small_Blacktesty.png

thumbnail_Blacktesty.png

Blacktesty.png

large_Whitetesty.png

medium_Whitetesty.png

small_Whitetesty.png

thumbnail_Whitetesty.png

Whitetesty.png

Like most websites, we store cookies for analytics, and well, that's it. You can read all about this in our privacy policy.

# Culture & process
###### Embed commercially focused behaviours and mindsets

### Our approach
Culture and process are the foundation of sustained commercial success. We work with leadership teams to embed commercially-focused decision making, behaviours and mindsets throughout their organisations.

Our work focuses on three key areas: designing decision-making frameworks that prioritise commercial outcomes, building experimentation cultures that drive continuous improvement, and creating operating models that enable teams to deliver results efficiently.

We believe that the right culture and processes turn commercial strategy into sustainable competitive advantage. Our approach ensures that commercial thinking becomes embedded in how your organisation operates day-to-day, not just at the strategic level.

> What you can expect:
> - Decision-making frameworks that consistently prioritise commercial outcomes
> - Experimentation practices that drive data-led, customer-centred improvements
> - Operating models designed for efficient commercial delivery
> - Cultural change programmes that embed commercial mindsets across teams

From recipes to retail: driving commercial transformation for one of the World's most loved chefs

Bespoke statistical analysis: How we doubled refurbishment capital efficiency for this FTSE hotelier

Driving revenue growth: how we used multi-lever optimisation for a student property firm

Analysing customer behaviours: a recipe for success for this high street food retailer

Unlocking £5.8m revenue for a global fashion retailer through customer analytics

Inhousing & replatforming for a global retailer

Success Stories

The simple task your AI can't learn

Everyone's claiming they drove the upside. Who's right?

They don't like a cold night in Stoke

Go early or get left behind: four lessons for the agent economy

The Listening Gap: how brands forgot their customers, and why AI is about to call their bluff

Culture & process


*Your sales model predicts almost to the dollar. It has no idea what happens if you cut the ad budget.*

You contract a data science team, and they just shipped a sales model. It's trained on three decades of clean, complete data: every ad dollar spent, every site visit, every sale closed. Backtests are spectacular and it predicts next week's sales almost perfectly. An R² of 0.95.

So you ask it the only question you actually care about: what happens to sales if we cut the ad budget in half?

The model's answer: **nothing**. Not "a little." Not "it's uncertain." A confident, precise `0.00`. Move the budget up or down as far as you like, your near-perfect model insists sales won't budge.

This isn't a bug. It isn't bad data. It's what the best predictive model in the world will do, every time, on one of the simplest cause-and-effect chains in business. Here's why, and why more data does not help.

## A three-link chain

Everyone in your company knows how advertising works. You spend more, more people show up, and more of them buy. Cause, then effect, then effect. In the real world it's often more complicated, but the aim of this article is to show how such a simple problem of just A→B→C is unsolvable for standard predictive AI/ML. In this case it's Ad-Spend→Traffic→Sales.

![The chain every operator already understands: ad spend drives traffic drives sales.](/images/your-ai-cant-learn-chain.png)

Now here's the setup that breaks everything. Your team, doing the seemingly obvious and definitely standard thing: feeds the model both ad spend and traffic and asks it to predict sales. Why wouldn't they? More features, better predictions. And it works - the predictions are stellar. But include Ad-Spend and Traffic and your near-perfect accuracy predictive model insists revenue won't budge when you change Ad-Spend.

So the model predicts well - really well. But it also thinks that Ad Spend has no effect on Sales, even when it does.

Remember, again, you have decades of high quality data, and the most advanced predictive model.

## The model found a shortcut. It's allowed to.

Think about what the model sees. To predict sales, traffic is all it needs: traffic sits right next to sales in the chain. And once the model knows traffic, ad spend tells it nothing new. Every scrap of information spend carries about sales already traveled through traffic and arrived first.

So the model learns the efficient thing: read traffic, ignore ad spend. And judged purely on prediction accuracy - the only thing it was ever asked to do - that's not a mistake. It's the correct answer to the question it was asked.

The problem is you're about to ask it a different question.

## dual-card-section
card1Label: QUESTION 1 — ACED
card1LabelColor: "#1A4D5C"
card1Heading: "What will sales be next week?"
card1Body: A forecast. The model watches the world and guesses what comes next. Traffic is a legitimate clue.
card2Label: QUESTION 2 — FLUNKED
card2LabelColor: "#8A4526"
card2Heading: "What happens if we change the budget?"
card2Body: An intervention. You're not watching the world anymore, you're reaching in and moving a lever.
footer: Same model, two questions. It can ace the first and flunk the second.

When you drag the budget slider in a "what-if" tool built on this model, you're changing an input the model learned to ignore. Spend moves, traffic stays frozen at whatever value is plugged in, and predicted sales don't twitch. In the real world, cutting the budget would cut traffic, which would cut sales. Inside the model, the middleman never gets the memo.

> The better the model predicts, the more confidently it tells you your levers do nothing.

## More data won't save you. Better data won't save you. A bigger model won't save you.

This is the part that surprises even well-trained data scientists who are used to prediction, but not trained in the scientific method. It's actually a question I get a lot from data scientists following talks on causal inference. Surely with enough data, the model figures it out?

But the answer is a resounding "No". No amount of data will make a predictive model learn the correct answer to the right question.

In fact, more data makes the model *more certain* that ad spend doesn't matter, because in every single row, traffic already explains sales. A fancier architecture doesn't help either. You can't buy your way out of this with scale, compute, or a better vendor. The model is answering the question it was trained on, flawlessly. It was just never the question you needed answered.

And this trap isn't exotic. It's hiding in plain sight for any business decision-making problem.

And this is just for a simple three-variable problem. Most problems are more complex and have more causal pitfalls that fool predictive models into giving confidently incorrect answers, with high accuracy.

## What to do instead

The problem isn't the data, and it isn't the model's brainpower. A predictive model learns one equation: everything in, sales out. The chain - spend drives traffic, traffic drives sales - gets crushed into a single shortcut, and the lever falls out of it.

The fix is to learn the chain itself. That's what a structural causal model does: instead of one shortcut from inputs to output, it learns each relationship in the system - how spend moves traffic, and how traffic moves sales - as its own piece.

## dual-card-section
card1Label: PREDICTIVE MODEL
card1Heading: Learns one shortcut
card1Body: Spend plus traffic to sales, in a single equation. Accurate, but the chain is gone. Move spend and nothing propagates.
card2Label: STRUCTURAL CAUSAL MODEL
card2Heading: Learns every link
card2Body: Spend to traffic and traffic to sales, each learned separately. Move spend, and the effect flows through the chain, just like reality.

Same data. Same three columns. But now, when you drag the budget slider, the model does what the world does: traffic responds, and sales respond to traffic. It predicts about as well as before, and it can finally answer the question you actually had. Forecasts and what-ifs, from one model.

This is what we do at Zebra. We build structural causal models that learn how each of your variables drives the next, so the model traces a change through the chain instead of collapsing it into a single shortcut to the outcome. Your team can test a budget shift, a price change or a policy tweak before betting real money on it.

Your model isn't lying to you. It's answering a question you didn't mean to ask.



Earlier in the year we shared [a piece](/articles/cold-night-in-stoke) on using causal mapping on football: separating what pundits talk about from what actually drives results. A few people asked the obvious question. That's fun, but what's it got to do with my business? Fair. So here's the same problem, somewhere it costs real money.

Let's take a (fictional) mid/upscale hotel group. They've had a good quarter: forward bookings are ahead of plan, revenue is up, and the commercial team meets to work out what worked so they can do more of it.

