Governing what we do not control: accountability as a pillar of performance management
Perspective, Governance
4conditions for
genuine accountability
the TAVC grid
Governing is not about understanding everything, it is about organising who answers.
Every organisation today handles things it does not understand from end to end: an algorithm, a complex financial asset, a subcontracting chain, a piece of data whose exact origin it cannot establish. The temptation is twofold: to prohibit out of fear, or to let things run out of fascination. There is a third way, more mature and more demanding: governing without controlling everything. This article proposes a framework for doing so, accountability, and a grid that makes it operational rather than merely declaratory. Events of 2026 serve as illustrations, but the principle itself will not go out of date.
Accountability is not responsibility
Let us begin with a word that is often used and rarely defined. In DiagTraM's vocabulary, accountability is the obligation of an actor to give an account of a power it exercises, before someone entitled to demand that account, with a consequence if the account is not given. It differs from responsibility in the everyday sense, which means being the cause of an outcome. One can be responsible without being accountable, when nobody has the power to hold you to account. And one can be held accountable without having controlled everything: this is precisely where the concept becomes useful.
Responsibility
Looks to the past
Being the cause of an outcome. It looks for someone to blame once the incident has occurred.
Accountability
Organises the present and the future
The obligation to account for a power, before whoever is entitled to demand it, with a consequence. A continuous mechanism, independent of any accident.
This nuance has practical consequences. A leader who confuses the two waits for an incident before naming a culprit. A leader who has grasped the difference builds, upstream, a chain that obliges everyone to document and to explain themselves, regardless of any accident.
Why total control is an illusion of governance
Conventional wisdom holds that one can govern only what one understands perfectly. This belief, however reassuring, is false, and it paralyses. No board of directors understands every line of code in an information system. No minister masters the detail of every contract awarded by their department. No central banker knows the future behaviour of every economic agent.
“Governing never means knowing everything. It means putting in place the mechanisms which ensure that, even in partial ignorance, someone will have to answer.”
Consider a concrete example. In 2026, the European Union brought into application a series of obligations for providers of general-purpose artificial intelligence models, with penalties of up to 3% of worldwide annual turnover or €15 million. The regulator does not claim to understand these models in their entirety; they are often opaque even to their own developers. It does something different, and smarter: it requires providers to document, to trace training data and to respond to requests, on pain of financial penalties. This is governance through accountability, not through technical mastery.
Regulating without understandingThe AI Act does not seek to pierce the opacity of the models. It requires providers to document, trace and respond, on pain of financial penalties.
The TAVC grid: four conditions for genuine accountability
Accountability that is proclaimed but not equipped is worth nothing. To make it effective, DiagTraM proposes a grid of four cumulative conditions, the TAVC grid. All four must be met; otherwise accountability remains a slogan.
The framework
The TAVC grid, in four cumulative conditions
01
T
Traceability
The action leaves a usable trace: a log, documentation, an audit trail. The raw material of any account rendered.
02
A
Attribution
The trace is linked to an identified actor. The antidote to dilution, where a collective failure becomes nobody's failure.
03
V
Verifiability
The account can be checked by a legitimate third party, against criteria known in advance. Blind trust becomes verified trust.
04
C
Consequence
A deficient account triggers a real effect, whether financial, statutory or reputational. Without it, the other three lose their force.
We trace, we attribute, we verify, we draw the consequences. Remove one link and the chain breaks: tracing without attributing means knowing what happened without knowing who must explain; attributing without consequence means naming responsible parties who risk nothing.
Example dashboardIllustrative model
Tracking accountability, not control
Traceability92%
Attribution68%
Verifiability45%
Consequence28%
Illustrative example. An accountability dashboard does not score your understanding of a system; it tracks the state of the four conditions. Here, the chain breaks at consequence: tracing is sound, but the absence of any effect drains the obligation of its force.
A cross-cutting discipline, in the public and private sectors alike
The framework applies well beyond artificial intelligence. Whether the subject is a budgetary windfall, the accounts of a listed group or a local authority's investment, the same grammar applies: make the use traceable, attribute it to an explicit objective and make its application verifiable over time, on pain of consequence.
3%of worldwide turnover: the penalty ceiling set by the EU AI Act.Source: AI Act (2026)
+34%in CAC 40 profits in the 1st half of 2026, to €73bn.Source: BFM Bourse / TradingSat
€84bnof local investment in France in 2025, under tight budgetary constraints.Source: AFL Barometer 2026
When the International Monetary Fund recommends that Brazil allocate its exceptional oil revenues to reducing debt rather than to current spending, it is applying a logic of fiscal accountability. When a listed group publishes its half-year accounts, the performance it reports matters less, to those who govern, than the quality of the account that accompanies it: a sharp rise in profits has governance value only if it is documented, attributed to identified causes and verifiable by the market. And for a local authority investing under constraint, accountability is what makes it possible to justify every euro spent to its citizens.
Practical lessons for leaders
Three lessons emerge, applicable to any organisation.
Key takeaways
Three reflexes for leaders
1
Stop waiting for control
It will not come. Instead, put the four conditions of the TAVC grid in place around your sensitive assets: algorithms, finances, subcontractors, data.
2
Reverse the order of control
Do not start by measuring the outcome; demand the trace and the attribution first. A dashboard with neither an audit trail nor an owner is mere display.
3
Attach a consequence
Accountability without effect is mere courtesy. It is the consequence, however modest, that turns a commitment from an intention into a requirement.
An open conclusion
Governing has never been about understanding everything. It has always been about organising who answers, for what, to whom, and with what consequence. Technologies change, assets grow more complex, chains grow longer, but this grammar endures. The real question every leader should ask is not “do I control this system?” but “am I able, today, to say who answers for it and how I verify it?”. Those who can answer it will govern. The others will be at the mercy of events.
Sources and references
The concepts used in this article are restated in our own words. The sources below document the illustrative facts cited for 2026.
Penalty regime applicable to providers of general-purpose AI models, up to 3% of worldwide turnover or €15m AI Act, enforcement of Chapter V
A budget adjustment should not be judged by the balance achieved, but by the capacity for action it leaves intact. Three filters to assess its quality.