AI in Sales: How to Get Your Works Council On Board and Sign-Off Fast

AI This article was created with the help of AI.

Key takeaways

• According to Bitkom, 36 percent of companies already use AI, which makes a solid works agreement all the more important.
• Co-determination applies as soon as an AI tool is objectively suitable for monitoring behavior or performance.
• A pilot project needs defined stop criteria, fixed deletion deadlines, and a clearly limited group of participants.
• Bliro processes conversations in real time and stores no audio files, so the data scope for the works council's review can be clearly bounded.

When does co-determination apply?

When you introduce an AI sales assistant as a sales leader, productivity is usually the focus: you want to spare your team tedious desk work in the daily sales routine, keep CRM data reliable, and steer the pipeline with more transparency. But as soon as software is involved that processes conversation notes or captures activities, co-determination kicks in. The key provision here is Section 87(1) No. 6 of the German Works Constitution Act (BetrVG), which gives the works council a mandatory right of co-determination in the introduction and use of technical systems designed to monitor the behavior or performance of employees^[1].

What matters in practice is a legal principle: it does not depend on whether you as a manager actually intend to monitor employees. For the statutory co-determination trigger, the objective technical suitability of the system for performance or behavior monitoring alone is sufficient. Since modern sales assistants structure conversation data, capture meeting durations, and fill CRM fields, they generally fall under this standard of review.

• Objective suitability is enough: the mere possibility of drawing conclusions about work behavior from system logs, activity records, or conversation metadata triggers co-determination.
• No circumvention through declarations of intent: a verbal assurance that data will not be used for performance control does not lift the co-determination right.
• Negotiation is required before rollout: without the works council's prior consent or a ruling by the conciliation board, the system may not be used regularly in the company.

Instead of treating co-determination as a brake on innovation, you should steer the process proactively. If you understand early on which technical data points are generated and how they can be decoupled from manager access, you create the foundation for a legally sound and fast rollout.

What information does the works council need?

A common cause of delays in software rollouts is incomplete documentation. Works councils rarely block new technology on principle - they demand understandable evidence of exactly how it works. In its guidance on AI and data protection, the German Data Protection Conference (DSK) recommends that controllers define the concrete purpose and fields of application in advance in a binding way and involve employee representation early and in a structured manner^[2].

To speed up the review process, you should put together a transparent information package. This package must show the works council and the company data protection officer precisely how data is captured, processed, and stored. Unclear interfaces or vague descriptions inevitably lead to follow-up questions and time-consuming renegotiations.

‍

If you put these documents together before the first official meeting, you preempt typical concerns. That transforms the discussion from an abstract risk debate into a factual assessment of defined protective measures in the mid-market.

Which analyses need to be limited?

The core conflict with AI-powered sales tools lies in the separation between operational process support and individual behavioral monitoring. When systems analyze transcripts or fill CRM fields, performance parameters are inevitably generated. For a rollout to be viable, the boundaries for analyses under Section 87 (1) No. 6 of the German Works Constitution Act must be clearly defined and technically safeguarded^[1].

For you as a leader, this means: the focus must be on aggregated process data and relieving your sales reps of desk work in the daily sales routine - not on individual employee monitoring. The following categories of analysis should be strictly separated in agreements:

• Ban on individual rankings: Comparative rankings based on automated AI metrics (such as talk-to-listen ratios, sentiment scores or objection rates per rep) must not be used for personnel decisions.
• Self-determined coaching: AI feedback on conversation structures or methodologies like MEDDIC should primarily be available to the individual rep for their own professional development, rather than flowing automatically into management dashboards.
• Aggregated quality control: Trend analyses of customer feedback, product requests or market objections should be summarized at team level to rule out conclusions about individuals.
• Exclusion of sanctions: Agreements explicitly state that no disciplinary measures or performance warnings may be derived from raw data or AI analyses.

This clean separation takes away your sales team's fear of surveillance. Your employees use the assistant features willingly and without hesitation, while you still get reliable market insights that measurably move your business forward.

When is an impact assessment required?

In addition to co-determination, data protection law requires a structured risk assessment. According to the guidelines of the European Data Protection Board (EDPB), a data protection impact assessment (DPIA) must always be carried out in advance whenever a form of data processing is likely to result in a high risk to the rights and freedoms of natural persons.

In the sales environment, this affects two groups at once: on the one hand, external customers and conversation partners whose statements are processed, and on the other, internal employees whose work performance is affected. The use of new technologies such as generative language models in combination with employee data frequently meets the criteria for a mandatory DPIA.

