The AI Everywhere Report 2026: How AI Became a Layer, Not a Category

Two years ago AI was a category you could point to. Now it is a line in almost every product's pitch. We measured how far that has gone by checking how each tool in our catalog describes itself, and the answer is that AI has stopped being a place on the map and become a layer underneath all of it. Nearly half of B2B software now reaches for AI language, and the categories where it shows up most have nothing obviously to do with artificial intelligence.

This is the positioning companion to our State of AI Adoption report. That one counts AI-first products. This one measures something different and broader: how many tools, in any category, now describe themselves with AI, because that is what buyers are being sold.

Methodology

We flagged a tool as AI-described if its tagline or description references AI as a standalone word, or the phrases artificial intelligence, machine learning, GPT, or LLM (9,960 published tools across 401 categories as of August 4, 2026, extracted from vendor positioning). This is deliberately a positioning metric, not a capability audit: it captures how a tool is sold, which is distinct from our adoption report's stricter count of AI-first products. Category figures exclude the explicitly-AI categories and use categories with at least 30 tools, so the numbers below show AI reaching into spaces that are not about AI on their face.

Where AI has spread the furthest

These are non-AI categories, ranked by the share of their tools that now describe themselves with AI.

Read the middle of that list, not the top. Transcription and NLP being AI-heavy is expected; they were machine-learning problems before the current wave. The interesting rows are Legal Document Automation at 80%, Revenue Intelligence at 87%, and Pricing Optimization at 72%. These are unglamorous back-office and revenue-operations categories, and they are now overwhelmingly sold as AI. That is what "AI as a layer" looks like in the data: it shows up strongest not in the tools built to be AI, but in the ordinary workflow tools that added it.

The 47% number, in context

Nearly half of all tools reaching for AI language does not mean half of all software is genuinely AI-native. Positioning runs ahead of capability, always. But positioning is also a leading indicator: vendors adopt the language of AI because they believe buyers now expect it, and they are usually right about what buyers expect. When AI moves from a selling point to an assumed baseline, the tools that do not mention it start to look dated rather than focused.

For buyers, that flips the useful question. Two years ago the question was "does this have AI." Now that 47% of the market claims it, the question is "what does the AI actually do here, and can I turn it off," because the label alone no longer separates anything.

What it means

If you are evaluating tools in 2026, treat AI in the pitch as table stakes, not a feature. The signal that matters has moved downstream: ask which specific task the model performs, whether the output is verifiable, and what happens when it is wrong. In the categories above, where two-thirds to nine-tenths of tools claim AI, the claim tells you almost nothing on its own. The demo, the failure modes, and the price are what separate them.

We refresh this report quarterly. To pull how any tool positions itself directly into your own agent, see Toolradar for AI Agents.