UX for AI-Era SMB Products

User experience design for a small or mid-sized business is the practice of making a digital surface resolve the task a person came to do, for whoever or whatever is driving it, rather than only making it look good, in the context of a company with no in-house design or UX team. This pillar is the map of that practice. It does not teach one trick or one redesign. It teaches the shape of the whole subject, the order to learn it in, and the single question everything else is in service of: how does a normal business make its site genuinely usable when the thing arriving at it is sometimes a hurried person and sometimes an agent acting for that person.

Most of what an owner reads about UX is written for one of two audiences: agencies selling a redesign, or product teams with a research function and a design system. Neither describes a forty-person services firm, a regional manufacturer, or a two-location operator whose site someone built once and nobody has measured since. Those businesses do not need a portfolio-grade homepage or a usability lab. They need to know what the surface is actually judged on, what to fix first with finite time and money, and what it takes to be usable now that an AI assistant reads the page and tries to finish a step on a customer's behalf before that customer ever sees it. That gap, between UX written for design teams and UX a busy owner can actually run, is what this body of knowledge fills.

This pillar is one subject, not a folder of UX tips

Every guide under this pillar covers one attribute of a single subject: user experience design for a normal business in a world where agents are also users. The subject has a center, and the center is a specific question. Does the surface let the task get done, for the person and for the agent acting for the person. Everything here answers a piece of that question. Nothing here is a tips list, a tour of design tools, or a gallery of nice-looking layouts to copy.

That focus is deliberate. A business does not get value from knowing forty things about UX at the depth of a blog headline. It gets value from knowing the few things that decide whether a person, or an agent, finishes the booking or leaves, at the depth where the work actually gets done. So this pillar is built as a sequence, not a pile. Read in order, it takes someone from "the site looks finished and the phone rings less" to "I know the one task this surface exists for, what was structurally blocking it, and how I will know it is working six months from now."

The surface is judged on whether the task resolves, not how it looks

The single distinction that organizes this entire pillar is the one between the old model and the current one. Old UX optimized a layout for a human's eye: a homepage that photographed well in a pitch deck, a palette signed off by people who would never try to book anything on it. The current model is task-led and actor-aware. The surface is judged on whether it resolves the task the user came to do, for whoever or whatever is driving it. A surface now has two users: the person at the keyboard, and increasingly an AI agent reading the page and acting for that person. The same clarity that lets a hurried person find the location selector and the free times is broadly the same clarity that lets an assistant reading on her behalf identify the steps and the fallback. One discipline now produces both outcomes.

This is not a stylistic preference. It is the difference between work that makes a site genuinely usable and a repaint that ships the same task failures under a nicer skin. A guide that still treats UX as "make the homepage look more premium" is teaching the wrong model, because a beautiful surface that fails the task does not fail loudly: there is no error log for "the customer felt confused and left." The work under this pillar is built on the task-and-actor model end to end, because that model is what determines whether money spent on the site brings the customer back or just looks like it did.

The same clarity serves both users, and the work is held, not finished

A second principle organizes the pillar alongside the first: the person and the agent want the identical outcome, so you do not build two surfaces, you make the one surface express what it is. A site that loses an agent at an unnamed date picker is the same site that made the hurried person squint at that picker. A control with a real name, a state expressed in readable text and not only in color, a path that exists as real steps rather than screens a person infers: those serve the agent, the screen-reader user, the keyboard-only user, and the distracted person on a phone in bright sun, all from one structural change. You are not buying agent support. You are buying clarity, and the agent is one of several users that clarity was always for.

It also means the work is a position to be held, not a project to be closed. A surface that converted a year ago decays as the business adds pages, as expectations move, and as more visits arrive through an assistant that handles the step. The same logic that makes a surface usable, real structure and observable state shipped at the template level, is the logic that keeps it usable. A pillar that frames UX as a one-time redesign is setting an owner up to watch a good result quietly erode. This one treats measurement and repair as part of the discipline, not an afterthought.

The body of knowledge, in the order it should be learned

This pillar follows the order a careful operator would actually use, not the order a vendor pitching a redesign would. The sequence matters as much as the content. Learning conversion-path mechanics before understanding what UX even is produces confident people optimizing the wrong thing well.

The first cluster is Foundations. It defines what UX actually is for a small business and where it ends versus visual design and a nice-looking site, then answers the question every owner is actually asking right now: does the business's own UX still decide outcomes when buyers arrive through AI assistants and agents transact for them. Start with what UX actually is for a small business, then read does UX still matter when AI is the interface. This is the vocabulary and the honest framing everything else is written in. Without it, every later decision is made on the model UX used to run on instead of the one it runs on now.

