xf

The end of the front end (and maybe brand as you know it, too)

Howard Pyle
10 min read

The post-interface web and how agentic systems will redefine customer experience and brand. Back in 2021, we released a white paper and wrote that websites would eventually disappear. Not the internet itself. But the idea of a “front end.”

We said that people would soon interact directly with the data underneath, through intelligent systems that understood intent instead of navigation.

At the time, there were two responses. Most people responded with either “What the hell are you talking about?” or, “Wow, this is visionary.” Honestly, it was probably closer to 95% confusion and 5% excitement.

We weren’t trying to predict the future, just describing what we saw as web builders: huge API ecosystems running underneath, but still funneled through a 1990s metaphor the website as “storefront” (and yes, companies literally had websites designed like a physical store with doors and windows). Everything sat behind a thin visual layer that brands controlled, even though most of the meaningful work was happening below it.

Agents and APIs are eating the web. The same iceberg. You can't escape it.

A piece in Fast Company just ran titled "In the Age of AI, Websites Will Be Transformed," which describes how AI agents are reshaping how people get information, with adaptive interfaces that build themselves around the user.

So it’s both exciting and a little familiar. The conversation finally caught up. The key point is that the idea of customer experience, and perhaps even the idea of brand itself, might be on the verge of changing forever. Not that brand is going away (I believe brand is the emotional relationship someone has with a company), if anything, it becomes more important when AI agents need to decide which organizations to trust and recommend. But brand expression is going to implode and take on entirely new meanings in the coming two years. Your carefully designed website, a controlled journey, your visual identity system.. Maybe these matter less when users never visit your site and their personal AI tools sit in between the customer and your company. What matters more is reputation, trustworthiness, and whether your data is accessible and reliable. Brand becomes less about what you show and more about what you are.

So here we are back at the point that values, purpose, and ethics are vital for modern businesses.

From Interfaces to Interactions

Fast Company talks about a world where people talk to intelligent systems that instantly shape content to their needs. No fixed templates, no clicking through menus. Think about that for a second. UX makes things easier but it also guides people through data and content the way the organization wants you to consume it. No CMO wants to users to access blob of information about their company that has no starting or ending point. That’s one reason “journey maps” are so popular as a strategic planning tool - you can try to shape how people move through content, marketing funnels, etc.

That’s exactly what we meant in our white paper when we said digital systems should be wrapped around the individual, not organized around the organization. We weren’t talking about personalization in the marketing sense, but rather systems that understand the context of that specific user and adjust in real-time. It’s essentially experience as a living exchange between a person and the data behind the scenes. Kind of like what happens when you want to chat with Perplexity or GPT about a company vs navigating their entire dumb website.

And this isn’t just Fast Company or thought leadership pieces like our white paper. Designers and researchers have started using terms like Agentic UX but I think that term feels like a bolt-on. I think what they mean is that we have to shift into designing for intelligent intermediaries rather than human clicks. So that means designing interactions that Agents or AI understand who will then pass it along to the user / customer / etc. No joke, I saw an ad for Postman (an API platform) the other day, and part of the ad copy was “AI agents are reshaping AI use, but human-centered design slows innovation.”

In 2024, UXDesign published "The Agentic Era of UX" about this exact shift, and UX Magazine has been exploring new patterns where interfaces dissolve into data conversations. McKinsey calls it the "Agentic AI Advantage," framing it as the next operational shift for business.

Agents as the New Intermediaries

The Fast Company article also talks about OpenAI’s Agent Builder. This is essentially a concierge that does the searching and sorting for you. There are ton of these emerging and tools like n8n, Make and even Manus give powerful agent capabilities to anyone who’s up for trying to wrap their heads around it.

One of the key things we were trying to describe in our work is the impact of a small number of technologies (e.g. chatGPT or perplexity, etc) sitting between the human and whatever collection of content and organizations you need for any task. So think about if you use chatGPT to answer a question about doing your quarterly taxes: it’s going to pull from content, describe tools you can use, and from its memory about you from prior chats, and organizations required for any task. That’s an experience 100% customized around the individual and whatever context I provide about my situation. There’s no journey map that can predict their context, collection, or solutions surfaced up to them. That’s the shift we’re seeing now. The UX layer is becoming a negotiation between the user's AI layer and agents, and a collection of data that is hard to predict.

