Good UX Always Needed the Whole Context. Now AI Can Hold It.

Of all the places we use AI in delivery, design is the one where we lean in hardest, and it is where we see some of our best results. That can sound surprising, because design is usually thought of as the most human part of building software. This post explains why we think AI suits design so well, and where the limits are.

Shouldn’t good UX start with users?

Yes. The common view is that design should be human-first, built on a clear understanding of users’ needs, and we agree with it. Nothing in this post argues against talking to users or testing with them.

In practice, though, users’ needs are only part of the picture. Most custom software has to work within constraints that users never see:

• Technical constraints, such as the shape and quality of the data, and legacy systems the new software has to read from and write to.
• Business processes, including the approvals, exceptions, and hand-offs that keep the organisation running.
• Compliance, such as audit trails, consent, data retention, and the rules of a regulated sector.

A design that meets users’ needs but ignores these constraints looks right in a review and then falls apart in the build.

Why was the best UX process always the one that understood the context?

In our experience, the best UX processes always had a full understanding of that context. The people involved knew what the data looked like and where it came from. They knew which legacy system held the truth, which steps in the process could not be skipped, and what the compliance team would ask for.

That knowledge shapes every screen. A list that might hold 4,000 records gets search and filtering rather than a scroll bar, and a form that writes to an older system validates what that system will accept. An approval step that the business depends on is designed in from the start, rather than bolted on when someone notices it is missing.

I have been lucky enough to work with some exceptional UX designers. In my years as a solution architect, the best products came from a great relationship and a shared understanding between us. That kind of partnership was the exception rather than the rule, because it depended on the right people working closely together for long enough to build it.

Why was that understanding so rare?

The context was spread across people who were rarely in the same room as the designer. Business analysts knew the processes and the compliance requirements. Engineers knew the data and the legacy systems. Designers often worked from a brief and produced screens before either group had been consulted in depth.

The constraints then surfaced during the build, and the design was patched to fit. Good teams closed the gap by putting designers, analysts, and engineers together, but design time was expensive, so it went into the main journeys and a few key screens. Everything else, including admin screens, error states, and reports, was built from the nearest example.

How does AI flip the order?

AI lets the people who hold the context do the design. Our business analysts and engineers, who already understand the processes, the compliance requirements, the data, and the systems, design with Claude at every stage, from discovery through the build. With good brand input from the client or their agency, we do not need a separate designer.

Holding the whole context at once

AI has a particular strength here. It can cross-reference requirements, process maps, data models, compliance rules, and user feedback, keep several viewpoints in play at the same time, and keep track of decisions recorded earlier in the project. In our view, it applies all of that more consistently than most people can, because no single person carries every constraint in their head.

That strength depends on what AI is given. It holds the context our business analysts and engineers supply, in requirements, rules, and the design system, and people check that each design reflects it.

For me, this is what brings something close to the shared understanding of the best design partnerships to every project, not only the ones lucky enough to have it.

Patterns first

The product’s patterns and components live in its design system, documented in Storybook. When AI designs a new screen, it builds from those parts. A search screen uses the product’s search pattern, a form uses its form components, and a table is designed for the real number of rows. The design starts from what the technology can do, rather than discovering it later.

Real constraints, early

Because the people designing know the context, the first draft reflects it. Long names, empty fields, slow legacy integrations, approval steps, and compliance rules shape the design from the start, instead of surfacing as surprises in a sprint review.

More options, every screen

AI is involved at every stage of our design process, so we explore and test far more options than a traditional process allows. The time and cost of design are now small enough that it is included in our discovery and build work, and we apply UX to every screen on our current projects, not a sample.

Our guide to UX design for custom software walks through each stage, and our guide to design systems, UX patterns, and Storybook explains the patterns that make this work.

What does this not change?

Letting the technology drive the design does not mean letting AI decide. Three things stay with people.

Users still have the final word. Patterns tell you how a screen usually works, not whether it works for your staff or your customers. People give the first input, users react to working prototypes, and people review every design before it is built. AI does not run user research.

Brands still need to be distinctive. AI falls back on common patterns when it has nothing specific to apply, so the distinctive parts of a brand have to be written down. That is why the brand input matters so much, and why we have written a guide to AI-friendly brand guidelines that clients can forward to their agency.

New problems still need new thinking. Patterns cover the common cases well. The screens that solve a problem nobody has solved before are where people should spend most of their time, and AI’s speed elsewhere is what frees that time up.

Our guide on where AI helps in UX design and where people decide covers that split in more detail.

What should buyers take from this?

If you are commissioning software, the question to ask a supplier has changed. It used to be ‘do you have designers?’ A more useful question now is ‘how do the people who understand our data, systems, processes, and compliance requirements shape the design, and how is AI used to explore and test it?’

Ask which screens get design attention, how users are involved, and who reviews what AI produces. A supplier that designs every screen from your brand, your patterns, and your real data, with people reviewing each one, will give you software that fits your constraints and the people using it. Our guide to UX on every screen explains why that is now affordable.

If you would like to see how this works on your own system, book a consultation.