UX Design Services for Web and Mobile Products
UX design is the work of deciding what a product should do, for whom, and in what order, then proving that decision with evidence before anyone builds it. Screens are one output of that work, not the substance of it. Mixcore Studio runs UX for web and mobile products: user research, information architecture, task flows, interaction design, accessibility conformance, and the measurement that tells you afterwards whether any of it worked.
We have been shipping software for more than 8 years with a team of around 20 people and 320+ projects delivered, which means our designers sit beside the engineers who have to implement the result. That proximity is the main reason our specifications survive contact with a sprint.
The ground moved sharply in 2025 and 2026. Producing an interface became cheap, accessibility became enforceable law across the EU, and AI disclosure became a UI requirement rather than a design opinion. What follows is what we actually do about each of those.
What changed in UX design during 2025 and 2026
Design tools became agentic rather than merely AI-assisted. At Config 2026 on 24 June, Figma shipped six things at once: Code Layers (executable code on canvas, in beta), Figma Motion (a native timeline with keyframes, open beta, exporting to CSS, JSON, React, MP4, WebM, animated SVG and GIF), AI shader fills and effects, generative plugins written from a prompt, Weave node-based generative workflows, and an upgraded agent with skills and connectors for Notion, Slack and GitHub. The agent is free during beta and moves to AI-credit billing at general availability, which turns design tooling into a recurring operating cost rather than a fixed seat price. If you are budgeting a 2027 design function, that line now exists.
The deeper structural change is that design-to-code stopped being an export. The Figma MCP server gives a coding agent a continuously queryable connection to a file's component hierarchy, variables, layer tree and Code Connect mappings, instead of handing it a flat image and a spec. The consequence is practical: token and component hygiene is now a hard prerequisite for AI code generation quality. A disorganised file no longer merely irritates a developer, it caps what any agent can produce from it.
Several things clients still have in their stack are finished:
- Adobe XD — no longer under active development, and Adobe has been retiring its XD support pages. Do not start new work in it.
- InVision — its design and collaboration products shut down permanently at the end of 2024, and invisionapp.com now resolves to Miro. An InVision prototype in a handover pack is dead infrastructure, not a reference.
- Static redline handoff — PNG-to-developer and Zeplin-style exports are superseded by Dev Mode, Code Connect and the MCP server. Feeding a codegen pipeline screenshots instead of tokens is a self-inflicted quality ceiling.
- Bespoke token JSON schemas — a homegrown format now costs you the entire tooling ecosystem, which has standardised on DTCG.
Penpot has become the credible open-source alternative, which matters when procurement blocks a particular cloud tool. Role boundaries moved too: the Figma 2026 AI Report found 60% of developers now participate in design work, up from 44%, and 41% of designers participate in development, up from 21%.
What you are paying for when a model can produce screens in minutes
This is the most common question we field in 2026, usually from someone who has already produced something that looks 70% right in Figma Make, v0 or Lovable. The honest answer is that generation collapsed the cost of producing an interface and did nothing to the cost of knowing which interface to produce. The Figma 2026 AI Report, drawn from 8,403 survey responses and 639 qualitative interviews across ten markets, puts it plainly: "AI can build almost anything. What it can't do is determine what's worth building."
The same report found that 41% of product builders say AI meaningfully changes how their teams work together today, against 7% two years earlier; that 90% say design is at least as important as it was before AI, with roughly 60% saying it is more important; and that 65% of developers agree design's importance has increased.
So the invoice covers research with real users, the decisions that follow from it, a system somebody can maintain after we leave, accessibility that survives an audit, and named accountability when something ships broken. We use generation tools ourselves and we tell you where they were used. What we do not do is hand over output that no human verified against a measurable bar.
There is a related objection worth naming directly, because buyers raise it before we do: everything is starting to look the same. The default AI aesthetic is recognisable on sight now, and visual polish alone has stopped signalling investment. We treat a distinctive register as something to argue for from your brand, category and audience, not something to select from a menu of templates.
Designing AI features that people can actually operate
Chat is no longer the default surface for an AI feature. The 2026 consensus is that conversational UI underperforms graphical UI for structured tasks, because linear reading, context recall and multi-turn error correction are all worse inside a transcript. Bolting a chat box onto a product now reads as a failure to design the interaction. The patterns that hold up are agentic and ambient:
- Planning visibility — show the intended steps before the system acts, not in a log afterwards.
- Tool-use disclosure — name the system being read or written to, at the moment it happens.
- User-set autonomy — let people choose suggest-only, confirm-each-step or run-unattended, per task rather than per account.
- Streaming partial results — output that can be judged early beats a spinner followed by a finished wall of text.
- Multi-step workflow tracking — a durable, inspectable, resumable view of a long-running job.
- Cheap recovery and escalation — undo, correct a single step without restarting, and reach a person without losing state.
A declarative generative-UI stack has formed around these patterns: Google's A2UI, the OpenAI Apps SDK for React UI rendered inline in ChatGPT, MCP Apps and mcp-ui for host-rendered tool interfaces, and AG-UI for streaming agent UI into your own frontend. Which of those you target is a distribution decision as much as a design one, and it should be settled before the interface is drawn.
