AI answers reviewed by our team (80–100% at launch to calibrate with you, then at least 4% in steady state). Human and AI tickets are reviewed the same way by your account manager. Every hand-off logged with its reason. CSAT tracked per path.

Hand-off to a person on negative sentiment, on an escalation rule built into your processes, or as soon as the customer asks for a person.
You choose the mode: strict (AI hands off as soon as it has no validated answer) or guided (AI steers the customer toward an existing process).
Your agents and your AI share one knowledge base, updated in under 24h.
Routing by contact reason: AI alone / AI-assisted agent / human only. Set on the platform, with us or for you.
Run your own quality checks in the Onepilot app, by channel, reason or agent. Co-built QA grid and calibration sessions.
Weekly DSAT review and root-cause analysis shared with your team. One accountable account manager owns quality and cost.
Work can stay entirely inside your own tools.
Email and chat are automated end-to-end. Automation can start from day one.
Once we have access to your helpdesk data, we map your tickets by contact reason within 48 hours, then estimate what can be automated.
of support automated
The other 60% stays with people
to start
Book a meeting
One blended price per interaction.
Automation roadmap in the contract: target rates at month 1, 3 and 6, set from an analysis of your historical tickets, with a review clause if missed.
Every hand-off to a person is logged with its reason (missing process or knowledge, technical error, or requested by your process) and reviewed with you each quarter.
You own the routing rules: you decide which contact reasons are AI-eligible and can extend the list yourself.
The proof over time: on stable contact reasons, the human share should go down month after month. You track it on your dashboard.
Platform hosted on AWS Paris. EU-hosted model option available.
You choose the model provider (Google, OpenAI, Mistral, Anthropic), all under enterprise contracts.
Your tickets are never used to train the providers' models. Quality reviews by your account manager improve your own AI setup only, never another client's.
Every AI answer and action is logged and auditable. You define which actions and pages the AI can access.
Access limited to shift hours, passwords never visible, cache wiped after each shift, sensitive pages blockable.
Virtual desktops hosted in France. Data processing agreement in every agent contract.
ISO 27001 · ISO 18295 · ORIAS accreditation. Security audit at onboarding, then every 6 months.




