Samuel JohnsonLead UX & Product Designer
Samuel Johnson · Lead UX & Product DesignerLondon / Kent

Intelligent systems people actually want to use.

I’m Sam, a Lead UX Designer at Tesco Technology. I’ve spent 12 years designing enterprise tools. Now I work out what UX needs to do about emerging capabilities, and turn them into products people trust and actually use.

By intelligent systems I mean machine learning, AI, large and small language models, and agentic workflows, working together as one system.

See the work How I lead
  • Lead UX DesignerTesco Technology, since Feb 2026
  • 12 yearsEnterprise tools, retail, media, the arts
  • ~30 designers10 I lead directly, 20 more through the wider team
  • 650 peopleUsing the agentic UX frameworks I helped shape
A glimpse of the workScreens, workshops, prototypes
01 · ScreensRange ReviewRange Review, scenario comparison and the impact summary, with the storyboard that started it.
02 · ProcessWorkshop wallDiscovery boards, a fishbone of buyer pain points, a printed process map and the MVP wall.
03 · ConceptsAgentic prototypesPricing and range prototypes, the relationship model and the shared commercial hub.
04 · PeopleThe peopleCo-design workshops, card sorts and wall reviews, and the team behind the work.
Image · ScreensRange ReviewRange Review, scenario comparison and the impact summary, with the storyboard that started it.
01 / 04
My workThree case studies · Password protected

Three case studies. Three things I can do.

All from Tesco, each building on the last. Range shows I can ship: machine learning that buyers trust, live at scale. Beyond the chatbot shows I can set direction: frameworks a whole department now uses to design intelligent systems. From delivery to direction shows I can scale design: designers who lead their own domains, with one shared language.

How to get the password
ImpactTesco Technology, 2022 – today

What it’s added up to so far.

Two results from shipping a product, and two from changing how a whole department works.

01 · Range1–2 days

To build store shelf layouts, down from 2 weeks, once ranging became science-led.

01 · Range~20%

Of Tesco’s ranges now run through it, grown from a single trial in cheese.

02 · Beyond the chatbot650

Practitioners across every job family in the department using my frameworks, now reaching around half of Tesco Technology.

03 · Delivery to direction~30

Designers I lead: 10 directly, and 20 more indirectly through the wider team.

About meLondon / Kent

Curious, quick to learn, still hands-on.

Most of my work has been enterprise tooling: the internal systems people rely on to do their jobs. I don’t see myself as an AI expert. I’m a UX designer who stays across emerging capabilities and works out what they mean for the people using them. I get into the detail of how a business really works, and I’m usually the one asking whether users still have control, and whether we’ve earned the right to automate yet.

Experience2014 – today

12 years, from agencies to enterprise.

Tesco TechnologyLead UX DesignerSenior UX Designer (contract to permanent)2026 – present2022 – 2026
JellyfishSenior UX Designer2020 – 2022
CACISenior UX Designer (contract)2020
HatchedUX Designer2019 – 2020
TateUX Designer2018 – 2019
CACIDigital Designer2016 – 2018
Constant CommerceUX Designer2014 – 2016
How I lead · 1 of 2My story

I’d rather write the story than be the hero of it.

Every stage of my career taught me something about trust. It’s why I lead the way I do.

  1. 01Junior

    I craved
    responsibility.

    I started in UI, asking for more than I was given. The people who trusted me early made the biggest difference.

  2. 02Mid to senior

    I wanted
    autonomy.

    I moved from screens to whole journeys, and learned to own a problem end to end.

  3. 03Senior

    I wanted to be the hero.

    I led the big projects. Some went well, some didn’t, and I learned I couldn’t be the hero of every story.

  4. 04Lead

    Now I help others be the hero.

    I shape the narrative so others can thrive in it, and show them the same belief that helped me succeed.

In practice, I’ve sat on hiring panels for at least 10 of the current team, supported designers through progression and promotion, and mentored people into their first UX roles.

How I lead · 2 of 2Working with me

Trust people early. Then back them.

Ownership

Give people real ownership.

The designers I work with own their area’s direction, not just its screens. I’d rather guide a decision than make it for them.

Voice

Get their work into the room.

I make sure designers are heard where direction and budgets are set, and that their ideas travel beyond the project.

Craft

Stay close to the work.

I still work from research to UI, so I can coach in the detail, while leaving designers room to own the craft.

Support

Adapt to the person.

People do their best work in different ways. I change how I support someone, rather than asking everyone to work like me.

Credit

Let the credit land with them.

I’d rather see a designer I work with shape a leadership decision than take the credit myself. Recognition is part of the job.

ThinkingHow I approach intelligent systems

What intelligent systems need from UX.

When someone says “we need AI”, the useful question isn’t which model or which screen. It’s what UX has to put in place for the opportunity to land. Three ideas sit underneath my work, and a set of habits for putting them into practice.

