversation was practical, candid and grounded in the realities of operating within a highly regulated financial environment.
The panel was moderated by Steve Miller of CreditSights, with contributions from Aani Nerlekar (SS&C Advent), Eugene Grinberg (solve), Joe Bao (Proskauer), John Borse (Octus), Joseph Burdis (AIR Platforms), TJ Unterbrink (barrow Hanley) and myself, representing Lamina. What stood out most was how aligned the group was on a core principle: AI is already delivering value in CLOs - but only when it's implemented with discipline, transparency, and the right guardrails.
1. Start Deterministic
One of the most important themes we discussed was the need for deterministic and explainable AI - particularly in CLOs.
Not all AI needs to be probabilistic or opaque. In fact, many of the highest-value use cases today rely on supervised, explainable models - especially when outputs will ultimately be reviewed by auditors, compliance teams, and regulators.
From my perspective, if you can't explain what the system did and why it did it, you're creating risk instead of reducing it. Auditors are going to ask what happened. Regulators are going to ask how you know it's correct. AI needs to support those conversations, not complicate them.
That's why guardrails matter. CLOs operate within a heavily regulated environment, and AI should respect that reality from day one.
2. AI Should Accelerate Work, Not Replace Verification
Another point I strongly agree with is that AI should augment research and operations, not replace human judgment.
Whether you're using AI to summarize documents, extract data, or surface insights, the system should always be able to:
- Show the original source
- Point back to where the information came from
- Allow users to quickly verify outputs
If an AI tool can't tell you where the data originated, that's a problem - especially in a market where auditability and traceability are non-negotiable.
3. Efficiency vs. Insight: Both Matter
The panel also spent time discussing the difference between efficiency-driven AI and insight-driven AI.
Efficiency use cases are where we're seeing immediate ROI today:
- Faster document processing
- Reduced manual workflows
- Chat-style interfaces that surface answers quickly.
These improvements may not move markets on their own, but they meaningfully reduce operational friction and risk.
Insight-driven AI is where things get more complex. These models begin connecting data across private credit and broadly syndicated loans, continuously ingesting new datasets and identifying patterns that humans would struggle to spot at scale.
But none of that works unless the underlying data is clean, structured, and reliable. So make sure you have a robust Data Management program in place.
4. The Reality of CLO Data
One challenge we all acknowledged is how fragmented and unstructured CLO data still is.
Agent bank notices, amendments, rate changes, and borrower communications arrive in countless formats. In some cases, firms still rely on outdated delivery methods because they're perceived as secure.
For AI to work in this environment, it must:
- Securely ingest unstructured and structured data
- Normalize and validate information across sources
- Respect strict privacy and compliance requirements
Without this foundation, even the most advanced AI models will struggle to deliver meaningful value.
In Part 2, I'll look ahead to how AI is enabling a shift toward real-time credit infrastructure and how these ideas are being applied in production today. Make sure sure to follow Lamina on LinkedIn to catch the next post.