Typewise AG, a Swiss AI startup, launched Nova today, a system that builds, runs and continuously improves customer-service AI agents without requiring dedicated operations staff. The platform addresses a persistent bottleneck: the operational work needed to keep AI agents accurate and current often prevents pilot projects from reaching full production.
How Nova works behind the scenes
Nova functions as an orchestrator, not a customer-facing chatbot. It maintains agent knowledge, instructions, workflows, evaluations and guardrails as business conditions shift. The system monitors agent performance, tests proposed changes and adjusts operations automatically. Customers retain control over what agents are permitted to do.
When a request arrives, an AI supervisor identifies the customer's intent and routes the case to a specialist agent. That agent pulls information from more than 3,500 connected knowledge bases and business systems - including CRM, ERP, billing, commerce and logistics platforms. It can check orders, change bookings, update addresses or issue refunds.
Companies authorize connections and describe in plain language what the AI should do, avoiding the need to program every workflow. Administrators set automation limits and require approval for sensitive or high-value actions, with each action logged for auditing. Cases needing judgment transfer to a human employee through Typewise's inbox or an existing help desk, with full conversation context attached.
Early adoption and performance metrics
Typewise reported concrete results from two customers. At HealGreen LLC, a Hamburg telemedicine platform, AI agents now handle 70% of incoming inquiries across email, chat and WhatsApp, fully resolving between 75% and 85% of them. Beurer North America LP autonomously resolves 65% of incoming inquiries with a success rate above 90%, and average handling time for human-fielded requests dropped from 12 minutes to six.
Nova initially drafts replies alongside human employees, letting teams review its work before assigning more request types. An initial use case can go live within minutes, though most customers expand deployments over several days. The platform integrates with applications from Intercom, Zendesk and Shopify, and supports email, chat and WhatsApp channels.
Pricing and data handling
Nova uses outcome-based pricing: fully resolved requests incur a full charge, partial resolutions cost half, and unresolved requests are free. Customer data can be hosted in Europe or the U.S., with zero data retention and no customer information used for model training.
"Every company that deploys AI agents creates a new operations job," said Typewise CEO and co-founder David Eberle. "Someone must monitor them, update their knowledge, test changes and decide when they are ready for more responsibility. Nova is that 24/7 operations team."
Why this matters for customer support professionals
Nova targets the operational burden that falls on support teams after AI tools go live. Instead of manually updating knowledge bases, testing agent responses and monitoring drift, teams can define guardrails in plain language and let the system handle maintenance. For supervisors managing AI Customer Service Training, this shifts the role from constant oversight to strategic governance - deciding what agents can do, not micromanaging how they do it. The outcome-based pricing also removes the financial risk of paying for failed resolutions, which matters when justifying AI investment to leadership.
Your membership also unlocks:
