For the last fifteen years, software competed first on functionality, then on experience. Today it’s beginning to compete on something less visible but far more consequential: understanding.

This isn’t an incremental shift in enterprise architecture. It changes what UX actually is, where design creates value, and what it means for a company to differentiate in the age of AI.

In 2010, Dave McCrory introduced the idea of data gravity — as data accumulates, applications and infrastructure move closer to it. That single idea shaped an entire generation of enterprise software: cloud warehouses, lakehouses, ETL pipelines, modern BI, governance tools. For over a decade, the assumption was simple. Structure your data well, and you gain leverage.

But AI doesn’t just consume data. It consumes meaning. And that’s where the old assumption breaks.

A CRM records pipeline stages, not how your salespeople actually qualify a deal. A policy document lists rules, not when experienced people quietly ignore them. Humans have always filled these gaps — with instinct, culture, memory, the stuff nobody writes down. Software stored information. We supplied the understanding.

AI removes that. It can’t absorb tribal knowledge by osmosis the way a new hire does over six months of hallway conversations. It needs things made explicit. That single requirement changes everything downstream.

If data gravity shaped the last decade, contextual (or semantic) gravity will shape the next.

As organisations encode their operational meaning — definitions, workflows, decisions, institutional memory — into shared systems, AI capability starts aligning around that shared understanding. Whoever owns the context ends up owning the intelligence layer, because models are becoming interchangeable. Context isn’t.

We’re already watching this go wrong in a specific way. Every CRM, ERP, and support tool is bolting on AI, and each one is quietly building its own version of “the truth.” One system’s definition of “customer” isn’t another’s. Nothing breaks visibly — dashboards load, agents respond — but every AI is reasoning from a slightly different picture of the business. I’ve seen this happen inside client organisations long before anyone called it an AI problem: three teams, three spreadsheets, three “official” numbers, and nobody quite sure which one to trust in the room. AI just makes that old problem load-bearing.

This is context fragmentation, and it’s why semantic platforms are becoming inevitable rather than optional. Enterprises already centralised identity, infrastructure, and data. Understanding is the next layer.

There’s a deeper shift underneath all of this, though.

For thirty years, UX designed interaction with software:

Human → Interface → Application → Database

The interface was where meaning got translated, and design’s job was to make that translation invisible — fewer clicks, lower cognitive load, faster task completion.

The emerging architecture looks different:

Human → AI → Semantic Layer → Enterprise Systems

Someone simply states an intent — “show me every enterprise customer likely to churn because onboarding exceeded 45 days” — and the AI figures out where the data lives, what “enterprise customer” means here, which exceptions apply. The application becomes an execution endpoint inside a larger reasoning system, not the place where understanding gets built.

This is what I call Operational UX: designing how intent gets interpreted, how context gets retrieved, how confidence gets represented, how humans and AI teach each other over time. The product isn’t the interface anymore. It’s the organisation’s operational understanding — how it’s created, governed, and evolved as people and AI work side by side.

The designer’s central question changes too. It’s no longer just “can the user complete this task?” It becomes: does the organisation have enough shared understanding for AI to complete the task correctly — and how do we keep building that understanding as things change?

What happens to UI?

It doesn’t disappear. It still builds trust, communicates reasoning, gives people control over increasingly autonomous systems. But like cloud infrastructure today, good UI becomes table stakes rather than a source of advantage. Differentiation moves to how an organisation understands its customers, coordinates its teams, and evolves — and every interface simply becomes an expression of that.

Here’s a misconception I keep running into: that AI’s biggest contribution is efficiency.

Efficiency is valuable. It’s rarely a moat. Your competitors have access to the same models. If everyone gets equally fast, fast stops being a differentiator — it just becomes the price of entry.

The real opportunity isn’t doing today’s work faster. It’s discovering work worth doing that didn’t exist before. Every company that reshaped an industry started by asking “what’s now possible?” — not “how do we cut costs?”

Once an organisation has real operational context, AI starts surfacing patterns people miss: where work stalls, which approvals add nothing, which decisions consistently win. Efficiency is the first layer. Learning comes next, and evolution after that.

So can AI eventually match human judgement?

Depends what you mean by judgement. AI gets very good at recognising patterns and refining decisions within a known frame. What it doesn’t do is reframe the frame itself — question the assumption nobody’s questioned, notice the weak signal before it’s obvious, imagine something that isn’t in the data yet. That’s still where people create the most value.

Which raises an uncomfortable follow-up: if every company trains AI on similar data, similar KPIs, similar best practices — don’t they all start converging on the same answers? I think they will. Widespread AI adoption compresses variation across industries, not expands it.

Which means real differentiation has to come from somewhere else entirely.

Representational Gravity

Context explains how an organisation understands itself today. It doesn’t explain how an organisation imagines what it could become tomorrow. That’s a different force, and it’s the one I think matters most.

I call it representational gravity — the pull exerted by the mental models an organisation carries about reality itself. What market do we think we’re in. Who we think our competitors are. Which problems we consider worth solving. