AI startups don't make money the way SaaS did.

For a decade, the playbook was the same. Build a SaaS product. Charge users. Hit $1M ARR. Raise on revenue multiples. Repeat. The CAC and LTV math was well-modeled, the funnel was predictable, the rounds were priced.

AI economics broke that. The network is the asset and the buyer is somewhere else.

In the AI network plays we operate at Innovative Group, the founders we acquire in Year 1 are not the customer. They are the supply side of a two-sided network whose Year 2 revenue comes from third parties, data buyers paying for intent signal, enterprise subscribers buying access to the data layer, ESO advertisers paying to reach founders, retail consumer products charged at low ARPU through the IdeaMaker-style funnel.

Here is the math from a current Customer Zero engagement, the AI funding platform that just raised $3.5M at $20M valuation with a two-step performance ratchet.

Imagine you charge founders $79/month for a Pro tier in Year 1. Every founder you convert at $79 is a founder who did not refer three others because they are now rationing. Every subscription tier you push is friction on the supply side of a two-sided platform that needs both sides to scale.

The math gets worse, not better, with Year 1 monetization on AI network plays. The right move is to keep the network free in Year 1, build the supply and demand sides to scale, and turn on third-party monetization in Year 2 once the network has the density to support it.

SaaS asks one question: how much does the user pay per month? Network-economics businesses ask seven. Each layer has a different timing, a different audience, and a different unit economic.

Patient capital with milestone-based ownership unlock is the structural innovation behind some of the best AI startups of 2026. Here's how it works in practice.

At raise, the investor takes a notional 30% to 40% of the cap table. Two milestones are agreed in writing: a user growth milestone (typically 100K to 250K active users) and a revenue milestone (typically the first $1M to $5M in annualized network revenue).

When milestone one is hit, the founders earn back the first ratchet, reducing the investor's share to the mid-twenties. When milestone two is hit, the second ratchet kicks in, leaving the investor at 15% to 20% and the founders holding the majority of the company.

The structure works because patient capital wants to own a meaningful piece of a successful business. They don't need to lock in their share at the priced round if the milestones are credible. For founders, the structure trades upfront ownership protection for execution risk on milestones they would have to hit anyway.

Traditional venture math compounds returns on investor ownership. The bigger the slice taken at seed, the better the IRR if the company exits at $1B+. This math works for SaaS companies where the funnel is predictable and seat-based ARR scales linearly with marketing spend.

It does not work for network-economics AI startups. The first 18 to 24 months are network-building, not revenue-building. Heavy investor dilution at seed produces founder economics that incentivize an early exit rather than the long arc the network play requires. The two-step ratchet aligns investor returns with milestone-based value creation, which is the value creation pattern the business actually has.

For the 4 to 6 AI startups per year that fit this profile, the math is structurally better for everyone. The investor still ends up with a top-decile return because the company they're backing actually compounds. The founders stay motivated and stay running the business. The capital is patient enough to let the network economics unfold on their natural timeline.