The diagnosis is settled: AI does not create organizational debt, it calls it due. What remained to be written was the sequel: how you repay. This article ports the machinery of financial turnaround onto the organization: covenants instead of an inventory, three repayment regimes, a triage between avalanche and snowball, and the option no one dares take, restructuring. A case for keeping a register of organizational debt at the same rank as the risk register.
Your transformation is in default. The budget is approved, the licenses are deployed, the pilots are multiplying, and yet nothing scales. The reflex is to look for the cause in the technology, or in that rhetorical convenience we call resistance to change. It lies elsewhere: your organization is insolvent.
In these pages, we have documented this diagnosis from four angles. The shift toward AI-Native reveals an organizational debt that your tools were masking. Critical knowledge concentrated in a handful of heads is a complexity rent, a liability disguised as expertise. Legacy architecture levies a viscosity tax on every initiative. And the true debt of AI will not be technical but organizational. Those four articles are the bailiff's report. This one is the turnaround plan. Because diagnosing over-indebtedness has never paid down anyone's debt.
A listed company knows its financial debt to the cent. It publishes its ratios, negotiates its banking covenants, provisions for its litigation. That same company knows almost nothing of its operational liabilities: the monthly close that rests on one person and two spreadsheets, the client exceptions handled from memory, the architecture decisions whose rationale left with their author, the business rules that exist nowhere but in the practice of those who apply them.
None of these liabilities appears in any reporting. Yet they display the three attributes of a debt. A deferred cost: yesterday the organization saved itself the effort of formalizing, documenting, transmitting. Compound interest: each passing year buries the knowledge a little deeper and raises the cost of extracting it. And a repayment still possible, at least while the knowledge-holders are still around.
What makes the situation urgent is that the creditor has just spoken up. As long as nothing changed, this liability slept and its interest rate stayed close to zero. The decision to deploy AI turns it into a debt called due, immediately and in full: you cannot augment a process that exists nowhere, nor hand an agent a rule no one can state.
Faced with this, the natural temptation is to map. Launch the great process inventory, mobilize consultants, produce the complete atlas of how the company actually works. This is a mistake, for two reasons. The exhaustive inventory is impossible: tacit knowledge escapes precisely the exercise that claims to capture it. And it is obsolete before it is finished: the organization keeps living while you photograph it.
Banks solved this problem long ago. When a creditor cannot audit its debtor continuously, it does not measure the stock of debt: it tests solvency. It imposes covenants, simple and verifiable ratios whose breach triggers an immediate alarm.
The organizational transposition holds in three clauses. Every critical process must be executable by at least two people. Every structuring decision must be recoverable, with its rationale, in under ten minutes. Every critical system must be redeployable without its historical maintainer. Three binary tests, verifiable in an hour, that do not claim to measure the debt but to detect insolvency. A broken covenant triggers an action, not a line in the annual audit. That is the strength of the device: it turns a diffuse worry ("we depend too much on certain people") into an actionable signal.
That leaves the question of why so many transformations stall even when everyone acknowledges their necessity. The usual answer invokes culture. The accounting answer is more precise: it is a debt overhang.
In finance, an over-indebted company can no longer invest, even in profitable projects, because every future gain is already pledged to debt service. No investor funds growth whose first beneficiaries would be the creditors. The organizational equivalent is recognizable by a symptom every executive has heard: "we don't have the bandwidth." When synchronization meetings, exception handling, and decision archaeology (recovering who decided what, and why) consume most of the capacity, nothing remains to invest in the transformation, whatever its expected return.
This is not a motivation problem. It is a state of operational insolvency, and it has a perverse feature: it is invisible in the accounts. An organization can be in operational bankruptcy while showing an impeccable financial balance sheet, because the liability that paralyzes it is recorded nowhere.
Once liabilities are detected (by their broken covenants, not by inventory), treating them is a matter of management, not heroism. For each liability, three regimes are possible, and the choice must be explicit.
Repay the principal: externalize the knowledge, document, automate, transfer. It is the most expensive regime, to be reserved for the liabilities carrying the highest interest rate, the ones the whole organization crosses every week.
Service the interest: contain without reducing. Double up the knowledge-holders, freeze the perimeter of the system in question, forbid adding new complexity to it. The debt does not shrink, but it stops growing, and that is sometimes exactly what is needed when repayment capacity is committed elsewhere.
