What gets VC funding now?

A few years ago, a founder could walk into a pitch meeting with a good idea, a tidy deck, a few product mockups and, if they were ahead of the pack, a working demo.

Pair that with a large, boring workflow, an underserved market and a story about how automation could change the economics of the category, and investors would usually lean forward.

Recruitment. Outbound sales. Legal work. Customer support. Logistics. Back-office admin. Anywhere you found repetitive, high-volume knowledge work, especially across a mess of disconnected systems, there was usually a startup being formed around it.

The pitch was simple enough: this work is expensive, manual and dull, so let’s use AI to automate it.

For a while, that was enough to get a serious conversation going. A founder might raise a million dollars on the back of a smart idea, a few promising customer conversations and a believable vision of the product. They would then disappear for six months, hire a small team and build the first proper version. If the beta customers liked it, if a bit of revenue started to appear, and if the market still looked big enough, they might be on track for a good seed round.

That basic pattern has shaped early-stage investing for a while. In some categories, having the money and team to spend six months building the thing was part of the moat. It didn’t keep everyone else out, but it filtered out a lot of casual competition. Plenty of people had the same idea. Far fewer had the capital, technical team and stamina to turn it into a polished product.

That has changed.

Over the last few years, investors have been pitched countless AI recruiters, automated SDRs, legal assistants, agentic teammates and workflow automation tools. The issue is no longer just volume. The average pitch has got much more sophisticated. Eighteen months ago, many teams were still pitching a vision, a deck and maybe a rough prototype. Now founders are turning up with working products they have vibe-coded in their spare time, plus a handful of beta users or design partners.

That shifts the fundraising bar.

If a competent founder can build a reasonable version of your product in a few weekends, the idea itself is not much of a moat. Nor is the demo. Nor, in many cases, is the fact that AI can automate the workflow. That may be true, but it is also obvious.

So investors have become more sceptical. If you’re pitching an AI recruiter, assume the investor has already seen twenty. If you’re pitching an AI SDR, assume they’ve seen even more. If you’re pitching an AI legal assistant, assume they have invested in one, passed on five and watched a couple more struggle to break through.

That doesn’t mean these categories are dead. Good companies will still be built in crowded markets. But the burden of proof is much higher. You can’t just point at a tedious workflow and say “AI will fix this.” You need to show why this wedge, why this buyer, why this team, why this moment, and why this becomes a company rather than a feature inside somebody else’s product.

That last distinction is where a lot of AI founders get into trouble.

There are plenty of useful AI features. Tools that summarise, classify, draft, search, route, generate or automate. Some are genuinely handy. Some save time. Some are impressive the first time you use them. But useful is not the same as fundable.

A venture-backed company needs a chance of becoming large, defensible and hard to displace. If the thing you’ve built is likely to become a button inside Salesforce, Workday, Figma, Adobe, Microsoft, Google or an existing vertical SaaS platform, that doesn’t make it a bad idea. It may be a very good idea. It just may not make it a venture scale business.

Investors are trying to work out which side of that line you’re on.

The bar has also gone up because building has become cheaper. A few years ago, a non-technical founder with an idea usually needed to find a CTO, pay an agency, convince an engineer friend to help, or raise a small pre-seed round just to get something working. Now a designer, teacher, consultant, operator or industry specialist can get surprisingly far on their own. They may not be able to build the final production system, but they can create a convincing demo, test a workflow, put something in front of users and learn whether anyone cares.

That changes the investor conversation. A deck is less persuasive than it used to be. A clever idea is less persuasive than it used to be. Even a basic demo is becoming table stakes in many categories. If AI makes it possible to build something tangible in a few weeks, investors will increasingly expect founders to have done that work before they raise.

This is especially interesting for design founders and other non-technical founders. Designers have often had good instincts around user needs, workflow, friction, behaviour and adoption, but they lacked the ability to ship without engineering support. AI has narrowed that gap. It gives people with taste, domain insight and customer understanding a way to make their ideas real much earlier.

That doesn’t make every designer an engineer. But it does make “I need funding before I can test this” a much weaker argument. If you can’t be bothered to prototype the thing with today’s tools, that tells investors something. Not a good thing.

So what is getting funded?

Infrastructure is still attractive, if the team has the technical depth to match the ambition. The picks-and-shovels argument still holds. In AI, that means developer tooling, evaluation, observability, deployment, security, data pipelines, compute optimisation, agent reliability and the other unglamorous problems sitting beneath the application layer. These companies can benefit from the wider shift rather than betting everything on a single workflow. But a thin wrapper around an API is not infrastructure.

Application-layer companies can still get funded too, but they need a sharper story. Either they have breakout traction, a specific wedge into a valuable market, or a team with unusually strong founder-market fit. “AI for finance” is not enough. “AI for healthcare” is not enough. “AI for legal” is definitely not enough. The more crowded the category, the more precise the wedge needs to be.

Some verticals are clearly getting attention. Defence is hot for obvious geopolitical reasons. Finance and insurance remain strong in the UK because there are budgets, complex workflows and a long history of buying software. Healthcare is enormous, but anything that relies on selling into the NHS needs to be treated with care. Education depends heavily on whether you are selling to individuals, companies or schools. Creative tools are exploding, partly because AI has unsettled some of the old assumptions about where design and production work happen.

But the category matters less than the evidence.

The best founders don’t just say “this market is big.” They make you feel that they understand the customer better than the room does. They can explain the pain in plain English. They know who owns the budget. They know what people use today, even if it is a spreadsheet, a WhatsApp group or a junior employee doing repetitive work badly. They know why the buyer would switch. They know what trust needs to be built before the product is adopted. They can explain how they’ll get their first ten customers, then their first hundred.

That last bit matters more than many founders realise.

AI has made building faster. It has not made distribution easier. In some ways, it has made distribution harder. Every buyer is being pitched AI products. Every LinkedIn feed is full of AI demos. Every category has become noisier. Paid acquisition is expensive. Outbound is saturated. Social media channels have collapsed. Content is harder to make distinctive. Communities are tired of being sold to.

So you get this odd situation where a team can build a credible first version in six weeks, then spend eighteen months struggling to work out how to make anyone care.

This is where go-to-market becomes the real moat. In a world where almost anyone can build a passable version of an idea quickly, the advantage shifts to the team that can reach customers fastest, onboard them cleanly, learn from them quickly and turn early usage into momentum before the next wave of competitors arrives. It is not enough to ship. You have to create adoption faster than the market can copy you.

That is where a lot of startups fail. Not because the product is impossible to build, but because the founders mistake product progress for business progress.

This is basically why I wrote The Growth Equation. Most early-stage founders are good at building something they think should exist. Far fewer are good at building something enough people care about, can discover, understand, trust, buy and keep using.

That is the real funding bar now. The demo still matters. It just doesn’t protect you for very long.