The current MVP is rules-led, explainable, and adviser-bounded.

This page explains where the current logic sits today, how risk scoring is treated, and how AI will be introduced responsibly as the product matures.

Current MVP method

• The current MVP uses a structured rules engine to score cross-border readiness based on residence, business country, customer location, income source, payment platform, filing history, revenue band, and selected compliance goal.
• The output is intentionally framed as a preparation report: profile summary, risk reasons, exposure areas, record checklist, deadline checklist, and adviser questions.
• The current MVP does not claim autonomous tax advice, official filing, or final legal/accounting conclusions.
• The AI direction is to support explanation, document classification, adviser-note generation, and jurisdiction-pack guidance under professional-review boundaries.

Risk score inputs

• Multi-jurisdiction exposure increases risk because residence, company location, and customer location may trigger separate review questions.
• Estonia-linked business with non-Estonian residence increases the need for management-location and adviser review.
• EU or multi-country customers increase VAT, invoicing, customer-location evidence, and OSS-readiness questions.
• Unclear filing history, notices, or penalties increase escalation urgency.
• Payment platforms such as Wise, Stripe, PayPal, marketplace wallets, and crypto increase record-reconciliation needs.
• The score is a triage indicator only; it is not a tax decision or legal classification.

AI roadmap

• Phase 1: rules-based readiness checker and structured report generation.
• Phase 2: AI-assisted plain-language explanations with source and professional-review boundaries.
• Phase 3: document classification for invoices, bank/payment exports, contracts, and customer-location evidence after secure storage is enabled.
• Phase 4: adviser-note generation and missing-record detection for accountants and corporate-service providers.
• Phase 5: jurisdiction-specific knowledge packs reviewed by qualified professionals before public use.

AI governance

• No generated output should be presented as final tax, legal, accounting, immigration, or official filing advice.
• High-risk profiles should be escalated to qualified professional review.
• Jurisdiction-specific prompts should be reviewed by local advisers before commercial release.
• Sensitive records should not be ingested until secure storage, retention, deletion, and access-control workflows are ready.
• AI outputs should remain explainable: show inputs, assumptions, limitations, and adviser questions.