By the end of the hour, three teams have each claimed the win:

1. Pricing says it was the new rate strategy. The numbers moved right after they changed approach.
2. Marketing says it was them. They retuned the campaign and the platform's been delivering ever since.
3. And someone quietly points out a third thing: last year's quarter was weak, so the comparison was always going to look good this time around.

All three are plausible. Each has a nice chart to back themselves. And here's the problem: add up what each team is claiming and it comes to more than the actual gain. They can't all be right. *But the room has no way to tell who is.*

## Three claims, three different bets

This matters because what you do next depends entirely on who actually drove it.

If pricing did it, you lean into the rate strategy and roll it out wider.

If marketing did it, you give them more budget and scale the campaign.

If it was just the weak comparison, you do nothing different, because there's nothing to repeat. The upside wasn't a lever anyone pulled. It was the base. And maybe you want to think about your metrics to strip out this effect.

So the claims aren't only a fight over credit. Each one points you at a different decision for next quarter. Scale pricing, fund marketing, or hold steady. Back the wrong one and you pour money into a lever that did nothing, and bank a win that was never going to repeat.

## Why nobody in the room can actually tell

The honest answer is that all three probably happened at once.

Pricing changed its approach in the same quarter marketing retuned the campaign, while you were lapping a weak base. Three things moving together, landing on one revenue line. You can't read three overlapping causes off a single chart. The line went up. It doesn't come with a breakdown of who moved it.

Each claim also has a trap of its own. The marketing one is a measurement trap: the platform reports the conversions it thinks it drove, but a platform claiming credit isn't the same as a platform causing the result. The pricing one is the in-house version of that, the team that changed something assumes the change is why, because the timing fits. And the base effect is the sneakiest, because it isn't really a cause at all, just last year's weakness flattering this year's numbers. Everyone gets to feel good, and nobody actually did anything.

## Same problem, simpler pitch

In our football series we saw how, across thousands of matches, the factors pundits talk about most; form, weather, travel, fixture congestion, barely moved results. The ones they rarely mention, the wage bill gap and the crowd, did most of the work. Confident and correct turned out to be different things.

Our hotel meeting is the same gap. The three claims on the table are the easy ones to reach for, the levers someone owns and can point at. The real driver might be one of them, a bit of all of them, or something nobody put on a slide at all: a shift in channel mix, a change in who's booking, a cohort behaving differently from last year's. None of it makes it into the room, because nobody can measure it.

## What telling them apart would take

You can start to resolve this without a model. The trick is to stop treating the three claims as a multiple-choice question with one right answer. The technique behind that is **causal mapping**, the same approach we used on the football data. It's just a way of laying out what you think drives what, so you can see where two claims are tangled together and what would tell them apart.

The first move is to notice they're tangled. If the rate change and the campaign retune landed in the same quarter, you can't credit either without separating them. That's called *confounding*, and it's why the chart can't settle the argument.

![Causal map of the same quarter: one push drove both the rate change and the campaign, so their effects can't be separated, while a weak prior year inflates the reported figure rather than causing real revenue.](/images/who-drove-the-upside-diagram.png)

The second is to ask what evidence would actually distinguish them. The base effect is testable: look at absolute booking pace against a normal year, not just the year-on-year line. The pricing claim is testable: did the gain show up specifically where the new rates applied, or everywhere, which would mean something else drove it. The marketing claim is testable: hold out a market or a spend level and see whether the upside holds without the platform marking its own homework.

None of that is exotic. It's the difference between a room that argues about who deserves the credit, and a room that asks what would have to be true for each claim to hold.

And that's what a causal map gives you. It won't hand you the answer, but it shows you which claims are tangled together and which ones you could actually test.

## The question to leave with

Next time the numbers are good and three teams claim the win, ask one thing before you reward anyone: can we actually tell who's right, or are we backing the team with the best story?

