Start with the business and the security you are buying
This is Kooya’s educational investment framework, not an offer, a recommended deal or an assessment of a particular issuer. Kooya develops AI systems and therefore has a commercial interest in AI adoption. Financing a startup means accepting ownership or contractual investment risk; purchasing its software is a separate decision. No target return or personal allocation is proposed here.
Identify the entity receiving the money, its jurisdiction, the instrument, the bank-account beneficiary and the evidence establishing the offer’s legal basis. Ask qualified counsel to review the actual terms and any applicable registration or exemption. Do not treat incorporation, a demo-day invitation or another investor’s participation as sufficient evidence. For a cross-border investment, include tax, currency, transfer restrictions and dispute enforcement in that review.
The U.S. SEC’s private-placement bulletin warns of total loss, limited disclosure and difficulty reselling private securities. Those are useful risk questions, not a statement that U.S. Regulation D governs a Philippine offering. Decide whether capital could remain unavailable indefinitely before spending time on the product narrative.[1]
Reconcile revenue with contracts, usage and cash
Kooya analysis: request a customer-level schedule that separates free trials, paid experiments, implementation fees, recurring subscriptions and usage charges. Trace a sample from signed contract to invoice, collection and product activity. A slide labelled annual recurring revenue needs a written definition, especially when short pilots or one-time services are included.
Read renewal, cancellation and refund clauses. Compare successive customer cohorts and ask which accounts continue paying after founder-led implementation ends. Identify revenue from related parties or a small number of customers. A strong reference can support one relationship; it cannot establish retention across the business.
For a pre-revenue startup, label evidence honestly: interviews, letters of intent and pilot agreements test different assumptions. Build the investment case around the next falsifiable milestone and the cash needed to reach it. Do not annualize a pipeline as though it were contracted recurring income.
- Request a reconciliation of reported revenue, invoices and bank receipts.
- Separate recurring product income from bespoke delivery work.
- Ask for customer concentration and renewal evidence, with permission to validate references.
Calculate economics after the work required to deliver
Kooya analysis: begin with collected customer revenue and deduct the costs directly needed to serve that customer: model and tool calls, hosting, retrieval, retries, human checks and support. Call the result a contribution calculation and state its inclusions. Do not silently compare that measure with another company’s differently defined gross margin.
Ask for production traces and invoices for representative workloads. A cached demo can conceal costs from long documents, difficult cases, repeated agent actions and escalations. Test what happens when discounts or cloud credits end, usage rises, and a higher-quality model is needed to meet the customer’s promised result.
Use explicit scenarios rather than a single forecast: base usage, heavier usage, lower selling prices and slower collections. For peso revenue with foreign-currency infrastructure bills, test exchange-rate sensitivity. Calculate cash runway using unrestricted cash and a stated net-burn assumption; rerun it after planned hires and commitments rather than assuming today’s burn stays constant.
Test model dependency and evidence of defensibility
A startup can create useful value on a third-party model. The investment question is what it owns or controls beyond access: a difficult workflow, distribution, integrations, permitted proprietary data, evaluation systems or sustained customer relationships. Require a demonstration of that advantage and the resources needed to maintain it.
As one concrete supplier example, Anthropic’s commercial terms link fees to its pricing terms and allow published rates to change with notice. They also place responsibility on customers to assess output suitability. These provisions illustrate why a startup’s supplier agreement matters to its economics and operating obligations; they do not describe every provider’s contract.[3]
Kooya analysis: ask the team to run the same representative tasks on its current model and a credible fallback. Record quality, latency, cost, integration work and failure behavior. A diagram with several providers is not proof that switching works. Include the risk that a supplier adds a competing feature or access becomes unavailable.
Use our AI pilot evaluation checklist to structure the technical evidence behind the investment case.
Inspect data rights before treating data as an asset
The Philippine NPC’s Advisory No. 2024-04 applies data-protection obligations to AI development and deployment involving personal data, including training and testing. It addresses lawful processing, transparency, accountability and security. Its scope makes data governance a diligence concern throughout the product lifecycle, not merely at customer onboarding.[4]
Kooya analysis: request an inventory connecting each material dataset to its origin, permitted purpose, contract or licence, retention policy and deletion process. Separate a right to process customer data for service delivery from a claimed right to train a reusable model or sell derived products. Have specialists assess those rights against the actual agreements and applicable law.
Review ownership assignments from founders, employees and contractors, plus dependencies on open-source code, model weights and third-party content. Ask what must be removed if a customer withdraws access or a licence ends. The relevant evidence is a traceable rights record and tested controls, not an unsupported claim that all training data is public.
Model dilution, control and the absence of an exit
Y Combinator describes its SAFE as a contract for future equity, with no maturity date; it is not stock before conversion. Its documentation distinguishes valuation caps, discounts and optional pro rata rights. These are descriptions of YC instruments, not a substitute for reading a proposed agreement or confirming its suitability in the issuer’s jurisdiction.[2]
Kooya analysis: obtain the fully diluted capitalization table, outstanding convertible instruments, option commitments and investor side letters. Model a future financing at several valuations, an expanded employee pool and any participation rights. Compare your resulting ownership and distribution priority, not just the headline cap.
Ask who approves major decisions, which information you will receive and what happens on a sale, shutdown or new financing. Read transfer restrictions and any redemption provisions carefully. A planned acquisition or stock-market listing is a scenario, not a repayment schedule. Maintain a no-exit case alongside every attractive outcome.
Turn the findings into an investment decision record
Use a short evidence register: claim, supporting document, unresolved question, owner and decision impact. Distinguish verified facts from management forecasts and your own assumptions. Missing customer evidence, unclear data rights or an unexplained capitalization table should remain visible rather than disappearing into an averaged score.
Kooya analysis: a useful conclusion can be decline, wait for a milestone, or proceed to specialist review under stated conditions. Product evaluation is only one input. A technically strong system can still be an unsuitable investment because of valuation, financing terms, concentration or illiquidity. Record what would change your decision before negotiating the price.