Where there is still room to build in AI

Everyone building in AI right now is answering the same question in private: is this space already over? The honest answer is that it depends entirely on which space you mean. Across the 29 markets we track, the distance between the most open and the most contested is wider than most people building in the middle of it realise.

What “room” means, precisely

Two scores drive everything here. Opportunity asks how much unmet demand a tool addresses relative to what already exists in its space — high when a product is solving something people are visibly struggling with and few others have solved well. Saturation is a property of the category, not the product: how contested the space already is, how many credible entrants are fighting over the same buyer.

Subtract one from the other and you get the number this piece is built on. We call it room. A market with opportunity 82 and saturation 11 has 71 points of room, which in plain terms means demand is running well ahead of supply. A market with negative room means the opposite: more people are building than the demand currently supports. Both scores, and the heat score that blends them, are documented in full on our methodology page, including the weights and the decay we apply to stale signals.

The shape of that chart is the finding. There is no dense middle — markets cluster either in the open top-left or along a crowded diagonal, and very little sits between them. That matters for anyone choosing where to start, because it means the decision is closer to binary than gradual. You are either entering a market that has room or one that does not.

The six markets with the widest gap

These six clear 60 points of room. Read the two bars in each row as supply and demand: the top bar is what the market is worth solving, the bottom is how hard the fight for it currently is.

Notice how small these markets are. Healthcare is the largest of the six at 137 live tools, and Gaming the smallest at 57. Against a market like Coding — 422 live tools and 16 points of room — the open categories look almost empty. That is the trade the data is describing: room exists where fewer people went, and fewer people went because these markets are harder to enter.

Healthcare: the widest gap in the catalog

Healthcare scores 82 on opportunity against 11 on saturation — 71 points of room, the widest of any market we track. Across 137 live tools we found zero that we classify as a thin wrapper on someone else’s model. That is not a coincidence. A clinical documentation product has to survive a compliance review, integrate with an electronic health record, and satisfy a buyer who is personally liable if it is wrong. None of that is solved by wrapping an API.

The consequence for builders is specific: the moat in Healthcare is distribution and approval, not intelligence. The four tools below all score at or above 88 on opportunity, and each one earned it by taking on the non-model work — coordination across a stroke care team, prior-authorisation logic for an insurer, or protein structure prediction released openly to the research community.

Legal: 83 tools, no wrappers, and a buyer who pays

Legal is the second-widest gap at 69 points — opportunity 80, saturation 11 — and it has the clearest commercial logic of the six. Law firms bill by the hour, which makes time saved directly and unambiguously convertible into money. That is rare. Most AI buyers have to argue their way to an ROI number; a firm can simply count hours.

Yet only 83 live tools compete for it, and again none of them are wrappers. The barrier is the same shape as in Healthcare: a legal product that hallucinates a citation is not a slightly worse product, it is a liability. The tools that have won here invested in retrieval over verified case law and in workflows partners already trust.

3D, Gaming, Real Estate and Research

The remaining four open markets share a trait that is easy to miss: in each one, the output has to be usable by a specific pipeline, not merely impressive. A generated 3D asset that arrives with broken topology cannot go into a game engine. A generated listing description that misstates square footage is a regulatory problem. Passing that bar is what keeps saturation low.

3D & AR/VR (77 opportunity, 13 saturation, 68 live tools) and Gaming (78 / 14, 57 tools) are the two most technically demanding. Real Estate (76 / 13, 82 tools) is the most distribution-bound — the product is easy, reaching agents is not. Research (76 / 14, 87 tools) sits closest to becoming contested, because more of its value can be delivered by a good model alone.

Where the gap has already closed

One market in the catalog is underwater. Writing scores 40 on opportunity against 64 on saturation: 351 live tools chasing demand that no longer justifies them. Image and Social are effectively level, which is its own kind of warning — a market at parity today is a market that goes negative on the next cohort of entrants.

Wrapper density explains a great deal of the difference between the two ends of this chart. We flag a tool as a wrapper when its product is a thin interface over a general-purpose model with little proprietary data, workflow depth or distribution of its own. Every market that clears 10% wrappers sits in the contested half of the catalog.

Support is the densest at 18.7%, with Writing just behind at 17.7% and the largest absolute count at 62 tools. At the other end, Healthcare, Legal, Real Estate, Avatars and Human Resources each register zero. Image is the instructive exception: 392 live tools but only 2.6% wrappers, because generating an image well still requires owning a model. It is crowded for a different reason — too many people did the hard thing.

If you are deciding whether a market is genuinely open, wrapper density is the fastest proxy available. It tells you whether the hard part of the problem is the model, and if it is, someone will commoditise you.

How to use this if you are building or investing

If you are building. A wide gap is not an invitation, it is a description of a barrier you will also have to clear. Healthcare has room because clinical validation is slow and expensive; that cost does not disappear because you noticed the opportunity. The useful read is the inverse one: if you already hold domain access — a clinical network, a panel of firms, a brokerage relationship — these six markets are where that asset is worth the most, and it is worth considerably more than model quality.

If you are investing. Wrapper density is the cheapest diligence signal on this page. A company competing in an 18% wrapper market needs a story about what it owns that its model provider does not. A company in a 0% wrapper market usually has that story structurally, because it could not have shipped without one.

If you are running product in a crowded market. The 351 Writing tools are not all going to survive, and pricing is where that resolves first. The exit from a contested market is almost always downward into a workflow — owning a step nobody else owns — rather than upward into more features.

Every market, scored

The complete table, sorted by room. Tool counts are live tools only; products we have confirmed as shut down or acquired are excluded from both the counts and the averages.

Photo: Airam Dato-on / Pexels. Colour-graded for Falcoscan.

← All articles