DUE DILIGENCE

What Investors Actually Look For in Due Diligence Before Writing a Check

Published August 23, 2026 · LaunchCode Editorial

Due diligence is the point in a fundraise where an investor stops evaluating a pitch and starts verifying it. It typically happens after a term sheet is signed and before funds actually move, when the investor confirms that the claims made during pitching hold up, and that nothing material was left out that would change the terms of the deal. Knowing what investors actually check helps founders prepare for it instead of being caught off guard mid-process.

Corporate and legal structure

Investors review incorporation documents, confirm the capitalization table is accurate and complete, including every SAFE, note, and option outstanding, and check that intellectual property assignment agreements exist for every founder, employee, and contractor who has ever contributed to the product. This last point is one of the most common sources of diligence delay: if the company itself doesn't clearly own its own code and designs, that's a real problem, not a technicality. Investors also review any pending or threatened litigation and material contracts the company has signed.

Financials

Expect a close look at historical financial statements, burn rate and runway, revenue recognition practices, outstanding liabilities, and any existing debt. Investors are checking whether the numbers a founder presented in the pitch match what the underlying books actually show.

Team and hiring

Key employment agreements, the actual vesting status of founder equity (see the guide on how ESOP vesting works for the mechanics), and any non-competes or prior-employer IP claims that could affect ownership of the startup's core work all get reviewed here. A founder who built an early prototype while still employed elsewhere, without a clear release from that employer, can create a genuine ownership problem that surfaces during this stage.

Product and technology

Investors typically want to understand how the product actually works, not just how it's marketed: a technical architecture review, security practices, and the company's dependency on third-party platforms or providers all factor in here.

Market and customers

This includes real usage data rather than projected usage, actual customer contracts, concentration risk (how much revenue depends on a small number of customers), and churn. Investors are looking for the gap, if any, between the growth story in the pitch deck and what the underlying data shows.

What's different for AI and ML startups

Diligence on an AI-native company covers everything above, plus a set of questions specific to how the product is actually built:

• Model IP and provenance. Is the core model built in-house, fine-tuned from an open or licensed foundation model, or largely a wrapper around a third-party API? This shapes both defensibility and any licensing obligations the company carries forward.
• Data rights. Where did training and fine-tuning data come from, does the company have a clear legal right to use it commercially (licensed, properly consented user-generated data, purchased, or public), and does any of it carry restrictions that limit commercial use or redistribution?
• Compute costs and unit economics. How do inference and training costs scale with usage, and do gross margins hold up as the customer base grows? AI-heavy products can carry a meaningfully different cost structure than traditional software, and that shows up directly in margins.
• Model evaluation. How does the company actually measure whether the model is working, through benchmarks, held-out evaluation sets, or human review, rather than relying on demos alone?
• Vendor and platform dependency. How reliant is the product on a single foundation model provider or cloud vendor, and what happens if that provider's pricing, availability, or terms change?

How founders can prepare

Keep an organized, current data room so documents don't need to be assembled under time pressure. Get IP assignment agreements signed early, for everyone who has ever touched the code. Keep the cap table reconciled after every SAFE, note, or option grant rather than reconstructing it later, since inconsistencies here are one of the fastest ways to slow a deal down or change its terms. And be ready to walk through unit economics with real numbers, not projections dressed up as facts, since that gap is exactly what diligence is designed to find, and it directly affects how dilution and future rounds get modeled once the deal actually closes.