If you can't tell them apart, you're guessing. You might guess right. But you won't know that you did. You'll fund the wrong lever, scale a campaign that did nothing, and plan next quarter around a win that was never going to repeat. Then when it doesn't, you'll be in the same room again, working out who to blame.

Football has the excuse that it's only a game. Your P&L doesn't.


Who drove the upside?


Pundits have a stock phrase for the flair team who supposedly can't handle a wet, cold night under the lights in Stoke. We wondered how much of that kind of received wisdom actually holds up. So we took a full Premier League season, 2,140 matches, and mapped what really moves a result against what gets talked about.

## The picture is the point

Plot every factor two ways: how much media attention it gets, and how much it actually changes the outcome. You get four quadrants. Things that are talked about and matter. Things rightly ignored. And the two awkward ones: the myths, discussed constantly while doing nothing, and the hidden drivers, quietly deciding matches while nobody mentions them.

![Causal map of Premier League factors, plotting media visibility against actual causal effect. Form, weather and travel sit in the myth quadrant. Wage bill gap and crowd size sit in the hidden quadrant.](/images/cold-night-in-stoke-map.jpeg)

## What the data shows

A few things fall out.

Weather doesn't matter. We tested rain, wind, temperature and combinations of all three, and found nothing that survives scrutiny. The cold night in Stoke is folklore. The data doesn't show it.

The wage gap does the quiet work, and almost nobody puts a figure on it in the moment. Annual wage tables get their day of coverage, but match by match the commentary talks about "quality" without saying out loud that one side is paying far more for its squad than the other. It's the least discussed of the big factors and it does more work than any of them.

## Where we stopped ourselves

Kickoff time is where we had to stop ourselves. The late television slots look like they carry less home advantage, which would have been a tidy finding. But broadcasters pick the stronger away sides for those games, so the slot and the quality of the opponent are tangled together. We can't separate them, so we don't claim the effect. That restraint is the whole discipline. The moment you can't tell two explanations apart, you say so, rather than crediting the one you prefer.

Form is the one we're most careful about. Momentum is the pundit's favourite word, and hot streaks do look a lot like illusion, with teams drifting back to their level. But a few cold-run examples made us hesitate before writing the whole thing off. So we treat it as a question worth testing properly, not a result. We'd rather under-claim and be right.

## The gap is the message

The map's real message is in the gap. The biggest bubbles, the factors that get the most airtime, sit on the wrong side of the picture. The things that decide matches sit quietly over on the right with barely a mention. That distance, between what gets said and what the evidence shows, is the gap between received wisdom and proof.

## It's not really about football

None of this is really about football. Every business has the same map. There's a version of the cold night in Stoke in most Monday meetings: a comfortable story everyone repeats that the evidence never supported. Forget which pundit is right. The question worth asking is whether the things you believe about your own numbers would survive being plotted like this.

One note on rigour. This is the accessible version, done to make the idea land, not to survive peer review. Taking the same discipline to publication standard, where every claim has to hold up to scrutiny, is a harder craft again. That's the work our causal scientists do, and it's where the real proof lives.


A cold night in Stoke

# Introduction: The Agent as Customer's Advocate

In my [last article](https://www.zebra.associates/blog/article/the-listening-gap) I set out some different routes that brands may take in response to the rise of agentic search. Here I’m going to stand back a little and consider what these changes may mean for brands and eCommerce specifically.

While nascent it’s pretty clear that ‘Agent mode’ is here to stay, and from here user expectations for how it should feel are going to grow. Less Minority Report, more Her.

Today this agent layer is vanilla, objective, cold. 

As we get more expert at interacting with them - and they get better at listening to us - this will change.  

In this world the agent becomes not just a passive gateway, but an active **advocate and filter for the customer**. Consider this future interaction:

“Place a weekly shop at Tesco. Read me my favourites for me to choose which I want to add. Then let me know what’s on offer that I might consider adding”

Here the agent has native access to my account at [Tesco.com](http://tesco.com) (ideally not via an agentic browser). It then retrieves first a list of favourites, then polls a service for suitable offers, curating them for me.

