What Private Credit strategy is the best according to AI ? The results are in! A Highly Scientific Experiment in Decision-Making
Adam Grimsley
Working with family offices and pension schemes to access the opportunity set in private and alternative markets.
Humans, that’s me, have long been celebrated, remarkably again even including myself occasionally, for their original thinking and hard work, well ..let’s move on, yet perhaps even more remarkable is our talent for copying, no comment about myself and GCSE economics homework, absorbing, adapting, and reapplying ideas to individual bespoke situations. This propensity for imitation, excluding GCSE plagiarism, allows us to rapidly learn from successes and mistakes alike, effectively accelerating innovation by building on what works and what doesn’t … well at least that’s my theory.?
In a similar vein, AI platforms like ChatGPT extend this natural human strength by aggregating and synthesizing vast amounts of information, enabling us to quickly replicate and refine the best ideas. Essentially, while the spark of originality ignites progress, it is our capacity to copy and improve upon existing ideas that propels our society forward at pace.
So… now I’ve fully provided a water-tight case for plagiarism as a moral necessity for humanity, let’s put it into practice!?
In my research lab of private credit investments, e.g my back bedroom in South London, I conducted an experiment: I fed several advanced AI systems—Claude, Gemini, Co-Pilot, ChatGPT, and DeepSeek—the same challenge. The task? Recommend the top three strategies for a professional, sophisticated, diversified investor with ample liquidity.?
And the winners are …..
The results converged on three core strategies:
1st place -? Direct Lending
2nd place - Mezzanine Financing / Distressed Debt??
3rd place -? Speciality Finance / Asset-Based Lending / Structured Credit
Individual model results 1st to 3rd
1st Direct Lending: The Baseline Control
Our AI models unanimously identified direct lending as the favourite in the private credit equation.?
> "Direct lending typically involves senior, secured loans to mid-market companies, offering favorable risk‐adjusted returns due to collateralization and robust covenant protections."??
> (Claude)
In the controlled environment of our experiment, direct lending is our baseline and it seems AI agrees.?
2nd Mezzanine Financing / Distressed Debt: The Reactive Compounds
Here the AI responses begin to diverge like the outcomes of a bifurcated experiment. Gemini and Co-Pilot lean toward mezzanine financing, noting:
> "It is a step up in risk from direct lending, as mezzanine debt is subordinated. However, the potential for equity upside through warrants can be attractive to experienced investors. Therefore, it is placed second, as it is a step up in risk."??
> (Gemini)
In contrast, Claude and ChatGPT focus on distressed debt:
> "A well-executed distressed strategy can generate outsized returns compared to traditional credit, especially when managed by experienced teams.It’s a natural complement to direct lending—enhancing yield while still offering some downside protection."
> (ChatGPT)
3rd Specialty Finance / Structured Credit: The Niche Elements
The third category, variously termed Specialty Finance, Asset-Based Lending, or Structured Credit, represents the niche elements discovered on the periodic table of private credit. As one AI noted:
> "Asset-based lending is listed third because, while it offers attractive risk-adjusted returns, it is more specialized and may require a deeper understanding of the underlying assets being used as collateral."
> (Co-Pilot)
For investors, this is where qualitative analysis meets quantitative data. It’s not merely about crunching numbers but about understanding the underlying asset quality, market dynamics, and the strategic fit of these niche investments in a broader portfolio.?
The Synthesis
Actually the results were pretty much as I expected, although one can argue the specifics of 2 and 3. The reasons behind the AI’s suggestions are also kinda reasonable and reflect an inherent bias toward well-documented, historical strategies that have been validated over time. No surprise there.
The order begins with direct lending as the starting point, “a solid foundation in private credit, delivering stable returns without excessive risk” - seems sensible. Distressed and mezzanine being complementary to direct lending “enhancing yield while still offering some downside protections” maybe, not so sure distressed is the ideal downside protection but…lastly speciality finance while “potentially lucrative, being niche, higher-risk allocation that requires deep expertise and careful manager selection.”?
Of course there are going to be limitations and biases. For example it is limited by the availability, or lack of, of public data, the results therefore mirror trends in institutional investor reports (e.g., S&P, KBRA, Preqin, McKinsey etc ), suggesting AIs are trained on widely available market analyses, I’m not blinding you with science here! Plus given its general public usage the AIs avoid highly speculative strategies, venture debt etc, so we have some conservative bias.
It’s obvious AI is becoming an increasingly valuable tool for LPs that can assist with data analysis, trend identification, and even risk modeling. However, their recommendations are frameworks, not substitutes for expert human judgment, a fine observation I might add, at this point. The variations in third-strategy picks highlight that optimal allocations depend on investor-specific goals and risk appetite. LPs should use AI insights to narrow focus, then validate with deep due diligence on managers, sectors, and terms.
While AI might help us identify the “what,” it’s the LP who must answer the “why?” In the world of private credit, the smart investor blends the empirical power of AI with their own individual requirements..?