By Roberto Pernicone
Private credit has become one of the most important financing channels outside traditional banking. Yet for most lenders, the real constraint is not borrower demand or available capital. It is the desk.
A loan file rarely moves cleanly. A borrower reaches out by email, text, broker referral, or intake form. From there, someone on your team has to interpret the request, apply your credit box, chase documents, coordinate third-party checks, review title, confirm insurance, validate entity information, handle exceptions, prepare terms, and get the file to the right decision-maker. Each step is manageable on its own. The problem is what happens when you try to grow. More files do not simply mean more underwriting decisions. They mean more handoffs, more follow-ups, more missing documents, more vendor coordination, and more places for something to fall through.
Why the Desk Became the Bottleneck
The default response to growth in private lending has always been hiring. When volume increases, you add processors, analysts, coordinators, ops staff. That helps in the short term but does not change how the work actually flows.
The desk stays the place where too much depends on a person knowing where things stand. Someone has to know which document is missing. Someone has to remember which exception applies. Someone has to check whether a file still fits your guidelines after something new comes in. Someone has to move data from one place to another.
That works when you are doing a manageable number of loans a month. It becomes expensive and fragile when you are trying to scale. The constraint is not your credit box or your capital. It is the number of touches required to get a file from inquiry to funded.
More Software Has Not Fixed It
Most private lenders are already running some combination of a loan origination system, CRM, document storage, spreadsheets, pricing tools, and vendor portals. The problem is that most of those tools organize work rather than do it.
A system can flag that a file is incomplete. It does not go get the missing document, verify it, update the file, recheck conditions, surface the exception, and log what happened. So the bottleneck stays. Someone still has to connect the dots between every tool, chase every piece, and keep the file moving by hand.
Private lending is also too specific for generic automation. Your credit box is not a simple checklist. It reflects borrower experience, leverage, property type, geography, liquidity, exceptions you have made before, and judgment you have built up over time. A tool that flattens all of that into a generic workflow does not solve the problem. It creates new ones.
Where Agentic AI Changes the Picture
The shift worth paying attention to is from systems that track work to systems that do it.
In an agentic setup, your credit box gets encoded once. Agents then handle the operational work around each file: sizing the deal, collecting and verifying documents, coordinating checks, preparing the package, flagging exceptions, and moving the file toward the decision that needs a human. The point is not to replace your judgment. It is to stop asking your team to spend their time on work that does not require it.
That is the real opportunity for private credit. Not a chatbot. An operating layer that knows your guidelines, works the file, and hands you something ready to approve or decline.
The Decision Still Belongs to You
The strongest argument for AI in private credit is not that it removes human judgment. It is that it gets your team to the moments that actually require it, faster, with a cleaner file in front of them.
Private credit works because lenders can evaluate situations that do not fit a template. That discretion is a feature, not a problem to be automated away. The right architecture keeps humans at every gate that matters and takes the coordination work off their plate.
Faster files are only useful if they hold up. A lender that moves quickly but cannot show how a decision was made has not improved anything. Speed and auditability have to travel together.
Antal as One Example of Where This Is Heading
Antal is building an AI operating layer for private credit that lets lenders encode their credit box, run files through agents, and keep approval gates in place for every decision that counts. The thesis is operator-led: the founding team built origination infrastructure firsthand and experienced the desk bottleneck directly before deciding to build something that removes it.
The broader point extends beyond any one company. Private credit is entering a phase where operational capacity matters as much as capital capacity. The lenders positioned to scale are not necessarily the ones with the most money to deploy. They are the ones with infrastructure that lets them deploy it consistently, quickly, and with a clear record behind every decision.
What Comes Next
Private credit’s growth has put real pressure on the desk model. More capital means more origination pressure. More origination means more operational complexity. At some point the people and the process cannot keep up with the volume.
That does not mean lending gets automated. It means the coordination work around lending becomes something a system can handle, while the credit call stays where it belongs. For lenders trying to grow without just adding headcount, that distinction is the whole game.
