\u26a1 Quick answer:<\/strong> The biggest gap in design today isn’t between fashionable and unfashionable \u2014 it’s between what gets discussed and what actually ships. 94.8% of the top million homepages fail accessibility checks<\/a>, even though everyone talks about inclusivity. Meanwhile 71% of users expect personalisation, and Gartner projects 30% of new applications will run AI-driven adaptive interfaces \u2014 up from under 5% two years ago. Innovation right now isn’t a visual device; it’s how far the interface adapts to the person using it.<\/p>\n The word “innovative” in design usually means “looks unusual.” In reality the things that genuinely change products are nearly invisible in a screenshot \u2014 they show up in how an interface behaves, not in its static image.<\/p>\n The numbers show how fast this moved: 93% of designers already use generative tools daily, and 73% consider AI collaboration the most influential factor of the year.<\/p>\n It’s worth separating two very different applications that often get conflated. The first is AI as a designer’s assistant: generating variants, suggesting palettes, drafting layouts. That speeds up work but changes the end product very little.<\/p>\n The second is more interesting \u2014 AI inside the product itself: an interface that reshapes around a specific user. Different starting screens for a newcomer and an experienced user, prompts appearing exactly where people usually hesitate, menu ordering that shifts by scenario.<\/p>\n That second kind is what Gartner means by 30% of applications running adaptive interfaces. Growth from 5% to 30% in two years isn’t a trend, it’s a change in the baseline.<\/p>\n The data here is unambiguous: 71% of users expect a personalised experience, and companies that do data-driven personalisation well earn 40% more<\/a> than those that don’t.<\/p>\n For a small business that sounds out of reach, but the basic levels need neither AI nor a large team:<\/p>\n None of that is innovation in the literal sense. It’s basic hygiene many sites still haven’t reached \u2014 which is exactly why implementing it produces a visible effect faster than anything more sophisticated.<\/p>\n That 94.8% failure figure matters for practical rather than moral reasons. Accessibility requirements overlap almost entirely with the requirements of good UX.<\/p>\n Sufficient text contrast isn’t only for people with visual impairments \u2014 it’s for anyone reading on a phone in sunlight. Clear form field labels help more than screen readers. Keyboard navigation rescues people who simply don’t like reaching for a mouse.<\/p>\n The minimum set that closes most typical problems: text contrast at or above the recommended ratio, labels on every form field, alternative text for meaningful images, a visible focus state during keyboard navigation, and never conveying meaning through colour alone \u2014 a red border with no error text explains nothing.<\/p>\n It’s increasingly a commercial question too: large clients and EU tenders already require compliance, and a site without it simply doesn’t pass screening.<\/p>\n For a five-page site a system is overkill. For a product that will grow for years it pays back by the second dozen screens \u2014 which is why large companies invest in it first, before any visual experimentation.<\/p>\n Multimodal interfaces \u2014 where you interact with a product by more than a finger \u2014 are gradually moving beyond the big ecosystems. Voice search inside an app, gesture control, photo input instead of text.<\/p>\n For most business products this isn’t a priority yet, with one exception: voice search inside a catalogue genuinely helps stores with large product counts, where typing a name is awkward.<\/p>\n More practical right now is making sure basic input works flawlessly. Address autocomplete, the right keyboard for the field type (numeric for phone numbers), recognising a pasted card number that includes spaces. These details affect checkout completion far more than any voice command.<\/p>\n For a long time speed was treated as the developers’ domain: the designer draws, the programmer optimises. That no longer holds, because most of a page’s weight is committed at the design stage.<\/p>\n Decisions a designer makes that can’t be “optimised” later without rework: full-screen video in the hero, three typefaces at two weights each, high-resolution background images on every block, complex shadows and blurs the browser recalculates on every scroll.<\/p>\n The working approach is to build constraints into the design from the start: two typefaces maximum, images sized for the blocks that hold them, animations that don’t block scrolling. A beautiful layout that takes five seconds to load loses to a plain one that takes one and a half \u2014 and that’s a measured pattern, not a matter of taste.<\/p>\n Not everything written about is worth adopting. A few cases where fashion conflicts with sense.<\/p>\n Elaborate scroll animation.<\/strong> Impressive in a designer’s portfolio, heavy on speed, and irritating on repeat visits. We covered how speed affects results<\/a> separately.<\/p>\n Unconventional navigation.<\/strong> A menu that has to be decoded always lowers conversion. Familiar structure is an advantage, not a constraint.<\/p>\n Glass effects on everything.<\/strong> Transparency over a busy background almost guarantees broken text contrast.<\/p>\n The general rule is simple: a trend should solve a user’s problem, not demonstrate the team’s awareness.<\/p>\n The riskiest scenario is redesigning everything at once. A product that was working changes completely, and if the numbers drop it’s unclear which part caused it: new navigation, new palette or new block order.<\/p>\n The working approach is incremental, with measurement. First a hypothesis with a specific number: we expect the new form to lift conversion from 8% to 12%. Then one element changes, not ten. Then a long enough period to accumulate data \u2014 a week on modest traffic is almost always too little. Only after that comes the decision: keep, roll back, or try another variant. Without that cycle a redesign becomes a matter of taste, won by whoever insists loudest rather than whoever has data. At Netloria we try to tie every noticeable interface change to a specific metric before work begins \u2014 it removes a great deal of argument about shades.<\/p>\n How to read those numbers and how much data you need for a conclusion is covered in our piece on website metrics<\/a>.<\/p>\n Proper usability testing costs money and time, but the basic problems surface far more cheaply. A few techniques that work for both a site and an app.<\/p>\nAI Stopped Being a Topic and Became a Tool<\/h2>\n
Personalisation: A 40% Revenue Difference<\/h2>\n
\n
Accessibility: The Market’s Biggest Gap<\/h2>\n
Dark Mode: Function, Not Fashion<\/h2>\n
\n
\n <\/th>\n Pile of layouts<\/th>\n Design system<\/th>\n<\/tr>\n \n New page<\/td>\n Designer draws from scratch<\/td>\n Content manager assembles from blocks<\/td>\n<\/tr>\n \n Changing button colour<\/td>\n Edit every layout separately<\/td>\n One edit in the component<\/td>\n<\/tr>\n \n Risk of inconsistency<\/td>\n High: five different buttons<\/td>\n Low by definition<\/td>\n<\/tr>\n \n Upfront cost<\/td>\n Lower<\/td>\n 20-30% higher<\/td>\n<\/tr>\n<\/table>\n Voice, Gestures and Other Input Methods<\/h2>\n
Speed as Part of Design, Not a Consequence<\/h2>\n
Which Trends to Ignore<\/h2>\n
How to Introduce New Things Without Breaking What Works<\/h2>\n
Testing Your Interface Without a Research Budget<\/h2>\n