Audit the 552 entities in the Plan’s perimeter before abolishing anything

Why start with an audit rather than a list of bodies to abolish?

PROPOSALMeasure 3.01 opens the Agencies / operators volume because structural reform should begin with evidence, not with a name on a list. Before an entity is merged, reintegrated or abolished, the State needs to know what it actually delivers, what the full cost is, which legal obligations survive and which organisation would take over the work. The figure of 552 retained entities is the Plan's own working perimeter. It must not be presented as an official count of every French public body.

OFFICIAL DATAFor 2026, the State budget authorises 2,016,588 State jobs and finances, directly or indirectly, 401,389 full-time-equivalent jobs in State operatorsA French budgetary category for legally separate bodies closely tied to a State public policy and subject to State oversight.. ^[1] That is an employment measure, not a count of organisations. It illustrates the discipline this chapter applies throughout: every number needs a unit, a year and a perimeter.

What a State operator actually is

EXPLANATIONThe word “agency” is not a single legal form. An organisation called an agency may be a public establishment, an independent authority, an association or another type of body. Conversely, many bodies treated as State operators do not have “agency” in their name. The audit therefore starts with legal status and budget classification, not branding. ^[3]

The budgetary notion of a State operator is used for entities that carry out public-interest functions under strong State influence and are closely connected to a State policy. Separate legal personality may be justified by specialist expertise, regulatory independence, continuity of financing or operational flexibility. The relevant question is therefore not “How many agencies does France have?” but “What additional public value is created by keeping this activity in a separate body?”

ANALYSISA useful mission can be housed in an unnecessary structure, and a small structure can still create slow interfaces or blurred accountability. The reverse is also true: a costly body may be justified if independence or specialist capacity genuinely improves outcomes. The audit has to test those propositions rather than assume them.

Seven audit files: from the organisation's name to its full cost

1 — Legal identity, statutory basis and sponsoring ministry

The audit records the founding instrument, legal personality, sponsoring ministry, governing board, regional network, subsidiaries, borrowing powers and earmarked taxes. These facts determine whether a change can be made by regulation, needs primary legislation or requires a complex transfer of contracts and assets. ^[3]

2 — Mission and actual output

The second file describes what is actually produced: licences, controls, grants, research, datasets, technical advice, infrastructure or citizen services. The statutory mission must be distinguished from day-to-day activity. The mission may remain necessary even if the current corporate vehicle is not.

3 — People, money, property and systems

The third file consolidates budget funding, own-source revenue, payroll, jobs, cash, property, information systems and external procurement. French budget law already requires the State-operator report to publish a large part of this information, including execution data and property metrics. ^[2] Delta-Sierra's method should reuse existing official data before creating new reporting burdens.

4 — Interfaces and genuine overlaps

A genuine overlap is not established merely because two bodies work on the same policy area. The audit maps who prepares a decision, who signs it, who funds it, who holds the data and who is accountable to Parliament. An overlap exists only where comparable functions serve the same need and can be consolidated without destroying required independence or specialist capacity.

5 — Quality and time

Savings must be read alongside service quality. The audit therefore includes a small set of operational metrics: median processing time, backlog, error or rework rate, availability of critical systems, user satisfaction and, where meaningful, cost per output.

6 — Transfer liabilities

Every obligation that survives abolition must be listed: staff, litigation, live grants, contracts, archives, buildings, information systems and scarce skills. Ignoring these items is the fastest route to a fictitious saving.

7 — Counterfactual operating model

Finally, the audit describes the day after reform. Which directorate takes over? With how many staff? Which budget line? Which system? If those questions cannot be answered, abolition is not implementation-ready. If the answer is simple, legally sound and cheaper, the structural case becomes testable.

Five possible outcomes, not a binary verdict

How to cost reform without counting the same euro twice

COSTINGThe starting point is not the total amount of money passing through a body. An organisation may distribute a billion euros of grants while costing only a fraction of that to operate. If the grants remain, the billion does not become a saving. Conversely, reintegration may reduce governance and support costs while requiring the sponsoring ministry to recreate staff, systems and controls.

Time savings and clearer accountability may be important, but they should not be converted into euros without evidence. A reform can improve the chain of responsibility while producing only modest cash savings; the site should be able to say so.

What changes in practice?

For citizens and businesses, the audit matters only if it ultimately removes unnecessary interfaces: fewer duplicated portals, fewer repeated data requests and a clear owner for each decision. For public servants, the same reform must avoid the slogan “the agency is gone” without an answer on role, location, manager, tools or specialist career paths. Service continuity is part of the saving test.

For Parliament and the public, each completed audit should produce a standard record: mission, legal basis, budget, jobs, full cost, performance measures, recommended operating model, expected net saving, transition cost and final government decision. The audit then becomes a reusable democratic-control dataset rather than a report that disappears after publication. ^[2]

Serious objections

“The audit itself could become another bureaucracy.”

That risk is real if every ministry invents a different template and asks for hundreds of new pages. The response is a common dataset, maximum reuse of existing official records and a small central team focused on reconciliation and outliers rather than document production.

“Separate status can protect expertise from short-term political pressure.”

Sometimes it can, particularly in regulation, research or evaluation. The autonomy test must therefore look for legally necessary independence, rare expertise, long-term funding, neutrality or specialist operating needs. The correct recommendation may be to retain the body while pooling back-office functions.

“Savings will still be overstated.”

That is exactly what the counterfactual budget is designed to prevent. A saving cannot be treated as certified until the receiving organisation, transferred staff, migration cost and recurring replacement costs have been identified.