This is where it’s important to really prioritize AEO / GEO. Trying to get AI tools to ingest your content so it’s surfaced to users during whatever it is they’re using AI for. New standards like MCP (Model Context Protocol) and A2A (Agent-to-Agent) communication also become essential for any organization trying to insert itself into AI ecosystems that users create for themselves. A paper published in July 2025 by an international team of academics titled "Agentic Web: Weaving the Next Web with AI Agents" calls this the missing connective tissue. These protocols let systems talk to each other safely without a UI.

The End of the Walled Garden

Traditional digital strategy has always been about getting people into your system and keeping them there. That’s the walled-garden model. APIs are the secret plumbing behind many systems and products (and have been for a long time), but the agentic web breaks that completely for the end users.

If my AI assistant can get what I need without visiting your site, the walls don’t matter. Neither does your design or your brand experience. I’m just getting the data or information I need.

The dissappearing click How many times can you redesign your site before your content just is absorbed by AI?

We argued that the strategy of having a walled garden in your organization is directly opposed to the future of creating AI-oriented digital experiences. So organizations that already have practices around making their content and tools widely available (think of open source models) have a huge advantage now. And the opposite statement is that companies with brittle, highly intricate tools or reams of gated content are quietly becoming ignored by the users they’re trying to reach. It turns out that openness isn’t just feel-good, it’s good UX and good for business in this Agentic AI era.

We’ve been working with startups, enterprises and non-profits to help them build creative systems that connect brand, product, and AI workflows so organizations can operate in this open, agentic landscape that’s emerging. It’s a combination of business, brand, tech, and ethics.

Agents on a Human Web

The web was built for people, not machines. Every form, menu, and layout teaches humans how to find things. Agents hit that surface and have to pretend to be people to get data, and that’s where it breaks.

Most AI startups still rely on scraping stacks that have existed for decades. Headless browsers, Selenium, Playwright. They click, scroll, and parse HTML. It works until it doesn’t. Layout changes, bot detection, and render differences constantly break pipelines. Entire vendors exist to paper over the cracks: Browserbase sells fleets of headless browsers, Zyte and Apify offer scraping APIs that repackage page data as JSON (2025, browserbase.com; zyte.com; docs.apify.com).

Meanwhile, platforms are starting to close the door. Cloudflare now lets sites block AI crawlers by default (July 2025, The Verge). Publishers are testing paid access models to license data rather than have it scraped (2024, Wired). The idea that a public page is an open resource is quietly disappearing.

If agentic AI is the next step, the fix isn’t better scraping. It’s structured access. APIs, not HTML. But exposing APIs is hard work. You need to manage ownership, schema mapping, privacy, rate limits, versioning, and abuse. That’s why so many teams keep patching the browser layer (it looks faster until it collapses).

A few startups are rebuilding from the ground up. Exa offers semantic search and crawl APIs so agents query meaning instead of markup (2025, exa.ai). Diffbot turns web data into a live knowledge graph (2025, docs.diffbot.com). Perplexity and Firecrawl provide clean text and search endpoints for LLMs (2025, docs.perplexity.ai; firecrawl.dev).

There was an effort to legislate this shift. The ACCESS Act of 2021 proposed interoperability and data portability, but it stalled in Congress (June 11 2021, congress.gov). So the market is filling the gap.

If companies don’t invest in APIs now, agents will still reach their data, just through slower, more fragile paths they don’t control. The API has quietly become the new interface.

Automation as the New Design Layer

The other thing the Fast Company article talks about LLM-powered websites that rewrite themselves dynamically. The future is a world where design and content generation blend together. Where the interface isn’t handcrafted anymore, it’s composed on the fly by algorithms.

To see this in action, I built a simple demo showing how an AI agent completes a healthcare workflow (prescription refill + flu shot) in 2 minutes without visiting any brand websites. Try the demo → That doesn’t make design less important; it makes intent the new design material.