This future of eCommerce may seem complex and perhaps far-fetched, but it will be with us soon. The question is how brands should equip themselves, moving beyond traditional marketing to a world where they are equipped to communicate with (and even negotiate through) the agent, at scale.

Let’s think through what this might entail for brands.

# The New Inbound: Engineering for Experience

I’ve previously suggested there may be a dividing line between experiential and functional eCommerce. On both sides of the divide however the brand will have a requirement to ensure their core content is agent optimised or at least agent readable. Content here could include product specifications, descriptions, reviews. Given advances in computer vision, images and video too.

We could define this as the inbound content strategy. Its purpose is to ensure the agent can effectively find and present the brand’s best qualities when the user is actively searching for a product that might match.

Part of this is undoubtedly the place of GenAI Engine Optimization (GEO) - a term that’s mildly horrific to anyone like me old enough to remember the days of doorway pages, the rise of social optimization etc.

But it’s also an engineering challenge, to build the pipelines and systems that allow for seamless and efficient communication between the brand's data and the AI agent:

1. API Development & Maintenance: Engineers will be responsible for creating and maintaining robust APIs that provide agents with direct access to a brand's product catalog, customer data, and content. These APIs must be fast, reliable, and well-documented.
2. Structured Data Implementation: While marketing defines the content, engineering implements the schemas (e.g., Schema.org) and data structures that an agent can parse. This ensures that the rich metadata created by the marketing team is correctly formatted and delivered.
3. Dynamic Negotiation Systems: the team will need to build systems that can receive requests from an agent and, based on predefined rules or machine learning models, offer dynamic pricing, bundles, or add-ons in real time.
4. Data Feeds & Integrations: Engineers will manage the systems that deliver updated product information, inventory levels, and pricing to agents in a timely and efficient manner, often through continuous data feeds.

By adopting these engineering practices, a brand becomes the most compelling option not just because of its products, but because it is the most discoverable and intelligently articulated choice for a user's agent.

# The New Outbound: Communicating with the Agent

I’ve referenced an idea for a ‘dynamic negotiation system’ above as an engineering challenge. It’s also of course a commercial one.

We’ll need to decide how brands should ‘market’ to the agent. In this new frontier of outbound I see two key modes:

1. Notifications: a brand could send a notification to a customer’s agent about a relevant offer or new product. Here the agent may be in ‘sentry mode’, and given a brief to minimise noise may simply decide not to let irrelevant communications through to the user. So brands will need to work much harder to personalise this kind of promotion to the right customer: personalisation will need to get, well, personal.
2. Negotiations: here the brand’s system could interface with the agent to dynamically offer a better price, or potentially even to intervene in real time to win a competitive transaction. This moves from a static offer to a dynamic, programmatic negotiation. The implication is a far more complex set of pricing/discounting rules, able to evaluate net revenue considering a variety of variables - including, probably more importantly, the customer value to the brand.

# The Metrics of Trust and Negotiation

In this world of fluid inbound and outbound communications between users, their agents and brands we’ll need to start considering new types of business metrics that reflect the success of the relationship:

- For instance we may want to monitor the offer acceptance rate: the % of agent-relayed offers that result in a purchase.
- Or the profitable negotiation success rate: % transactions negotiated between a brand’s system and an agent at sustainable margin
- Agents themselves may start to build metrics that allow them to make better choices for their users: a ‘trust score’ based on a brand’s ability to serve relevant inventory or provide fair pricing for example

These next generation metrics will by their nature be more complex to operate than today's eCommerce metrics. Unlike click-through rates or conversion percentages, they'll require sophisticated data pipelines that can track multi-step negotiations and evaluate the quality of agent relationships over time. Early adopters will need to experiment with proxy measures whilst the industry develops standardised benchmarks - much as we did in the early days of social media marketing.

