Preparing an AI customer support platform for EU launch

Result

A clearer legal path to launch

The team received a prioritised launch checklist and draft customer-facing documents addressing AI interaction, user data and the external model provider.

  • ClientB2B SaaS startup
  • ProductAI customer support assistant
  • MarketEuropean Union

The client’s task

A SaaS startup planned to launch an AI assistant that answered end users’ support questions using a company’s help-centre materials and the context of each conversation. The product relied on an external model through an API. Before launch, the team wanted to understand what its role in the AI supply chain meant for the product, what users should be told and how to handle messages that could contain personal data.

Challenges and risks

  • The SaaS company, its business customers and the model provider each had a different role in the service.
  • End users interacted directly with an AI assistant.
  • Support messages could contain account details or other personal information.
  • The parties needed clear rules for data retention, provider access, human escalation and the limits of AI-generated answers.

What we did

  1. Mapped the product’s intended use, user journey and data flows.
  2. Assessed the parties’ roles and which AI Act transparency requirements might apply.
  3. Reviewed the model provider’s terms and data-processing arrangements, including prompt retention and reuse.
  4. Prepared user disclosures and service terms explaining the assistant’s role, handling of user input and access to human support.
  5. Created a launch checklist covering document updates, internal ownership and product controls.

Result in practice

The startup received a prioritised list of legal and product changes to consider before launch, together with draft documents for review and implementation. This gave the team a clearer view of data flows and responsibilities as it made launch decisions.

Adding AI to your product? Let’s discuss the legal work before launch.

AI and ML legal services · ← All cases

The scope and result of each project depend on its facts, target markets and the work agreed with the client.