As Figma's recent MCP server announcement shows, serious design players believe that interfaces are becoming dynamic representations of design intent rather than fixed screens. Future design will be generated on the fly based on context and patterns rather than handcrafted layouts. Designers won’t just be deciding how things look, they’re defining how systems behave.

This is why in addition to AEO/GEO and a clear API strategy, every brand needs a well-groomed design system that not just for designers to use, but optimized for AI and automation tools as the fodder for these generative interfaces. This means treating your brand system as a data collection, not just a 100-page PDF guidelines doc.

What's Still Missing: Values as Infrastructure

We’re all a little anxious about agents between us and the information we need. But even in navigating a sea of poorly designed websites to manage your healthcare or education, it’s frustrating but at least you could see what you were dealing with. You could spot the dark patterns, the buried disclaimers, the biased recommendations.

When an AI agent sits between you and the data you need, those problems don't disappear, they just become invisible.

Imagine an AI agent helping someone compare health insurance plans. If the underlying data comes from insurers who profit from denying claims, what gets surfaced? What gets hidden? The interface is gone, but the bias remains. Now it's just wrapped in the authority of an "intelligent system."

This is why values can't be an afterthought. In the agentic web, your values become part of the infrastructure.

When interfaces dissolve into data conversations, the algorithm IS the experience. If your organization's values aren't embedded in how data is structured, accessed, and prioritized, you're not just risking ethical violations, you're scaling harm invisibly.

This has been the central question that drives our work: How do we ensure that agentic systems serve people's actual needs rather than optimizing for outcomes that benefit organizations at the expense of users?

It's not enough to bolt on "AI ethics" after the fact or have it be a checkbox powerpoint slide living in a side room next to the CIO’s office. Organizations need to:

Define their values explicitly - Not as marketing language, but as operational principles that guide how data is structured and surfaced Embed values in data architecture - If your API prioritizes profit over user benefit, agents will learn that priority Design for transparency - When there's no interface to show your work, how do users know what's being hidden? Protect user agency - Agentic systems should empower choice, not replace it

The organizations with the best brands in this agent era will not necessarily be the ones with the best AI models. They'll be the ones whose values are clear, consistent, and embedded in every layer of their technology stack. So clear the users have trust, belief and loyalty.

Because when the interface disappears, values are all that's left.

The Same Roadmap Still Works

In 2021 (cough) we mapped a four-stage model for how people and organizations would mature into this new landscape: Baseline, Beginner, Intermediate, Advanced.

I think it still holds.

Brand + API Maturity

Where is your brand on this chart?

The point wasn’t technology adoption. It was trust, transparency, and adaptation. Here’s what we wrote in our 2021 white paper (Ethan McCarty suggested we call it the “Stepwise Model”):

Baseline — Most organizations operate as isolated systems with fixed front-ends and manual workflows. Digital experience is fragmented; data lives in silos. Personalization means static audience segments and basic content targeting. Users have little control or visibility.

Beginner — APIs and automation start connecting systems internally. Experience teams begin experimenting with adaptive interfaces, but still design around organizational structure rather than individual need. Governance and ethics are reactive, not embedded.

Intermediate — Cross-organizational integration emerges. People can complete multi-party tasks (for example, booking travel, healthcare, or education services) through a single digital flow. Early “agent” layers appear; software that brokers between systems. Ethical and accessibility frameworks start to formalize.

Advanced — Interfaces dissolve into data interactions. Agents mediate most engagement between individuals and institutions. Systems adapt automatically based on user context and declared values. Ethics, transparency, and inclusion are designed in from the start, not added later.

The distinction between front-end and back-end disappears.

So What Happens Next

If customers stop visiting your website, what happens to your brand? (And your budget for it)

If people interact directly with your data, what defines their experience of your organization?

When there’s no interface, there’s nothing left but information, algorithms, and intent.

To me, the conclusion is that it’s not enough to only look at how AI integrates with brand, marketing, and CX as data flows or prompts help internal teams generate new content, and instead look at it as a wholesale reinvention of how people interact with your organization overall.

That’s the real transformation.

Some references / further reading: