GPs aren't behind on AI tools; they're behind on AI thinking
Private markets firms focus on AI tools. The harder problem is the operating model. Juniper Square's Brandon Rembe on what rebuilding around AI looks like.
Author: Brandon Rembe, Chief Solutions Officer at Juniper Square.
The firms winning on AI in private markets share one characteristic: they rebuilt their operating model around what the technology enables, rather than layering it onto what already existed.
Brandon Rembe, Chief Solutions Officer at Juniper Square, on how that rebuild actually works.
Every private market leader I talk to has the same question about AI: "Are we behind?" And my answer is always the same: yes, everyone is. But the more urgent question is whether you're behind in the right way.
There's a common misconception that AI adoption in the private markets is primarily a technology procurement problem—select a tool, deploy a model, check a box. It’s actually a foundational problem: most companies try to retrofit electricity into a steam-engine plant when they should tear the building down and start over.
That analogy isn't hyperbole. The early industrialists who adopted electricity spent years trying to wire it into their existing factories before a handful of engineers, Henry Ford among them, recognized that the technology required a completely different architecture.
Too often, the conversation about AI centers on tools: which large language model to use, what to do with the data warehouse, which vendor to pilot. These are the wrong first questions.
Pulling ahead requires asking a different question: what does our operating model look like if we design it from the ground up around these new capabilities, rather than grafting them onto workflows that predate them?
The companies that figure this out in the next 12 to 18 months will define the competitive landscape in the private markets for the next decade.
What agentic means in practice
Most organizations have interacted with generative AI in some form—drafting emails, summarizing documents, running a Q&A against a data set. That is not agentic AI. Agentic AI can make decisions, take action, and execute multi-step tasks with minimal human intervention.
And it changes the math on what's possible for a private markets firm. Say a $10 billion private credit firm currently runs a war room at quarter-end. Their team spends hundreds of hours a week reviewing PCAPs and quarter-end workbook papers, verifying data, ticking and tying numbers across documents. The bottleneck is entirely structural: humans doing pattern-matching work at human speed.
An AI agent can ingest the same materials, run hundreds of checks, flag anomalies, and show exactly how it reached each conclusion. What used to require a war room now requires minimal human oversight. That’s not incremental efficiency. It’s a different operating model.
AI should be used to outsource outcomes rather than outputs. An output is a deliverable—a document, a report, a summary. An outcome is a result—quarter-end closed accurately, LP request resolved, fundraise faster with fewer resources. AI agents are increasingly capable of delivering outcomes, and that changes the staffing math for GPs.
The broader enterprise market is already moving in this direction at speed. Gartner projects that by the end of 2026, 40% of enterprise applications will include AI agents—up from less than 5% in 2025. Deloitte estimates that 50% of enterprises using generative AI will deploy autonomous agents by 2027, doubling from 25% today. GPs that treat agentic AI as a 2028 problem are calibrating against a market already in motion.
Today, a small, well-structured team, augmented by agents, can manage materially more capital, $10 to $20 billion, with greater speed and accuracy than a much larger, manual one.
The data foundation question—and why most firms are asking it wrong
One of the most common objections I hear is: "Our data isn't ready for AI." This is partly true and mostly a deflection. Data hygiene matters. Connected systems matter. A single source of truth across loans, borrowers, and exposures is a genuine prerequisite for meaningful AI outcomes.
If you try to layer agentic AI on top of a fragmented data architecture, you'll get a very confident system that produces inaccurate results, which is materially worse than the status quo. As a recent analysis of private markets AI adoption put it bluntly: AI can be a multiplier of operational weaknesses and risk exposure if implemented on an inadequate, old data infrastructure. The firms hitting walls on their AI initiatives right now aren't dealing with a model problem. They're dealing with a plumbing problem.
But "our data isn't ready" often becomes a reason to wait, when the actual answer is to start building the foundation and deploying AI simultaneously in workflows where data is already clean and structured.
The private markets—and private credit in particular—present an especially acute version of this challenge. No two deals look the same. Every borrower's story evolves. The documents behind those deals are often non-standardized. Critical information sits in PDFs, email threads, and Excel trackers that don't speak to each other. Transparency has historically been constrained by data latency and manual reporting: by the time a covenant breach or delinquency surfaced in a quarterly report, the warning signs had been present for weeks.
This is precisely where AI has the most to offer. It thrives in document-heavy, relationship-driven environments where the underlying information is complex but the tasks are structured: extract this term, flag this anomaly, standardize these inputs so a model can compare them across a portfolio. AI doesn't replace judgment—it gives teams better information, faster, with fewer gaps.
The latency between an event and the people who need to act on it can shrink to near zero when the data infrastructure and AI systems are genuinely connected.
As research on private credit and AI has found, agentic AI can act on approximately 60% of incoming LP requests without human intervention. For the remaining 40%, it's about making the human handoff faster and better informed. That ratio will improve as models improve and context deepens.
Trust but verify: the operating model that actually works
The most common failure mode I see isn't firms doing too much with AI—it's firms doing too little because they can't figure out how to trust it. And that's actually the right instinct applied in the wrong direction.
Ronald Reagan's phrase from the nuclear disarmament negotiations—"trust but verify"—is the right framework for agentic AI. You trust the agent to do the work. You verify through accountability and accessibility. The agent's reasoning must be legible and show its work. If a person cannot understand how an AI model reached a conclusion, that model will not influence real decisions, regardless of how accurate it is.
This is why I'm skeptical of AI implementations that operate as black boxes. The agent that's earning trust in a $10 billion firm isn't just getting the right answer—it's showing the data points it pulled, the checks it ran, even the code it used to execute those checks. After two or three quarters of 100% accuracy on that process, the team stops double-checking it at the granular level and starts checking it at the oversight level. That's the transition from output management to outcome management.
Think about the visibility consumers expect from something as simple as a pizza delivery. We’ve normalized real-time tracking for low-stakes decisions, while critical financial processes still rely on delayed reporting. That gap will close. Firms that build for transparency and auditability now will have a meaningful advantage as expectations evolve.
In conclusion
The firms that win won’t be the ones with the most AI tools. They’ll be the ones that rebuild their operating model around what AI actually enables: scalable expertise, continuous monitoring, and outcome-level accountability.
The technology is ready. The constraints now are mindset and execution speed. Waiting for perfect conditions isn’t risk management, it’s falling behind.
About the author
Brandon Rembe is Chief Solutions Officer at Juniper Square, where he works with private markets GPs to design and implement AI-driven operating models across fund operations, investor relations, and portfolio management. With 25 years of experience building financial technology, he leads Juniper Square's efforts to help clients translate AI capability into measurable operational outcomes.
Editor's note
Every piece published on The Bright Minded goes through careful verification, but mistakes can happen. If you spot an error, have additional information, or want to flag anything, write to rosalia@thebrightminded.com.
