Autopilot
// For works councils & social partners

AI in the Mittelstand. With the works council, not against it.

We know: AI rollouts rarely fail on the technology and almost always on the question „who gets to talk?". Four commitments we want to be measured against. Written for works councils, union secretaries, HR leads and everyone who sits at the table when a mid-sized company introduces AI.

// Four commitments

What we put on ourselves.

Not „employee-friendly" as a label, but as a position. These four sentences are in the standard retainer contract — not just on this page.

01

Works council from day 1

In co-determined companies, we include the works council from the first conversation — before the rollout, not after. § 87 of the German Works Constitution Act in practice, not goodwill. If the works council isn’t involved, we don’t come.

02

No employee-tracking mechanisms

We don’t build tools that track individual employee performance, login times or usage behaviour at the person level. Anonymised team aggregates yes — personal profiles never. Contractually guaranteed.

03

We come to the works council meeting

If the works council wants our method explained to the workforce — we come. In person, on site, in a works council assembly. We answer the questions that come up. The uncomfortable ones too.

04

Job reduction is not the purpose

We don’t take mandates whose declared goal is staff reduction through AI. We work with Mittelständler who want to keep and unburden their workforce — not replace it.

// Anti-tracking guarantee

What our tools NEVER capture about your workforce.

Specifics. The five points are in the standard retainer contract and AVV annex B. Anyone who wants to circumvent them doesn’t get a contract from us.

  • No individual performance profiles We don’t measure „who is how productive with AI". The question doesn’t arise.
  • No route or location tracking We don’t log where or when someone used the tool. Field workers aren’t monitorable.
  • No screen or keystroke capture Classic „workplace surveillance" mechanisms are technically not provided in our systems.
  • No person-level usage stats in reports Quarterly reports show team aggregates. Who clicks how often is not part of our reporting.
  • No transfer of behavioural data to third parties Not even anonymised in network studies. Employee behavioural data is not an asset for us.
// The job question

Job reduction is not the purpose. Here is why.

The most common question we hear from workforces is: „Does this make us redundant?" An honest answer deserves an honest position. We give it here — and present it in every works council meeting.

What we do, what we don’t do:

  • We don’t take mandates whose success criterion is „staff reduction". Period.
  • When a GM says in the first conversation „we want to cut 20% of jobs with AI", we decline. Even if the contract value is high.
  • We work with Mittelständler who have a skills shortage and want to unburden their workforce — not replace it.
  • We provide material for training that qualifies employees, not degrades them.
  • We recommend tools that enable knowledge preservation (RAG, service KB) so retiring knowledge holders can hand over well — not so juniors have to learn less.
  • We recommend adoption speeds that are culturally sustainable. „All at once" is not a methodology, it is a recipe for trust loss.

If this position doesn’t fit your use case — we’re not the right provider. That is meant honestly, not as a rhetorical trick.

// Works council agreement

WCA template as a starting point.

So that the works council doesn’t start from zero. A sample works council agreement on AI introduction with the clauses documented above — anonymised aggregates, clear purpose limitation, sign-off responsibility, grievance procedure.

Read the WCA template (v0.1, draft)

Note: This WCA draft v0.1 is a first proposal, not a legally final document. Before it enters into force, please have it reviewed by your law firm or the responsible union. **Roadmap to v1.0 (legally final):** planned Q3 2026 after the first productive mandate with works council. We will then have the template cross-reviewed by a labour-law-specialised firm and the IG-Metall legal department — updates on this page will be marked with a date.

// Co-determination in our approach

How works-council involvement actually happens with us.

Three phases, in each the works council is present. If a provider can’t show you this, you’re in for a bad rollout.

01

Before contract signing

First conversation between GM and us — afterwards presentation of the method to the works council BEFORE anything is signed. Works council reservation in the initial contract.

02

During build-up (first 4-6 weeks)

First tools are selected together with the works council. For high-risk applications (EU-AI-Act Annex III), the works council gets the classification before go-live. Workforce training is voluntary, not sanctioned.

03

In ongoing operation

Quarterly reviews with anonymised team aggregates — the works council gets the same reports as the GM. For new AI applications, renewed works council consultation. In case of conflict: joint meeting with works council + union secretary if desired.

// Employee data protection

When employee data is involved — EU hosting non-negotiable.

For applications in which employee data is processed, the stricter hosting class applies — no US-cloud defaults, no shadow-IT routes.

What this specifically means:

  • For employee data, no US-cloud default. Hetzner (Germany), IONOS (EU) or on-prem as standard.
  • Pseudonymisation before API call where technically possible — names + identifying data are removed before any model sees them.
  • The DPA with the mandate explicitly regulates the employee-data categories (annex A).
  • A grievance procedure via the mandate’s data protection officer is provided in the standard contract.
  • In case of DPO concerns: joint meeting before the tool goes live.
// For unions

We want to talk with you.

AI in the Mittelstand will shape how work looks over the coming years. We believe: it goes better if providers and unions talk to each other early — rather than in the dispute. If you want to talk with us about AI standards in the Mittelstand, get in touch.

Request a dialogue meeting
// Frequent questions

What works councils often ask us.

Is your sign-off block a substitute for co-determination?

No. The sign-off documents our quality control — it does not replace a works council agreement under § 87 BetrVG. The two mechanisms complement each other: the sign-off shows who is responsible on our side, the WCA regulates participation on your side.

Does the named reviewer in the sign-off track the performance of our people?

No. The sign-off reviewer is a person from OUR team, not from your workforce. Your employees do not appear by name in the sign-off.

What happens if the GM wants to use AI for staff reduction?

We don’t take the mandate. This is non-negotiable. If you experience a provider pitch that hides this: send us a short email, we are happy to share our evaluation criteria.

Do you offer training for the works council?

Yes. An initial works council introduction (45-90 min) on how it works, data protection aspects, EU-AI-Act is part of every mandate — free, on site or remote. Follow-up training as needed.

What if our DPO approves but we still have concerns?

Then we talk. The DPO approval is a precondition, not an end point. If co-determination concerns remain, we clarify them separately — with the GM, the works council and, if applicable, the responsible union.

Can we view the sign-off block or quarterly reports publicly?

Per mandate: yes. In the mandate portal, the works council can get access to the anonymised team aggregates and the risk-class statistics per output. Specific mandate content is subject to confidentiality.

// Let’s talk

Works council consultation or first conversation.

We come to your next works council meeting and introduce ourselves — free, with no sales pressure. Or we talk to the GM first and plan the works council involvement together.