1. Run a threshold analysis: Together with your data protection officer, check whether the use of AI meets the criteria for a mandatory DPIA due to systematic speech processing or behavioral monitoring.
2. Risk assessment for affected parties: Analyze concrete risks such as unauthorized data disclosure, misinterpretations by AI models or impermissible performance profiles.
3. Define remediation measures: Establish technical and organizational safeguards, for example role-based access controls, deletion concepts and confidentiality filters.
4. Involve the works council: Include the works council in the assessment to ensure transparency about the identified risks and countermeasures.

A diligently conducted impact assessment takes many worries off the works council's plate. It shows in black and white that risks were not ignored but effectively minimized through concrete technical safeguards before the system goes live.

What does a secure pilot look like?

Instead of pushing for a company-wide rollout right away, it makes sense to agree on a pilot phase with the works council that is limited in time and scope. The German Data Protection Conference (DSK) emphasizes that clear frameworks and protective rules must apply when testing new AI systems in order to protect employees from incalculable risks^[2].

A pilot project allows both sides to observe the actual impact of the AI sales assistant in practice without creating irreversible facts. For this, clear guardrails should be defined that give the works council maximum control.

‍

When the pilot rests on reliable criteria, it delivers objective insights instead of emotional fundamental debates. Your team tests how much time the assistant actually saves, while the works council can verify that the agreed protective limits are upheld in practice.

What belongs in the works agreement?

After successful testing, the works agreement forms the legal foundation for the permanent use of the AI assistant. It creates legal certainty for managers and protects the workforce. Following the guidelines of the German Data Protection Conference (DSK), all essential aspects of its operation should be bindingly fixed^[2].

A solid works agreement for AI tools in sales should not be overly complicated, but it must regulate the central protective rights and obligations unambiguously. The following building blocks belong in the document:

• Precise purpose definition: stipulating that the tool is used exclusively to ease workload, create visit reports, and update CRM systems.
• Transparency and information rights: regulating how employees are informed about the data processed and can access the summaries the system creates for them at any time^[2].
• Prohibition of misuse: an explicit ban on using automatically generated data for performance reviews, warnings, or salary decisions.
• Complaint procedure: establishing a simple point of contact for employees in case of suspected malfunctions, AI hallucinations, or data protection concerns.
• Regular evaluation: agreeing on annual review meetings between management and the works council to adapt to technical developments.

With such an agreement, you remove all legal uncertainty from your daily sales routine. Managers, employees, and the works council pull in the same direction to make processes more efficient and keep the focus fully on customer contact.

How does Bliro fit in?

No software is works-council-compliant in itself - the concrete introduction in the company only becomes co-determination-proof when purpose, data, evaluations, and access rights are bindingly agreed in advance. The technical architecture determines how many points need to be regulated at all.

The Bliro Sales Assistant automates meeting prep, creates structured conversation notes, and handles CRM updates directly after the customer meeting. What matters for the works council's review is what happens technically - and what doesn't:

• Audio: Speech content is processed in real time. Audio is not stored. There is no local audio recording on the device.
• Stored data: Deletion periods are freely configurable, so companies can adapt them to their own policies and the works agreement. Server location is Frankfurt am Main. A DPA and technical and organizational measures (TOMs) are in place.
• Access: Only the respective employee themselves can see conversations and the associated feedback. Managers receive only aggregated insights, for example about competitors or market objections, but no evaluations of individual performance. The role and permission model is defined together with the works council during the rollout.
• Evaluations: Bliro does not create individual performance rankings. Which analyses are actively in use is set out in the works agreement.
• Evidence: Bliro holds a SOC 2 Type 1 report. This and further security documents are available to data protection officers and works councils in the Trust Portal.
• Transparent CRM synchronization: Structured data flows selectively into Salesforce, HubSpot, SAP or Microsoft Dynamics. There are no further data flows to third parties beyond the subcontractors named in the DPA.

When you plan the rollout, involve the works council and the data protection officer from the very beginning, define the data flow, agree on clear boundaries for evaluations, and start with a limited pilot. That way you create the foundation for an implementation that the works council, data protection, and sales all support together.

Note: This article does not constitute legal advice.

Sources

1. https://www.gesetze-im-internet.de/betrvg/__87.html
2. https://www.datenschutzkonferenz-online.de/media/oh/20240506_DSK_Orientierungshilfe_KI_und_Datenschutz.pdf