By intelligent systems I mean machine learning, AI, large and small language models, and agentic workflows, working together as one system.

They grew out of retail, but they apply to any enterprise where people make high-stakes decisions with intelligent systems: banks, insurers, healthcare, logistics.

01 Relationships02 Context03 Principles04 In practice
Six human–AI relationship types
Context architecture for commercial decisions
Four agentic UX principles
01 · Relationships

Six ways people and AI can work together.

First, decide what relationship you’re designing: who leads, who decides, and how much the system is trusted to do.

Human-ledAI-led
Adviser Human explores, AI advises.

The person wants expertise, interpretation or guidance.

Design for: Memory, explainability, consistency

For example: Explaining why a forecast changed this week.

Storyboard and UI pattern for the Adviser relationship.

A mix, not a ladder.

This isn’t a march towards full automation. A mature product can use several types at once: a tool for routine work, an adviser to explain, a collaborator to explore, and an executor only where decisions, controls and ownership are clear.

02 · Context

Context is something we design.

An intelligent system is only as good as the context it understands. In most businesses that context lives in people’s heads, documents and old email threads. I design it like any other part of the experience: explicit, reusable and governable.

SchematicHow shared context forms around a scenario
Every scenario draws on the same five layers of context. People, tools and agents around the edge read from that shared layer instead of keeping their own copies, so they all reason from the same picture.
IntentGoalsThe outcome the business wants, with success measures and a time horizon.
PossibilityScenariosPossible situations or courses of action to explore before committing.
InfluenceDriversThe pressures behind a decision: strategy, markets, customers and commercial levers.
GuardrailsPolicies and agreementsRules, obligations and supplier commitments that shape which choices are viable.
LimitsConstraintsShelf space, capacity, resources and what’s operationally feasible.
ContinuityCommitments and conflictsEarlier decisions, dependencies and exceptions that affect what can safely happen next.

Explainable

A recommendation can show what it took into account, so people can challenge it.

Governable

Policies and agreements are checked, not assumed, before anything acts.

Shared and remembered

One context layer and one memory across the platform, so every tool, person and agent works from the same picture.

03 · Principles

Four principles for every intelligent system.

Transparent intent

Show what it will do and why, before it acts.

Reversibility by default

Design the way back before the action.

The human stays in the story

People own the outcome and can explain it.

First-class failure

Honest failure states and a clean handoff to a person.

The principles in one pattern
  1. 1Transparent intent. It says what it will change, why, what it drew on, and how sure it is, before anything happens.
  2. 2Reversibility by default. The change is scheduled, not instant, with a clear way back.
  3. 3The human stays in the story. A named owner approves, amends or rejects, and can explain the decision.
  4. 4First-class failure. What it couldn’t do is stated plainly and handed to the right people.
Before building anything, ask
  1. What is the AI actually doing in this step?
  2. Which relationship is that, and is it the right one today?
  3. Who stays accountable for the outcome?
  4. What context does it need, and where does that come from?
  5. How does a person see, stop or undo what it did?
04 · In practiceHow I approach intelligent systems

Earn the trust first. Then automate.

Four habits I bring to every intelligent system I work on, whatever the domain.

Trust

Autonomy is earned in steps.

I match how much a system does to how ready the people, data and controls are. Not the other way round.

Systems

See the whole system.

I zoom from one screen to the whole platform and back, so decisions, patterns and teams fit together.

Users

Keep users in the room.

The people whose work is changing stay involved from research to release, not just at testing.

Focus

Start where it genuinely helps.

I look for the decisions where AI adds real value, and leave the rest alone until it’s ready.

What this means for UX

Design matters more when the software decides.

Once a system can recommend and act, the experience isn’t just the screen any more. It’s the relationship, the context and the controls. Those are design decisions, and if designers don’t make them, they get made by default.

Why our role matters more

Models are fast becoming a commodity. What sets a product apart is how well people understand, trust and steer what the system does. Who leads, who decides and how a mistake gets undone: that’s UX work.

How our process has to adapt

We map decisions, not just journeys. We prototype behaviour over time, not static screens, design for uncertainty and failure, and work with data science from day one. Research doesn’t stop at launch: we keep checking that trust is earned.

Two kinds of user

We still design for the needs of people. Now we also design for the needs of intelligent systems: the context they need to reason well, the rules they must respect, and the signals that tell them to stop and hand back.

People need
Clarity: what the system will do, and whyControl: to challenge, change and undoConfidence: when to trust it, and when not toOwnership: a clear line of who’s accountable
Intelligent systems need
Context: goals, drivers, policies and constraints made explicitBoundaries: what they can do alone, and what needs a personFeedback: outcomes and corrections to learn fromHandoffs: clear moments to stop and pass back