The "debt" metaphor (borrowed from software's "technical debt") is doing something clever here. In software, technical debt means: this system works, but it's built on accumulated shortcuts, legacy decisions, and path dependencies that make it expensive to change. The system functions precisely because of the debt — it's load-bearing cruft.

You're applying that to human roles: certain jobs exist because the institutions they serve have accumulated decades of relational and procedural complexity that can't be refactored. The person isn't valuable because they're smart or even because their judgment is irreplaceable — they're valuable because they are the living index of a web of relationships, precedents, implicit agreements, and institutional memory that was never written down and never could be.

The key examples you listed reveal the structure:

Healthcare is the purest case. A senior physician's value isn't primarily diagnostic (AI will match or exceed that). It's that they know which specialist actually picks up the phone at 2am, which insurance reviewer can be reasoned with, which nurse's instinct to trust over the chart, which patient's "I'm fine" means they're not fine based on fifteen years of knowing their family. The entire system runs on informal trust networks layered over formal protocols, and the formal protocols themselves encode decades of regulatory negotiation, malpractice case law, and institutional memory. You can't just swap in a new system — the "debt" is structural.

Corporate law — the value of a senior partner at a white-shoe firm isn't legal reasoning (LLMs can already do most of that). It's that they know the general counsel at the counterparty personally, they've done three deals together, they know that this particular GC panics at aggressive language but responds to quiet pressure, and they have implicit credibility built over twenty years that lets them say "this clause is fine" and have it believed without a six-week review cycle. The relationship is the infrastructure.

Dealmakers and political operators — same pattern. The value is being a trusted node in a network where trust was built slowly and can't be transferred. A fixer in DC works because they've done favors for decades, they know who owes whom, they understand the unwritten rules of which committees actually matter, and their phone calls get returned. This is, in a sense, a human API that was built over thirty years and has no documentation.

What makes the "debt" framing specifically insightful rather than just saying "relationships matter" is that it implies several things:

First, it's not an asset — it's a liability that happens to be load-bearing. Just as technical debt isn't good code, relational/institutional debt isn't a well-designed system. Healthcare shouldn't require a specific doctor to know which administrator to call to get a patient into a bed. The fact that it does is a failure of system design. But you can't just rip it out, because the system would collapse.

Second, it creates a moat against disruption that is simultaneously real and pathological. These roles survive not because they represent the optimal way to do things, but because the cost of refactoring the entire institutional web they're embedded in is too high. An AI could theoretically replace the cognitive part of what a dealmaker does — but it can't replace the thirty years of accumulated trust, favors, and implicit agreements that the dealmaker carries in their head.

Third, it eventually gets paid down. This is the most important implication and the scariest one for people in these roles. Technical debt gets resolved when you finally do the rewrite, or when the whole system gets replaced. Institutional debt gets resolved when the institution itself gets restructured — or when enough of the relationship holders retire/die that the system is forced to rebuild on new foundations. AI might accelerate this by making it cheaper to build new institutional infrastructure from scratch rather than maintaining the old one.

The people who are most AI-proof right now under this framing are those whose roles sit at the intersection of: long time horizons for trust-building, high-stakes decisions where personal credibility matters, regulatory complexity that encodes decades of negotiation, and informal knowledge networks that were never formalized. But the framing also implies these roles are living on borrowed time — they're surviving on debt, not on intrinsic value, and debt eventually gets restructured.