Eighteen months ago, "we have an AI feature" was a sales slide on its own. A VC diligence call, a fintech RFP, a biotech lab-software demo: naming the AI feature was often enough to earn a second meeting. That window has closed. A prospect evaluating a SaaS tool, a professional services platform, or a biotech data product today assumes AI is already in there somewhere. The question is no longer "does it have AI," it is "what does the AI actually do for me, and does using it feel like using the rest of the product."
That shift shows up in how analysts are framing the market. Gartner projects that 40 percent of enterprise applications will feature task-specific AI agents by the end of 2026, up from less than 5 percent in 2025, according to Gartner's own research. When a capability moves from rare to default that fast, it stops being a reason to buy. It becomes a reason not to be disqualified.
McKinsey has started calling this directly: AI capability itself is becoming table stakes, and the firms pulling ahead are the ones building what McKinsey describes as a genuine competitive moat around it, not the capability alone, according to McKinsey's own analysis. For a product and design team, that moat is rarely the model. Most companies in a given category are drawing from the same handful of foundation models and the same class of agent frameworks. The moat is what happens around the model: how clearly the product explains what the AI is doing, how gracefully it handles the AI being wrong, how well it fits the rest of the product's visual and interaction language, and how much trust a user has left after using it three times.
This is the practical version of the McKinsey point. Two products can license comparable AI capability and land in completely different places with users, because the differentiation was never going to come from the model. It comes from whether the AI feature was designed as a first-class part of the product or bolted on as a checkbox. A chat panel dropped into the corner of an existing screen, disconnected from the product's design system and answering in a tone that does not match the rest of the brand, reads as exactly what it is: a feature added to satisfy a roadmap line item. An AI capability built into the actual workflow, using the product's existing components, failing gracefully, and explaining itself in the product's own voice, reads as the product simply getting smarter.
Two mid-market software companies in adjacent categories, one selling to operations teams at professional services firms and one selling to clinical research coordinators at biotech companies, both added an AI assistant that could answer natural-language questions about the user's own data this year, using comparable underlying models. The first shipped the assistant as a standalone chat window, separate from the main application, with its own visual style and its own way of showing sources. Users treated it as a side tool: helpful occasionally, ignored most of the time, and a common source of support tickets when its answers did not match the dashboard.
The second company took longer to ship. The assistant lived inside the existing workflow, answered using the product's own chart and table components, and every answer linked back to the exact records it drew from, in the same visual language the rest of the product used to show provenance. Adoption of the assistant tracked adoption of the product overall, because using it did not feel like switching tools. Support tickets about mismatched numbers were rare, because the assistant's answers used the same source-of-truth rendering as everything else. The underlying AI capability was not meaningfully different between the two companies. The outcome was.
This is Aero's own practical lens for auditing an AI feature before or after launch, not an established industry standard. Five questions worth running against any AI capability your product ships:
A team answering "not sure" to two or more of these is likely relying on the AI capability itself to carry weight it can no longer carry alone.
Differentiation through experience rather than capability is not a new idea in product design, but AI has made it urgent faster than most teams planned for. Getting users to actually adopt an AI feature once it exists is its own discipline, covered in why AI feature adoption is a behavior change problem. And as more products let a model assemble parts of the interface live, the design system becomes the mechanism that keeps an AI feature feeling native instead of bolted on, a shift walked through in our field guide to generative UI. Both point to the same conclusion: once AI capability stops being scarce, everything once dismissed as "just design" becomes the whole competitive argument.
It still matters as a baseline. A product that cannot do what competitors' AI features do will lose deals on that basis alone. The point is that clearing the baseline no longer wins the deal by itself, the way it often did in earlier stages of AI adoption.
Not more important, but more available as a place to differentiate, since most teams in a category have access to comparable models. The experience around the model is what a user actually judges the product on.
It tracks category-wide AI adoption closely. Once most credible competitors have shipped some version of the same capability, buyers stop treating it as a differentiator and start treating it as an expectation.
Wondering whether your AI feature is still doing the differentiation work you built it to do, or just checking a box? Talk to Aero.
Every product now has an AI feature, so having one has stopped being the pitch. Here's Aero's framework for finding real differentiation in the experience around the AI, not the AI itself.
AI platforms are starting to assemble interfaces on the fly instead of shipping fixed screens. Here's what generative UI actually means for product teams, and Aero's framework for using it without losing brand and accessibility control.
Usage-based and hybrid pricing are becoming the default for AI-powered products, but most pricing pages still assume a simple per-seat world. Here is Aero's practical framework for pricing pages that build trust instead of anxiety.
