The security review passed. The DPO signed off. Procurement got the contract to a sensible number. Then someone on the works council asks whether the new documentation tool records what employees do on screen, and the pilot stops for four months.
Almost nobody in this software category writes about this, because the vendors that dominate process documentation are American and their home market has no equivalent obligation. If you are buying in Germany, the Netherlands, France, Belgium or Spain, the co-determination question arrives before the product question.
Full disclosure: we build Loopyback, an EU-hosted documentation tool with a browser extension that captures clicks and screenshots when the user starts a recording. This is a rule we live under, not one we watch from a distance. It is also not legal advice: the point is to help you ask counsel and your works council the right questions early rather than late.
Why a documentation tool trips a monitoring rule at all
The instinct is that this cannot apply: the employee presses record, captures their own screen, and the output is a how-to guide. Nobody is being watched. That instinct is wrong about the legal test in at least two countries.
In Germany, §87(1)(6) of the Betriebsverfassungsgesetz gives the works council co-determination over introducing and using technical devices "intended to monitor the behaviour or performance of employees". Settled case law reads intended as objectively suitable. The Federal Labour Court restated it in July 2024 (1 ABR 16/23): the device is covered where it is objectively capable of collecting or recording behaviour or performance information, and the employer's subjective intention is irrelevant.
The Netherlands puts the same idea in the statute rather than the case law. Article 27(1)(l) of the Wet op de ondernemingsraden requires works council consent for facilities "aimed at or suitable for" observing presence, behaviour or performance. And 27(1)(k) independently requires consent for any arrangement on processing employee personal data, so arguing your way out of (l) lands you in (k).
A tool producing a timestamped click sequence and screenshots attributable to a named employee is objectively suitable for generating behaviour information. Whether you would ever use it that way is a different question from whether the rule applies.
What is owed, by country
| Body | Legal hook | What is owed | Threshold | If you skip it | |
|---|---|---|---|---|---|
| Germany | Betriebsrat | §87(1)(6) BetrVG | Co-determination, normally a Betriebsvereinbarung; deadlock goes to the Einigungsstelle | Where a works council exists | Council can seek an injunction to stop use |
| Netherlands | Ondernemingsraad | WOR art. 27(1)(k) and (l) | Prior consent (instemming), not consultation | OR mandatory from 50 employees | Decision void if the OR invokes nullity in writing within one month |
| France | CSE | Code du travail L2312-8 II 4° | Information and consultation before the decision (L2312-14) | CSE from 11; consultation attributions from 50 | Délit d'entrave; evidence risk under L1222-4 |
| Belgium | Conseil d'entreprise / CPPT | CCT nr. 39 (1983) | Information and consultation at least 3 months before rollout | 50 employees, and only where ≥50% and ≥10 workers of a job category are affected | Breach of a binding collective agreement |
| Spain | Comité de empresa | LOPDGDD art. 87.3; ET art. 64.5(f) | Representatives must participate in drawing up device-use criteria; prior committee report | Comité from 50 | Supreme Court STS 225/2024: unilateral adoption declared null |
Outside these five, Directive 2002/14/EC sets the floor: undertakings with at least 50 employees, or establishments with 20, must inform and consult on decisions likely to lead to substantial changes in work organisation. The name of the body changes; the obligation does not.
The Belgian nuance cuts the other way for once. CCT 39's three-month clock only starts if the technology has important collective consequences, tested as 50% of a job category and at least ten workers. A rollout to a six-person support team probably does not meet it, though general works council information rights still apply.
Does "the employee presses record" help?
It helps with the outcome. It does not remove the process.
No published decision squarely addresses employee-initiated, own-screen-only capture, so anyone answering with confidence is guessing. The honest position: self-initiation is a weak argument that nothing is owed, and a strong argument about what the agreement should say. A works council that would fight a silent background monitor usually approves a press-to-record tool quickly, provided the agreement is narrow.
What self-initiation definitely does not create is GDPR consent. Opinion 2/2017 of the Article 29 Working Party is blunt: employees are almost never in a position to freely give, refuse or revoke consent, so consent is highly unlikely to be a valid basis in an employment relationship. Pressing a button in a tool your employer told you to use is not free consent. Use Article 6(1)(f) with a documented legitimate interests assessment, and treat press-to-record as evidence of proportionality rather than as your legal basis.