Disclosure is now a legal UI requirement in the EU. EU AI Act Article 50 obligations apply from 2 August 2026, with Commission guidelines on those transparency obligations adopted on 20 July 2026 and a voluntary Code of Practice on Marking and Labelling of AI-generated Content published alongside them. Users must be told they are interacting with an AI in a way that is perceivable in the interaction itself. A line in the terms and conditions, a metadata-only watermark, or a euphemistic "assistant" label does not satisfy it. Deepfakes must be labelled, and AI-generated or altered content must carry machine-readable marks. The European Commission sets non-compliance at up to €15 million or 3% of worldwide annual turnover, whichever is higher. We design the disclosure surface and specify its states; classifying your system and confirming your exposure is your counsel's job, not ours, and we will say so in writing rather than let it fall between us.
Accessibility is enforceable law in the EU, not a quality bar
The European Accessibility Act (Directive (EU) 2019/882) has applied since 28 June 2025. The technical bar is EN 301 549, produced jointly by ETSI, CEN and CENELEC. The published version is still V3.2.1 from March 2021, which incorporates WCAG 2.1 Level AA. A V4.1.0 revision has been working through the ETSI process since November 2025 — the final draft is dated June 2026 and is in the vote phase, not yet published or cited in the Official Journal — and it updates clauses 9, 10 and 11 to align with WCAG 2.2, which adds six new success criteria at the conformance levels that matter — two at Level A and four at Level AA. We therefore write WCAG 2.2 AA into contracts as standard. Specifying 2.1 AA in a 2026 contract books a re-audit you will pay for twice. WCAG 2.2 is the current W3C Recommendation (5 October 2023, updated 12 December 2024), has 86 success criteria, and is backwards compatible, so meeting it also meets 2.1 and 2.0.
National enforcement regimes are in place, but published penalty decisions remain scarce, so a vendor quoting you an EAA fine as a settled fact is guessing. The exposure is nonetheless real, because penalties are set nationally: in Spain the sanctions regime that applies runs to a maximum of €1,000,000, and Germany's Barrierefreiheitsstärkungsgesetz sets its top fine at €100,000.
Two details decide scope more often than anything else. The microenterprise exemption requires fewer than 10 employees and a turnover or balance-sheet total of no more than €2 million, and it covers services only, not products. And an Article 14 disproportionate burden claim is not a shrug: it has to be documented against the Annex VI criteria, made available to enforcement authorities on request, and reassessed at least every five years.
In the United States, the DOJ Interim Final Rule published on 20 April 2026 extended ADA Title II web compliance by a year, to 26 April 2027 for public entities serving 50,000 or more people and 26 April 2028 for smaller entities and special district governments. The standard remains WCAG 2.1 Level AA, and the extension suspends neither the underlying nondiscrimination and effective-communication obligations nor the ability of private plaintiffs to sue.
We will not sell you an accessibility overlay, and we will put the reason in writing if you ask for one. The US Federal Trade Commission ordered accessiBe to pay $1,000,000 — announced 3 January 2025, final order approved April 2025 — over claims that its automated product could make any website WCAG compliant, and the order bars that claim outright. In litigation an overlay is now an aggravating factor rather than a defence, because plaintiffs cite the widget's own separate, inferior interface as evidence of a barrier. Hundreds of accessibility professionals have signed a public statement against them. Remediation belongs in the markup, in the components, and in the design decisions upstream of both.
We also do not build to WCAG 3.0. It is a W3C Working Draft, latest version 3 March 2026, whose own text states that requirements will be added, combined and removed; its conformance model replaces the pass/fail A/AA/AAA structure and is still being revised; and no legislation references it. Designing to it today is speculative work billed to a client.
iOS 26, Android 16 and the degraded path
Both platform design languages re-based, and Apple's is effectively compulsory. Since 28 April 2026, apps uploaded to App Store Connect must be built with Xcode 26 or later against an iOS, iPadOS, tvOS, visionOS or watchOS 26 SDK. That governs the build SDK, not the deployment target, so you can still support older OS versions. Building against that SDK applies Liquid Glass to native controls by default unless you opt out. Liquid Glass was announced at WWDC on 9 June 2025 and spans iOS 26, iPadOS 26, macOS Tahoe 26, watchOS 26 and tvOS 26, using real-time rendering with specular highlights that react to movement.
Shipping the default is a decision, not a safe harbour. Nielsen Norman Group published a critical usability read titled "Liquid Glass Is Cracked, and Usability Suffers in iOS 26", and legibility over busy backgrounds is the recurring failure. We check contrast per screen against real content rather than against a marketing render, and we opt out where the material costs more than it returns. On Android, Material 3 Expressive is now the reference: Google describes it as an expansion of Material Design 3 with research-backed updates to theming, components, motion and typography, complementing the Android 16 visual style and system UI. Material 2 and pre-Expressive specs are no longer the reference.
