Locally deployed AI model
Personal data, trade secrets, and internal documents are processed on the company’s own GPU servers in a Russian data centre. The model includes hardware CapEx, depreciation, electricity, colocation, and maintenance.
AI-powered financial control of documents
Three former employees of a major systems integrator launched AI-Kontur for mid-sized manufacturing and distribution companies. The product matches contracts, invoices, specifications, and acceptance certificates, identifies discrepancies, and prepares an explainable review for the CFO.
A customer pays between RUB 80,000 and RUB 95,000 per month, plus a one-time onboarding fee of RUB 120,000. Expected customer lifetime is approximately two to three years. Marketing spend per acquired customer equals five to seven monthly subscription payments and is recognized immediately, while subscription contribution is earned gradually.
Token costs
De-identified reference tasks are processed through external APIs. Token cost declines through caching, batching, and prompt improvements—from RUB 4,200 to RUB 2,200 per customer per month.
Q1Loan A: RUB 45 million. The first local GPU cluster is purchased.
Q5Loan B: RUB 35 million. The AI environment is expanded as customer acquisition peaks.
Y2 H2Free cash flow turns positive after growth slows and retention improves.
Y3 Q4A strategic buyer offers to acquire the business for 10 times MRR.
What AI-Kontur actually sells
AI-Kontur is a vertical AI system for financial control of commercial and accounting documents. It is not a general-purpose document chatbot: it sits between procurement, finance, accounting, and legal teams and helps determine whether a supplier payment is supported by the underlying contract and documents.
Target customer
Mid-sized manufacturers, distributors, wholesalers, construction firms, and engineering companies with many suppliers and a high volume of recurring documents.
- Contracts and amendments
- Specifications and purchase orders
- Invoices, acceptance certificates, and universal transfer documents
Operational pain
Employees manually reconcile documents, often under time pressure. A missed discrepancy can produce an overpayment, an incorrect tax document, or a payment made on terms that were never approved.
- Price, quantity, VAT, currency, or payment-term mismatches
- Slow approval and dependence on individual attention
- Poor visibility of financial risk before payment
Economic buyer
The CFO, head of financial control, or procurement director buys the product to reduce preventable losses and shorten the payment-approval cycle. Accountants, procurement managers, and controllers use it in daily operations.
- Subscription plus a one-time implementation fee
- Integration with the customer’s existing systems
- Recurring use before supplier payments are released
A typical control workflow
1 · IngestThe customer provides a supplier contract, specification, invoice, and closing documents through its ERP, EDI, or document-management system.
2 · UnderstandThe system extracts parties, amounts, dates, line items, VAT, payment terms, bank details, and contractual obligations.
3 · ReconcileAI-Kontur links related documents and checks prices, quantities, dates, terms, and approval rules across the complete document set.
4 · RecommendThe controller receives an explainable conclusion: approve, hold for clarification, or escalate—together with the discrepancy and estimated amount at risk.
AI is one layer of the system
The product combines OCR, PDF and spreadsheet parsing, an LLM for interpreting contract language, deterministic control rules, document matching, and integrations with accounting and workflow systems.
- AI handles variable language and document formats.
- Rules enforce customer-specific financial controls.
- Integrations place the result inside the payment process.
An intelligence layer above ERP and EDI
ERP and electronic document interchange systems primarily store documents, move them through workflows, and post transactions. AI-Kontur interprets document meaning, compares obligations with actual billing, and highlights exceptions before money leaves the company.
- ERP records the transaction; AI-Kontur tests its logic.
- EDI transports the document; AI-Kontur evaluates its content.
- The system supports a decision rather than making an autonomous payment.
Why customers can justify the subscription: a company processing hundreds of supplier invoices may approve hundreds of millions of rubles in payments each month. Preventing even a small share of pricing, quantity, tax, or payment-term errors can outweigh the recurring software fee. The product therefore sells measurable financial control—not document recognition by itself.
Excel model
AI_Exam_Case_FinModel.xlsx
Download the working financial model
The workbook contains five integrated tabs: Unit Economics, Balance Sheet, Profit & Loss, DCF, and Business Valuation across 12 quarters.