What to put in the agreement
The agreements that get signed quickly tend to contain the same six commitments.
1. Purpose limitation in writing: capture is for producing work instructions, and captured material will not be used to evaluate individual performance or conduct.
2. No aggregation across employees, and no dashboard of who recorded what, how often.
3. User control: recording starts only on explicit user action, and the user can review and delete before anything is shared.
4. Redaction by default for personal data in screenshots, both the employee's own and any customer data on screen.
5. A retention period for recordings and guide versions, with a disposal rule.
6. Data location and sub-processors, named.
That last point is the honest version of our sales pitch. Hosting in Belgium removes the international transfer chapter and shortens the sub-processor list. It does not exempt anyone from §87(1)(6) or article 27 WOR.
Do you need a DPIA?
Probably, though it is genuinely borderline. None of the national blacklists cleanly catch this: the CNIL entry requires monitoring "de manière constante", the Dutch list requires large-scale and systematic monitoring, the German DSK list requires extensive processing of behaviour data. Episodic, user-triggered capture arguably meets none of them.
But under the WP248 criteria the tool scores criterion 7 (employees are named as vulnerable data subjects because of the power imbalance) and criterion 8 (innovative use of new technology), and the Working Party's rule of thumb is that two criteria indicate a DPIA. Screenshots are also content-rich and will incidentally capture third-party data.
Do it. It is cheap relative to the exposure, and every one of the five processes above will ask for it anyway. Our post on documenting security procedures under NIS2 covers the adjacent problem of documentation that is itself sensitive.
Common mistakes
Starting the works council conversation after signing. In France consultation must precede the decision, not the rollout. Signing first inverts the sequence and the CSE will say so.
Treating a pilot as exempt. A 20-person pilot is an introduction and use of the technology. Several of these obligations attach to introduction, not to scale.
Relying on employee consent. The single most common error, and the one a supervisory authority spots fastest.
Assuming the German picture is stable. §26(1) BDSG is widely regarded as inapplicable after the CJEU ruling in C-34/21 (March 2023), a replacement employee data protection act has been drafted and abandoned once, and C-65/23 (December 2024) confirmed that a works agreement used as a legal basis is subject to full judicial review of necessity. Build so the processing survives being rebased on Article 6(1)(f).
Discovering hosting late. If the answer to "where is the data" is a US region, you have added an international transfer discussion to a co-determination discussion. Our European SOP and documentation software comparison sets out the vendor questions.
What Loopyback does not do
We do not offer silent or background capture. An administrator cannot start a recording on someone else's screen. There is no idle-time tracking, no activity analytics and no per-employee usage dashboard. That is a product decision, and it costs us deals with buyers who wanted exactly those things.
We also do not supply a legally reviewed template works agreement, and we are not an employment law firm. What we can do is answer in writing, before you start, the factual questions a works council will ask: what is captured, when, where it is stored, who can reach it.
FAQ
Our works council has never reviewed a browser extension. What do they need?
A plain description of what is captured and when, a data flow showing browser to server and in which country, the retention rule, the list of people with access, the DPIA, and the draft purpose limitation. Bringing all six to the first meeting shortens the process more than anything else.
We have 30 employees in Germany and no works council. Are we clear?
§87 requires a works council to exist. With none, there is no co-determination right to exercise. GDPR obligations still apply regardless of headcount.
Does an EU-hosted tool avoid the works council process?
No. Data location is a data protection question; co-determination is an employment law question. Hosting in Belgium answers the first one only.
What if we only document processes in one department?
The obligation usually attaches to introducing the technology rather than to its scale, with Belgium's CCT 39 threshold as a partial exception. Scope limits are worth writing into the agreement, but rarely a reason to skip the process.
---
If it helps, we will answer your works council questionnaire for our own product in writing before you commit to anything. The free plan is enough for a technical pilot; paid plans start at €16 per month, €24 per seat for teams. Details on pricing.
Founder of Loopyback
Daan is the founder of Loopyback, a Belgium based tool that turns workflows into step by step guides. He writes about documentation, SOPs and getting knowledge out of people's heads.