Financial Projections for Pre-Seed and Seed in the AI Era: From Prototype to Proof

The first check should not fund the first thinking

A useful way to frame pre-seed today is that the first check should fund the next proof, not the first round of thinking. Before raising, founders should ideally understand the problem, the first customer segment, the basic market logic, the likely pricing direction, the first go-to-market path, and the first version of the product. They do not need perfect answers, but they should be able to show that the business has been thought through beyond the pitch deck.

The model should make this visible. If the founder has already built a prototype, the financial plan should show what happens next: product hardening, onboarding, design partners, the first sales motion, usage testing, paid conversion, or technical reliability. If the founder has customer conversations but no product, the model should show how those conversations become pilots or paid tests. If the founder has a waitlist, the model should show how waitlist users convert, activate, retain, and eventually pay.

This is also where the financial model protects the founder from confusing activity with progress. A startup can spend money on product, content, tools, events, contractors, and ads without becoming more fundable. The model should force the question: what evidence will this spending create?

Prototype is not proof

A prototype is useful because it makes the idea concrete. It can help founders test workflows, gather feedback, show the product to customers, and start investor conversations. But in the AI era, prototypes are easier to create, which means investors may discount them faster unless they are connected to customer evidence.

For B2B SaaS, useful proof may be a design partner, a pilot, a paid pilot, a strong pipeline of qualified conversations, or evidence that the buyer owns the problem and has budget. For a self-serve subscription product, useful proof may be activation, conversion, retention, repeat usage, or willingness to pay. For an AI product, useful proof may include not only user interest, but also cost-to-serve, gross margin, and whether heavy users can be served profitably.

The financial model should separate these types of evidence. A waitlist is not the same as activated users. Activated users are not the same as paid users. Paid users are not the same as profitable users. A pilot is not the same as a repeatable sales motion. Investor projections become more credible when the model shows these transitions clearly.

What milestones should your model connect to the round?

The right milestone depends on the type of product, business model, and stage. A technical product, a B2B SaaS company, a self-serve subscription product, and an AI-native product should not all use the same fundraising model. The budget, timeline, and proof points should match the kind of evidence the startup needs to create.

These examples are not templates to copy directly. They show the level of specificity that makes a fundraising model useful. For a B2B SaaS product, a model that only shows website traffic and self-serve conversion may miss the actual sales process. For an AI subscription product, a model that ignores free usage and API costs may hide the real margin risk. For a marketplace, a simple customer growth curve may be weak if it does not explain how supply and demand will be created together.

How market size should connect to the revenue model

Pitch decks often include a market size slide. That slide can be useful, but it does not explain how the startup will actually capture value. Market size shows the opportunity; the revenue model shows how the company makes money; the financial model should connect both.

The self-serve subscription path depends on traffic, activation, paid conversion, churn, support load, and acquisition efficiency. The B2B SaaS path depends on pipeline, sales capacity, pilot-to-paid conversion, contract timing, onboarding effort, and customer concentration. Top-down market sizing is not enough for investor projections. A founder can show that the market is large, but the model still needs to show what pricing, customer count, acquisition, conversion, churn, and cost assumptions would make the revenue path possible. Use the market slide to show ambition. Use the financial model to show operating logic.

For the full framework, see the startup financial modeling guide and the article on top-down vs bottom-up startup forecasting.

AI makes cost-to-serve visible earlier

One of the biggest differences in AI-era financial projections is the cost side. A SaaS product can often grow users before direct serving costs become painful. An AI product can feel margin pressure much earlier if usage is expensive, free users are active, or heavy users consume more than the pricing model supports.

This is why AI startup projections should show cost-to-serve earlier than a classic pre-seed model might have done. The model should not wait until later-stage planning to include API costs, inference costs, usage caps, free usage, trial limits, plan-level usage, and gross margin. Bessemer's State of AI 2025 makes this point from the other side of the market: some AI companies can grow revenue extremely quickly, but top-line ARR does not automatically mean a healthy business. Margins, retention, engagement, and capital efficiency still matter.

For an early-stage founder, the practical question is whether the company can afford the usage if users love the product. The model should help answer that before the startup spends heavily on acquisition or gives away too much free access. See the AI/API cost forecast and cost of revenue and gross margin articles for how to model this in detail.

What should be in the pitch deck and what should stay in the full model?

Founders often try to put too much of the model into the pitch deck. The financial slide in the deck should not become a spreadsheet screenshot. It should show the few numbers that support the fundraising story. The full model should sit behind the deck and help the founder answer follow-up questions.

The pitch deck usually connects several slides: market size, business model, go-to-market strategy, traction, financial projections, use of funds, milestones, and fundraising ask. These slides should not tell separate stories. They should be connected by the same financial logic. If the deck says the startup is targeting enterprise customers, the model should not assume instant self-serve conversion. If the deck says growth will come from paid acquisition, the model should show CAC, conversion, spend timing, and payback logic. If the deck says the product is AI-powered, the model should show AI/API usage costs and gross margin pressure.

The deck should make the plan understandable. The model should make it defensible.

The first 6–12 months should be especially clear

For early-stage investors, the first months often matter more than the long-range revenue curve. This is where the model becomes operational. This is where the investor can see what happens immediately after the money arrives: when product work starts, which roles or contractors are needed, when acquisition begins, when pilots or customer conversations are expected, how long it takes before revenue appears, which costs are fixed from the beginning, which costs depend on usage or customer growth, and how much buffer the company has if milestones take longer.

A 36-month forecast may be useful, especially at seed. But if the first 6–12 months are vague, the model will feel weak. In the AI era, this is even more important because founders may already have a prototype before raising. The first months after the round should show what changes because of the capital.

For a technical product, that may mean moving from prototype to reliable product. For a B2B product, it may mean converting customer conversations into pilots or paid pilots. For a self-serve product, it may mean testing activation, paid conversion, and retention. For an AI product, it may mean proving that users want the product and that usage does not destroy gross margin. Cash flow, hiring, product timing, GTM activity, and milestones should be visible together before the founder is already spending the money.

What the model should include

A pre-seed or seed model does not need to be complicated for the sake of looking sophisticated. It should connect the main business drivers well enough to explain how the company could move from early evidence to stronger proof, and from stronger proof to the next stage. The core sections are usually revenue, acquisition, costs, team, cash flow, runway, and unit economics.

What makes AI-era projections credible?

Credibility does not come from adding more tabs or making the forecast look precise. A model becomes credible when the assumptions are connected, stage-appropriate, and explainable. A credible model starts from evidence, is driver-based, connects market ambition to bottom-up logic, ties the round to milestones, matches the stage, and stays cost-aware and scenario-aware. The final test is explainability: the founder should be able to explain why revenue grows, why the hiring plan starts when it does, why usage costs stay under control, and why the raise amount is enough.