A support desk that answers in its own product's language
A CRM company in Europe needed real answers to specific product questions — without any ticket content leaving an environment they control.
- Sector
- CRM software, Europe
- Service
- Conversational AI
- Deployment
- Private, no public LLM
- Constraint
- Customer data could not leave the approved environment
The client sells CRM software to businesses across Europe. Their support team handles questions from those businesses about how the product is configured and behaves — permissions, integrations, data model, edge cases in workflow automation.
Volume was routine and repetitive, but the questions were not simple. A useful answer often depended on how a specific account was set up. That combination is where generic support chatbots fail: high enough volume to be worth automating, specific enough that a canned response is useless.
There was a second constraint that ruled out most of the market. Support tickets contain their customers' customer records. Under their own contractual commitments, that content could not be sent to a third-party model provider, whatever that provider's retention policy said.
A conversational support agent grounded in the client's own material and connected to the systems that hold the answer.
Product documentation and resolved-ticket history, so answers reflect how the product actually behaves and how the team has answered before — with the source retrievable.
The agent can look up account state and raise or update a ticket, so it completes the task rather than describing what the customer should do next.
When the agent hands over, the conversation and the attempted resolution go with it. The human does not start from nothing.
Defined limits on what it will attempt, and an instruction to say it does not know rather than produce a confident guess about a customer's configuration.
Entirely on privately hosted open-weight models. No conversation, ticket or account record is sent to OpenAI, Anthropic, Google or any other model provider.
Every request and response is logged inside the approved environment, so the client's compliance team can inspect what the model saw without asking us for an export. The deployment target was settled with their security and legal teams before any code was written.
How we deploy privately →The privacy position is what made the project approvable at all
A build on a public model API would not have passed the client's own commitments to their customers. The choice was not between a cheaper approach and a more private one — it was between a private deployment and no project.
This is the pattern we see most often in European software and regulated industries. The pilot is easy to justify and impossible to ship.
Tell us the question your team answers fifty times a week.
A discovery session is a working conversation, not a demo. Bring your security lead — the deployment question is the one we want to start with.