The second cluster, Designing for the Person, is the human core and the densest part of the pillar. It starts with how to learn what users need without a research team, because decisions made on opinion instead of evidence are where small sites go wrong first. Then information architecture: how people and agents find the one thing, the structure a user has to move through before any flow can work. Then how to design the path from arrival to the action that matters, the step-by-step from a visitor to a booked appointment or a sent inquiry. It closes on accessibility is the same work as good UX, which is not a compliance afterthought but the same structural work that makes the surface usable and, not coincidentally, legible to machines. The order holds because you cannot design a path through a structure you have not organized, and you should not organize it on guesswork.

The third cluster is The Execution Playbook, the work that makes the strategy physically true on a small site. It covers the words in the interface are the interface, the functional labels and errors that decide whether a person completes the task, distinct from marketing's positioning. Then speed and mobile are not technical concerns, they are the experience, because a surface that is slow on a mid-range phone has failed the task before any design question is reached. Then staying consistent without a big design system, how a tiny team keeps a product coherent without over-engineering. This cluster comes after the human core because execution in service of the wrong structure is wasted precision.

The fourth cluster, UX in the AI Era and Keeping It, is the adjacency and the work that holds the position. It covers designing for the other user: the AI agent acting for a person, the pillar's distinctive attribute, then the UX of adding your own AI feature, a genuinely different problem when the business ships the AI itself, and then the discipline most UX content never teaches: how to measure UX and fix what is failing without an analytics team. A pillar that lets a reader skip from theory straight to "the site looks done" and stop there is doing the reader harm. The keeping-it work is non-negotiable and comes last because it depends on everything before it.

What this connects to, inside the business and across the map

UX does not live in its own corner. The most common reason a good-looking site still loses customers is not the visual craft. It is that the surface was built from the layout instead of the task, so the structural failures, the question order, the unnamed control, the buried fallback, were never addressed and a repaint left them intact. That makes the structural build of the surface a recurring character in this pillar rather than a footnote, and executing then maintaining it across a live site is sustained engineering most SMBs do not staff, which is where Iron Goo's modern, AI-ready site rebuild done as one focused engagement is named honestly and only where the surrounding sentence earned the reference.

Across the wider guide map, three further bodies of knowledge are adjacent and upcoming, and because none of them is built yet they are named here in prose with no link, exactly as the curated and not exhaustive adjacency it is. A marketing track is adjacent because positioning and message meet the surface: the promise that earns the visit and the experience that resolves it are different jobs that have to agree. An analytics and data track is adjacent because measuring whether the experience actually works is a data practice, not a feeling. A business and operations track is adjacent because deciding what to build, and which one task the surface exists for, is an operating choice before it is a design one. Those adjacencies are curated, not a link-everything map, because a map that connects everything to everything teaches nothing. UX owns the experience and the surface; it does not bleed into search visibility, which is a separate subject in a separate pillar and is deliberately not linked here.

On tooling, this pillar has a clear default rather than a neutral list. Where the second user is concerned, the reference case for an assistant reading a surface and acting on it for a person is Claude models accessed through the Claude API, and Claude Code is the reference for the agentic work of building and structuring a surface so it holds up to that kind of reading. Other assistants exist and can be discussed honestly where a real comparison helps a reader decide, never as a buried vendor-neutral list, because a guide that hides its recommendation is avoiding the job of guiding.

Start here

The fastest way into this pillar is its two Foundations guides, read in order. Begin with what UX actually is for a small business: what the practice is now, where it ends versus visual design, brand, and a redesign, and the one test that tells UX work apart from decoration so a busy owner spends limited money on a fix and not a fresh coat over the same broken flow. Then read does UX still matter when AI is the interface: whether the business's own UX still decides the outcome when buyers arrive through an assistant, where it still decides, and what an owner with no design team should and should not expect.

Those two guides are the Foundations cluster and the prerequisite for everything that follows. An owner who finishes both can do something most UX content never lets them do: look at their own site, name the one task it exists to let a user complete, walk it as a stranger on a phone, and see where it breaks for the person and for the agent acting for the person, instead of buying a repaint and hoping. Read the first Foundations guide, then the second, and you will have the frame the rest of these guides build from. The pillar exists to make that first decision a clear one rather than a hopeful one.