# 4 lessons for brands to get ahead of this curve

“Predictions are difficult, especially about the future” _Niels Bohr_

OK so some of this is a little guess work about where we are headed. But parts of it are happening right now, in beta features and early implementations. 

4 lessons for brands then:

1. Go early to get ahead: build these capabilities today and get ahead as consumer expectations evolve
2. Break down silos before they get broken down for you: view this transition not as a threat to traditional marketing channels, but as an evolution that requires new forms of partnership between marketing, engineering, and commercial teams
3. The human touch will be ever more valuable: while we may choose to hand off tasks to technology the brands that thrive will be those that can seamlessly blend human creativity with machine efficiency, creating experiences that feel both personal and scalable. The future belongs to brands that can speak fluent 'agent' whilst never losing sight of the human they're ultimately serving.
4. Product + trust + efficiency = value: in this new landscape, the strongest brands won't just be those with the best products, but those that become the most trusted and efficient partners for their customers' digital advocates. 

The question isn't whether this future will arrive, but whether your organisation will be ready when it does.


Last month, I needed to find a lightweight, battery-operated lawnmower for my elderly neighbours. Something easy to manoeuvre, reliable, and available for collection near their home within the week.

What followed was a depressing tour of generic ecommerce experiences. Endless filtering through "corded vs cordeless," scrolling past petrol-powered machines they'd never be able to start, comparing weights buried in specification tables, and manually checking store availability for each model. Not one site asked the obvious questions: Who's using this? What's your situation? What actually matters to you?

Twenty minutes later, I described the same need to ChatGPT. It immediately grasped the context—elderly users need lightweight, easy-start options—and provided three specific recommendations with reasoning, weights, and suggestions for where to find them locally.

The contrast was embarrassing. An AI with no memory of my previous searches understood my parents' needs better than retailers who've been serving customers like them for decades.

## UX has become lazy

This isn't an isolated failure. Somewhere along the way, we stopped asking "what does this user actually need?" and started asking "what colour should the button be?"

User experience design has been reduced to selecting from pattern libraries and copying successful templates. Teams now download Shopify themes, tweak the colours, and call it user-centred design. We've convinced ourselves that because ecommerce patterns are "established," we no longer need to understand our users.

The evidence is everywhere. Fashion retailers serve the same "new visitor" experience to someone buying their fifth black jumper. B2B suppliers make electricians navigate the same generic category pages as DIY enthusiasts. Luxury brands treat repeat customers identical to bargain hunters browsing for the first time.

We've traded user understanding for interface consistency, and called it progress.

## The death of user understanding

This laziness didn't happen overnight. As UI patterns became commoditised, product teams pushed user research further into the background. "It's just ecommerce" became the rallying cry—as if understanding your customers was optional once you had a search bar and some filters.

As Jakob Nielsen observes, "UX practitioners today echo the complacency of their predecessors in the mid-90s. They seem content to tinker with their personas and the deluge of deliverables from our constantly expanding methodology toolbox while designing yesteryear's UX instead of focusing on next-generation UX."

Take any major retailer. A customer buying winter coats gets the same faceted search experience whether they're a new parent looking for something practical, a cyclist needing weather protection, or someone replacing a coat they've owned for years. The system knows nothing about context, nothing about intent, nothing about the person.

We justified this by calling it "scalable UX." In reality, we'd simply stopped listening.

The principles of interaction design—understand the user, identify their mental models, create systems that match how they think—became afterthoughts. Personalisation became a "nice to have" bolt-on feature, not fundamental to how we design experiences.

Most ecommerce sites make no meaningful effort to understand users beyond remembering whether they shop in men's or women's clothing. That's the extent of our personalisation ambition.

## The cliff edge approaches

Here's the problem: users are about to remember what good service feels like.