Let it ride, with a review date: the process due to disappear with the planned migration deserves not a cent of repayment, on one non-negotiable condition: that the migration has a date, and that the decision is reexamined if that date slips.
The management failure is none of these three regimes. The management failure is the fourth, the one no one chose: the default regime, where the debt grows in silence because no decision was ever made.
How to prioritize repayments? Personal finance settled this debate long ago. The avalanche method repays the highest-rate debt first: it is mathematically optimal. The snowball method clears the small debts first: it is suboptimal on paper and formidable in practice, because each cleared debt frees up capacity and sustains momentum.
Ported to the organization, the avalanche points to the liabilities everyone crosses: onboarding, releasing to production, the close. The snowball points to the visible, bounded irritants a team can clear in a few weeks. And in an organization the snowball has a virtue finance ignores: it recruits allies. Each small debt cleared publicly creates a precedent, proof that repayment is possible, and volunteers for the next one. Organizational deleveraging is a political undertaking as much as an accounting one, and the plans that forget this die in committee.
One possibility remains that transformation plans almost never consider: not repaying. In finance, when a debt exceeds any capacity to repay, you do not dig in, you restructure. You renegotiate, you sell assets, you shrink the perimeter.
The transposition is brutal, but sometimes right: abandon the perimeter that made the debt callable. Kill the product whose business rules no one can maintain anymore. Close the channel that demands a process no one masters anymore. Stop serving the use case that ties up a quarter of support for a hundredth of the revenue. Heroic repayment, the kind that claims to document everything, modernize everything, and keep everything, is often the worst possible allocation of capital.
Restructuring is not an admission of failure. It is the accounting recognition that some liabilities are worth more dead than repaid.
Industrial history has already played this score. At the end of the nineteenth century, factories replace the steam engine with the electric motor and expect massive gains. They do not come: for decades, productivity disappoints. The economist Paul David documented this paradox. The gains arrived the day factories were reorganized around the new energy: one motor per machine, horizontal buildings, flows redesigned for logistics rather than for mechanics.
Vertical factories, built around the central drive shaft, were a structural debt. As long as it was not paid down, and it sometimes took razing buildings to pay it down, electricity did not deliver what it promised. Installing electric motors without reorganizing the factory: this is precisely deploying AI copilots on undocumented processes. The technology is present, the return absent, and the difference between the two has a name: debt.
That leaves the discipline to be institutionalized. Our proposal holds in one artifact: the register of organizational debt, kept at the same rank as the risk register, and reviewed on the same cadence.
Four columns. The liability. Its covenant. The chosen regime (repay, service the interest, let it ride, restructure). The review date. Not an exhaustive inventory: the short list of liabilities whose covenant has broken or threatens to break, each with a dated decision.
The difference between an indebted organization and an over-indebted one is not the amount. All high-performing organizations borrow: debt taken on knowingly is leverage, and refusing all borrowing would mean refusing all speed. The difference is the existence of the register. The first knows what it owes, at what rate, and what it has decided to do about it. The second discovers its liabilities the day a creditor starts to speak.
That creditor now has a name: the transformation you have just launched. The bill is on the table. The only question left is the one every debtor eventually faces: what is your repayment plan?
Analysis
The era of the interface as a "cover-up" for organizational complexity is ending. With the rise of the agentic, design does not disappear: it shifts from the surface of the screen to the structure of the contract. The click no longer carries authority — semantic clarity does. A case for moving from designing journeys to architecting robust business models.
Analysis
Organizations invested millions to deploy AI. Their teams refuse to use it. This isn't a training problem, or a change management problem. It's a coherence problem: incentives, processes and metrics still reward the very behaviours AI was meant to replace.
Analysis
L'avènement de l'IA agentique rend obsolète l'immense infrastructure méthodologique bâtie par l'industrie Tech depuis vingt ans. Alors que la production de code devient une commodité, le véritable défi des dirigeants n'est plus technologique, mais structurel : il s'agit de démanteler l'encombrant "exosquelette Agile" devenu un frein, pour réallouer massivement le capital financier et humain de l'usine à code vers la stratégie et la découverte de valeur.