Motion is where an impressive demo and a real experience diverge most, so we treat the degraded path as a deliverable rather than an afterthought. CSS scroll-driven animations (animation-timeline with scroll() and view()) run on the compositor with zero JavaScript and are supported in Chromium since 115 and in Safari 26.0; Firefox added same-document View Transitions in version 144 (October 2025) but still does not support the cross-document form, so a cross-fade fallback is required there. We state a frame budget, measure it on a mid-tier Android over a throttled connection, and specify the reduced-motion and no-JavaScript behaviour explicitly. A desktop screen recording is not evidence that a thing performs.
When UX work is the wrong spend
Three situations where we say so before taking the work:
- The constraint is not the interface — if the real problem is pricing, stock availability or a support queue measured in days, redrawing a flow will not move the number. We would rather scope a smaller diagnostic than sell a redesign that cannot succeed.
- Research to confirm a decision already made — if the decision is locked for commercial reasons, a study is theatre. Say so, and spend the money on usability testing the build instead.
- A full design system before product-market fit — systems are maintenance commitments. Built too early, they ossify assumptions you have not tested yet.
That last one is worth some numbers. zeroheight's Design Systems Report 2026, a survey of 147 design-system practitioners, found design tokens in 86% of systems but only 40% of teams with any token pipeline automation and only 11% with any code-to-design sync. Just 8% describe their system as very stable, while 34% say unstable and a further 10% very unstable. Satisfaction with organisational buy-in fell from 42% to 32% year on year, and driving adoption has topped the challenge list for five consecutive years. A documented system that drifts from production is the default outcome, not an edge case, and the difference is a pipeline plus a named owner.
When a system is the right call, we build tokens in DTCG format. The Design Tokens Specification reached its first stable version, 2025.10, on 28 October 2025, with the Format and Color modules published as Final Community Group Reports; the Color module covers Display P3, Oklch and the other CSS Color Module 4 spaces. We generate with Style Dictionary, Tokens Studio or Terrazzo, and maintain Code Connect mappings so the MCP server serves real components rather than approximations. One procurement note, because it catches people out: DTCG is a W3C Community Group publication, explicitly not a W3C Standard and not on the Standards Track. If your contract language demands a "W3C standard", amend the language rather than the format. The Resolver Module that handles context-dependent values, such as light and dark themes, is still a preview draft carrying an explicit "do not attempt to implement this version" notice, so we handle those contexts in the build pipeline instead of pretending the spec is ready.
Our expertise
- User research and usability testing
- Information architecture and flows
- Interaction and motion design
- Accessibility to WCAG 2.2 AA
- AI and agentic interface patterns
- Design tokens and developer handoff
Frequently asked questions
What is the difference between UX design and UI design?
UX design decides what a product does and in what order, based on research into who uses it and what they are trying to finish. UI design decides how that resolves visually and how each control behaves. A product can be attractive and still fail because the underlying flow is wrong, which is a UX failure, and it can be well-structured yet unusable because contrast, hierarchy and states are wrong, which is a UI failure. We run both, but we scope and price them as separate work.
Does the European Accessibility Act apply to our website or app?
If you sell in-scope products or services to consumers in the EU, it has applied since 28 June 2025 across the EU, regardless of where your company is based. The microenterprise exemption covers fewer than 10 employees together with turnover or a balance-sheet total of no more than €2 million, and it applies to services only, not products. The technical bar is EN 301 549, whose published version is still aligned to WCAG 2.1 AA; a V4 revision that aligns it to WCAG 2.2 is in the ETSI process but not yet published. Whether your specific service is in scope is a legal determination, so we scope the conformance work and your counsel confirms the obligation.
Will an accessibility overlay or widget make our site compliant?
No, and installing one can make your position worse. The US Federal Trade Commission ordered accessiBe to pay $1,000,000 in a case announced on 3 January 2025, with the final order approved in April 2025, and that order bars the company from claiming its automated product can make any website WCAG compliant. Plaintiffs now cite an overlay's separate, inferior interface as evidence of a barrier rather than a remedy. Real remediation happens in the markup, the components and the design decisions upstream of both.
How should an AI feature disclose itself under the EU AI Act?
EU AI Act Article 50 obligations apply from 2 August 2026 and require the disclosure to be perceivable in the interaction itself. A clause in the terms and conditions, a metadata-only watermark, or an ambiguous "assistant" label does not meet it. Deepfakes must be labelled, and AI-generated or altered content must carry machine-readable marks. The European Commission sets non-compliance at up to €15 million or 3% of worldwide annual turnover, whichever is higher. We design the disclosure surface and its states as part of the interface, while classifying your system remains a legal question for your counsel.
If AI tools can generate screens, why hire a UX team?
Because generation reduced the cost of producing an interface and left the cost of knowing which interface to produce untouched. The Figma 2026 AI Report, based on 8,403 responses and 639 interviews across ten markets, frames it as "AI can build almost anything. What it can't do is determine what's worth building." What you are buying is research with real users, the decisions that follow, a system your team can maintain, accessibility that holds up to an audit, and someone accountable when something ships broken.
Contacts
We are always happy to talk with you.
Feel free to contact us in any suitable way
Request a quote
Let's discuss your project!
Please, provide us with a brief description of what you
already have and what you are going to achieve.
Mail us contact@brainiacminds.com