A product lead at a mid-market SaaS company put it plainly in a recent planning meeting: marketing had quietly turned on an AI personalization layer that rewrote onboarding copy, reordered feature callouts, and swapped hero images per visitor segment, and nobody in design had seen a single variant before it went live. The tool worked exactly as advertised. It just did not know the product had a voice, a visual system, or a set of promises the rest of the experience had to keep.
Aero sees versions of this across every industry it works in: a fintech dashboard that personalizes which metrics surface first, a biotech portal that tailors onboarding content by research area, a professional services site that rewrites its own case studies per visitor, a media brand generating different headline variants for different readers. The personalization engine is usually a marketing or growth decision. The consequences land on product and design, because the visitor experiencing all of it assumes it is one coherent product, not a set of independently generated variants.
This is not a fringe concern. Gartner predicts that 60 percent of brands will use agentic AI to deliver one to one personalized interactions by 2028, a sharp jump from where most organizations stand today, according to Gartner's own research. That means most product and design teams have two or three years, not ten, to figure out how personalization at that scale stays recognizably on-brand instead of turning into a different experience for every visitor.
The starting point is not encouraging. In a Gartner consumer survey, 49 percent of U.S. consumers said generative AI has made content quality worse overall, not better, according to Gartner's own survey findings. That is the backdrop every team scaling AI personalization is building against: a public that already associates AI-generated content with a drop in quality, not a lift in relevance. Personalization that is technically accurate but visually or tonally off-brand does not read to a visitor as "tailored for me." It reads as "something is a little wrong here," even when they cannot say exactly what.
Two B2B software companies, one selling analytics tools to operations teams and one selling scheduling software to professional services firms, both rolled out AI-driven onboarding personalization this year, tailoring the first-run checklist and empty states to each new account's stated use case. The analytics company let the personalization tool generate its own copy and choose its own layout for each variant, independent of the existing design system. Within a few months, new users were seeing checklists with inconsistent button styles, empty states that did not match the illustration set used everywhere else, and copy that occasionally contradicted the help docs. Support tickets about "is this the real product" started showing up in triage, an odd complaint for software that was, in fact, entirely real.
The scheduling company constrained its personalization layer to a fixed set of approved components and copy blocks, with the AI choosing which combination to show rather than generating new markup or wording from scratch. Every personalized variant looked, and read, like the same product, because structurally it was. Onboarding completion improved in both cases, since personalization itself works. Only one company kept its brand intact while it worked.
This is Aero's own practical lens for reviewing an AI personalization rollout before it reaches customers, not an established industry standard. Five questions worth running against any personalization engine your product or marketing team is using:
A team answering "not sure" to two or more of these is scaling personalization faster than it is governing it, which is exactly the gap the Gartner consumer data above suggests visitors are starting to notice.
Personalization guardrails are really a design system problem wearing a marketing hat. A design system that was never built with AI systems in mind will struggle to constrain a personalization engine the same way it struggles to constrain AI-assisted coding, the gap covered in is your design system agent ready. And once personalization capability is common across a category, it stops being the differentiator by itself, the same shift Aero has written about in AI is table stakes now. The teams that will stand out are not the ones with the most aggressive personalization. They are the ones whose personalized experience still feels like one product no matter which variant a visitor lands on.
No. The risk comes from letting a personalization engine generate freeform layout and copy without constraints, not from personalization itself. Constrained to an approved component and content library, AI personalization can scale without drifting off-brand.
No. B2B products personalizing dashboards, onboarding flows, or proposal content face the same risk. A biotech portal or a fintech dashboard showing inconsistent variants raises trust questions just as fast as a consumer retail site does.
Design and product, alongside whoever owns the personalization tool. Marketing or growth teams typically own the rollout, but without a design reviewer in the loop, nobody is checking whether the variants still look like the product they came from.
Wondering whether your AI personalization rollout is still on-brand once it scales past the demo? Talk to Aero.
Agentic AI is about to put one to one personalization inside every product experience. Here's Aero's framework for keeping that experience recognizably on-brand instead of visibly stitched together.
AI coding assistants are shipping features straight to production, but security review isn't catching design, accessibility, or brand drift. Here's Aero's framework for closing that gap before customers do.
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.
