FAQ
Frequently asked questions
Clear answers about wallet credit, usage, subscriptions, and how Tycoon charges for work.
What happens when the AI gets a classification wrong and the customer suffers?
Two guardrails. First, the AI only auto-sends responses for high-confidence routine tickets (password reset, factual how-to, refund within policy). Anything borderline goes to human review. Second, any negative signal after AI response (customer replies 'that didn't help', 'escalate', 'can I talk to someone') instantly escalates to human + flags the original classification for training. Typical false-positive rate (AI handled something it shouldn't): <2% after 60 days. The 2% gets caught by the escalation signal, not by the customer churning silently.
Our customers expect a human response. Will they get frustrated by AI-handled tickets?
Depends on quality, not on whether AI is involved. Customers don't want to talk to a human for a password reset — they want the password reset. For routine tickets, AI-handled is actually PREFERRED (faster, same quality). For emotional/complex issues (billing dispute, outage impact, cancellation), human-handled is preferred and the AI routes there. Tycoon optimizes for the outcome customers want (resolution speed for routine, human empathy for complex), not for 'AI or human' as the metric. Customer satisfaction usually goes UP, not down, when routed correctly.
We have specialized support (technical SaaS, regulated industry, enterprise with custom SLAs). Does this adapt?
Yes, with configuration. For technical SaaS: AI Customer Support integrates with your engineering logs and deployment status, so tickets like 'something's broken' get diagnosed with real signal (recent deploy? error spike in Sentry? affected customer's usage pattern?). For regulated industry (HIPAA, PCI, SOC 2): AI handles only non-PHI/non-PCI routine tickets, all sensitive-data tickets escalate to human. For enterprise custom SLAs: per-account SLA rules enforced.
What about tickets in languages we don't officially support?
AI Customer Support handles multi-language natively. Ticket in Japanese → translated for your team + response drafted in Japanese + back-translated for your review. Accuracy is solid for common languages (EN, ES, FR, DE, JA, ZH, PT). For rarer languages, AI flags for translation verification before sending. Adding language support is effectively free — customers in new markets get same-day responses in their language without you hiring bilingual support.
How does this compare to Zendesk's built-in auto-tagging or Intercom's Fin AI?
Zendesk's tagging is rule-based (keyword matching) and shallow. Intercom Fin handles common questions via your help center content. Both solve slices of the problem. Tycoon's difference: end-to-end ownership (classification + prioritization + routing + response + escalation) across platforms, with context from your other systems (Stripe, Linear, PostHog). Most teams find Fin handles 20-30% of tickets; Tycoon handles 50-70% because it can route to Linear for bugs, Stripe for billing, and your KB, plus use customer context from everywhere. Fin is a bolt-on for the help widget; Tycoon is the support system.