What to Look for in Backing the Pivot to AI-Native Startups

How founders earn capital and customers when the old SaaS playbook no longer applies

A different diligence problem

Part one of this series looked at how AI-native pricing is displacing the SaaS subscription model, and what that means for boards. This second part turns to the people who fund and mentor the companies making that pivot. If usage-based and outcome-based pricing genuinely changes the unit economics of software, it changes what a credible investment thesis looks like too. The metrics that made a SaaS pitch fundable, seat growth, net revenue retention, a clean 3x rule of thumb, do not translate cleanly onto a business whose costs move with inference rather than headcount.

Angels and mentors are often the first check on that thesis, well before an institutional term sheet is on the table. Getting the framework right at that stage matters, because a startup that raises on SaaS-era assumptions and only discovers the mismatch at Series A has already wasted a funding cycle.

What credible traction looks like now

Growth numbers that impress on a slide can mask a business that loses money on every customer it signs. For AI-native companies, the diligence conversation has to go past top-line revenue and into the mechanics underneath it.

Gross margin at scale matters more here than in traditional software, because compute and inference costs do not shrink the way software delivery costs did. A startup running well below 70 percent gross margin because every customer interaction calls an external model is not describing a temporary inefficiency, it is describing its cost structure. Mentors and early investors should be asking founders to show a credible path to margin expansion, not simply revenue growth, and to explain whether that path depends on falling model prices, proprietary efficiency gains, or genuine pricing power with customers.

Customer economics need the same scrutiny. A minimum three-to-one ratio of lifetime value to customer acquisition cost, with payback inside twelve months for business customers, is a reasonable bar to hold founders to. If acquisition cost is climbing while lifetime value stays flat, that is a scaling problem hiding behind a growth number, and it tends to surface only once a company has spent the round finding out.

The team question has changed shape

Domain expertise now sits alongside technical depth as a genuine investment criterion, not a nice-to-have. Investors are looking for founding teams that pair someone who has actually built and shipped AI systems with someone who understands the buyer’s world well enough to sell into it, whether that is healthcare, legal, financial services or logistics. A team that can talk fluently about model architecture but not about how a target customer actually buys software is a common and avoidable failure pattern.

Model dependency is worth probing directly. A startup built entirely on a single external model provider carries a structural risk that a startup with a portable, provider-agnostic architecture does not. This is no longer a technical curiosity, it is a question mentors should be raising in the first substantive conversation, because it affects both margin and negotiating leverage down the track.

How capital-raising itself is changing

Investor targeting has become more specialised. Generalist funds with an AI thesis are common, but far fewer have the enterprise experience to judge whether a product will survive a security review, clear procurement, and convert from pilot to signed contract, the points where AI-native enterprise deals most often stall. Mentors advising founders on fundraising should be pointing them toward investors who have done this before in the founder’s specific category, rather than toward whoever currently has the most visible AI thesis on their website.

Non-dilutive capital, government grants and innovation funds, is playing a larger role than it did in the SaaS era. For a founder, it preserves equity while providing a credibility signal to later-stage investors. For a mentor, it is worth raising early and often, because founders chasing their first institutional round rarely think to look for it unprompted.

At the earliest stages, it is increasingly operator-angels and micro-VCs who move fastest and understand the category well enough to write a check quickly, while larger funds have generally migrated toward later-stage, de-risked opportunities. Founders and mentors alike should calibrate expectations about who is available to write the first cheque.

What to ask before writing the cheque

A short, consistent set of questions travels well across most AI-native pitches: what does gross margin look like once the business is at meaningful scale, not just today; what happens to unit economics if a key model provider changes its pricing; how much of current traction depends on a pilot that has not yet converted to a contract; and does the founding team have the domain credibility to close enterprise deals, not just the technical credibility to build the product.

None of this makes AI-native businesses a worse bet than SaaS was. It makes them a differently shaped one. Mentors and investors who bring the old checklist unmodified to this category will either miss good companies that do not fit the old pattern, or back companies whose economics were never going to hold up once the novelty wore off. The diligence must change because the business model genuinely has.

About: Gary Morgan is a director, board advisor and principal consultant at MPT Innovation Group, specialising in governance, technology strategy, and organisational transformation for private and not-for-profit organisations. He is a Fellow and Member of the Queensland State Council of the Governance Institute of Australia, and an Adjunct Fellow and Member of the Griffith University Industry Advisory Board for the ICT School. Gary publishes regularly on board governance, AI, technology, and cybersecurity.

Acknowledgment: This article represents the author’s independent views and incorporates AI-assisted research and drafting.