FAQ
Frequently asked questions
Clear answers about wallet credit, usage, subscriptions, and how Tycoon charges for work.
How do I trust the AI's SQL — what if it writes a query that's wrong and I make a decision on bad data?
Every query the AI writes is visible and auditable. You can see the exact SQL it ran, the raw results, and the interpretation. The AI also runs sanity checks: 'This query says revenue dropped 80% month-over-month — this is likely an error (checking...). Found: the Stripe data sync was delayed for 3 days last month. Adjusted numbers show 12% growth.' For critical decisions, you can require human review of the SQL before the analysis is considered final. Over time, you build trust by spot-checking queries in the first few weeks — most teams find the AI's SQL accuracy is 95%+ for standard business queries after the first month of schema learning.
What about data security — does the AI have access to all our customer data?
AI Data Engineer connects with read-only database credentials scoped to the specific tables and columns needed for analysis. You configure what it can and cannot see: 'access the users table but not the passwords column; access the events table but anonymize user IDs.' For PII, you can configure masking rules (emails → hashed, names → masked). The AI runs queries against your database; it doesn't copy or store your data. For companies with strict compliance requirements (HIPAA, SOC 2, GDPR), you can configure data access policies that the AI enforces at query time. The credentials are stored in your Vault, not in the AI's memory.
How is this different from asking ChatGPT or Claude to analyze my data?
ChatGPT and Claude can't connect to your database. You'd have to export data to CSV, upload it, and ask questions — and re-upload every time the data changes. AI Data Analyst has a persistent, read-only connection to your live database, so questions always run against current data. It also has persistent memory of your schema, your business logic, and your metric definitions — it doesn't need to be re-taught what 'MRR' means in your business every conversation. ChatGPT can help with one-off analysis; Tycoon's AI Data Analyst is the always-on analytics layer that updates your dashboards and catches anomalies while you sleep.
Can it handle our complex data model with 50+ tables and custom metric definitions?
Yes, but there's an onboarding period. AI Data Engineer documents your schema automatically on first connection. You then define your business logic in plain English: 'MRR = sum of active subscriptions where status is active or past_due, excluding trials. Churn = customers who were paying last month and are not paying this month, excluding those who downgraded to free.' The AI stores these definitions and uses them consistently. For very complex schemas (200+ tables, custom ETL, multi-source), the AI may need 1-2 weeks of guided querying to learn the nuances — you review the first batch of queries and correct any misunderstandings. After that, it operates autonomously.
Does it replace our data team or augment them?
For companies without a data team, it replaces the need to hire one for standard analytics and reporting. For companies with a data team, it augments them: AI handles the ad-hoc founder questions ('what's retention by cohort?'), the recurring reports (weekly KPI dashboards), and the anomaly detection — freeing human analysts to do deeper work (predictive modeling, experimentation design, custom research). Data teams that adopt this typically see their backlog of ad-hoc requests drop by 70% within a month, and their analysts spend more time on high-value work instead of 'can you pull this number for the board deck.'