FAQ
Frequently asked questions
Clear answers about wallet credit, usage, subscriptions, and how Tycoon charges for work.
Will my customers know they're talking to AI?
Tycoon's default is transparency: the support agent identifies as an AI when asked, and signs messages as the company's support team. Most customers don't care — they care about fast, accurate resolution — and studies (Intercom, Zendesk 2025 reports) consistently show AI-resolved tickets hit higher CSAT than slow human-resolved tickets. If your brand requires a named human persona for compliance or brand reasons, the COO configures a handoff rule: AI handles until complexity or escalation threshold, then routes to a human. Either model works. Hiding AI when customers directly ask, however, we don't do — that trust cost is not worth the short-term illusion.
What if the AI support agent gives a wrong answer?
Same containment loop as a human team: low-confidence answers trigger escalation rather than guessing. All responses reference the knowledge base, so corrections happen at the KB level and propagate instantly. Any founder or COO correction on a ticket updates the operating memory so the mistake doesn't repeat. Compare this to onboarding a human support hire, who makes the same mistake for weeks before patterns emerge. Tycoon's support agent reaches human-level accuracy on product-specific questions within the first 40-60 tickets as the KB fills in. Generic support (password reset, billing) is accurate from hour one.
Can it actually process refunds and cancellations?
Within policy limits set by the COO and CFO. The refund policy is codified (e.g., within 14 days, under $200, no prior refunds) and the support agent processes matching requests end-to-end through Stripe without founder involvement. Anything outside policy gets escalated with a recommendation and the full customer context. This is a huge founder-time saver: refund processing is one of the most common founder interruptions on small teams, and most refunds don't require judgment — they require a decision to be made in the customer's reasonable timeframe, not three days later when the founder logs in.
How does it detect churn risk?
Several signals in combination: a cancellation intent in a ticket, a sudden usage drop versus a customer's baseline, failed payments on a paying account, negative CSAT, or a feature request the customer has raised multiple times unresolved. When two or more signals fire, the support agent flags to the CEO with context and a drafted save attempt — usually a personal message from the founder, sometimes a policy exception (free month, onboarding call). Recovering even a fraction of would-be cancellations this way is often the single highest-ROI part of the support role economically, because churn compounds against LTV harder than almost any other metric.
How is this different from tools like Intercom's Fin or Ada?
Fin and Ada are AI chatbots bolted onto a support inbox. They answer questions and hand off to humans. Tycoon's AI support agent is an employee: it owns tickets from inbox to resolution, maintains the help center, coordinates with the CTO for bugs and the CFO for refunds, runs onboarding follow-ups, and reports insights to the CMO. The chatbot model solves 'deflection'; the teammate model solves 'function ownership'. For a one-person company where you don't have a support function yet, the teammate model is the unlock — you go from 'no support' to 'real support function' without hiring. For a larger team, the comparison is more nuanced and a hybrid stack can make sense.