AI agents are teaching people to expect systems that actually listen. When someone asks ChatGPT for help, it asks follow-up questions. It remembers context. It tries to understand what they actually need, not just what they typed.

These aren't abstract future possibilities. Right now, someone can ask an AI agent: "I'm an electrician. I need a consumer unit from Screwfix that's available for collection within 10 miles of SG13 8AZ tomorrow. It needs to be at least 10-way and IP2XC rated. Give me options and help me compare."

The agent delivers a complete report with recommendations, availability, pricing, and the option to place a click-and-collect order.

Compare that to the current Screwfix experience: search "consumer units," scroll through generic results, manually filter by IP rating (it's the 10th filter in a long list), check boxes for 10, 11, 12, 13, 14 ways, then click each product individually to check store availability.

The difference isn't just efficiency—it's that one system tries to understand what you need, while the other makes you navigate its database structure.

## The reckoning

Users' expectations are shifting rapidly for the first time in years. Google results that seemed perfectly adequate suddenly feel "comically irrelevant and one-dimensional" compared to conversational search. The same shift is coming for ecommerce.

When AI agents can provide genuinely personalised shopping experiences—understanding context, remembering preferences, asking clarifying questions—your generic category pages will look embarrassingly tone-deaf.

This isn't about technology replacing humans. It's about technology finally doing what we should have been doing all along: listening to users and responding to their actual needs.

## Know what game you're playing

The choice isn't simply "understand users or die." It's more nuanced: **decide what kind of commerce you're in, then optimise accordingly.**

**Functional commerce:** If you're selling solutions to problems—tools, supplies, utilities—accept that you're building for efficiency, not exploration. Users want to get things done and leave. AI agents will handle these transactions because they're fundamentally chores. Your job is to make your products and services easily accessible to these agents through APIs, structured data, and clear specifications.

**Experiential commerce:** If you're selling products that users enjoy choosing—fashion, lifestyle, luxury—then genuine user understanding becomes your only sustainable advantage. When users choose direct interaction over agent delegation, it's because they value the experience itself. Generic UX patterns won't compete with truly personalised, contextual design.

The Screwfix example reveals the deadly middle ground: they're clearly functional commerce—electricians aren't browsing for fun—but they've built consumer-facing UX that forces professionals to navigate irrelevant faceted search. They think they're in the "browsing" business when they're actually in the "procurement" business. That's why the AI agent easily outperforms their own website.

**The battleground:** Many purchases could swing either way. Brands with genuinely personalised, user-centric experiences might keep users engaged directly. Those with lazy, pattern-library UX will lose users to agents who understand their needs better than the brand's own website does.

The reckoning hits hardest in this middle ground—where brands pretend they're experiential while delivering commodity UX.

## Beyond the grid

The solution isn't to panic-buy AI features or bolt chatbots onto existing experiences. It's to honestly assess what game you're playing, then design accordingly.

**For functional commerce:** Stop pretending users want to "discover" your industrial equipment or professional supplies. Build efficient, data-rich interfaces that agents can navigate easily. Invest in APIs, structured product information, and clear specifications. Make it trivial for an AI to understand your inventory, pricing, and availability.

**For experiential commerce:** This is where user research and genuine personalisation become essential. When users choose to engage directly rather than delegate to an agent, they're telling you the experience itself has value. Generic UX patterns won't compete with brands that genuinely understand what makes each customer tick.

Start asking the right questions again. What does this specific customer need right now? Are they here to get something done or to enjoy the process of choosing? How can we serve them better than a generic experience—or better than an AI agent could?

The brands that survive the next wave won't be those with the most sophisticated AI—they'll be those that understand which game they're playing and design brilliantly for it.

The grid was never the problem. Our willingness to accept it was.

The reckoning is coming. The question isn't whether AI will change customer expectations—it's whether you'll be ready when it does.


Culture & process

Embed commercially focused behaviours and mindsets

Our approach

Customer & product strategy

Conversion optimisation

Proposition & experience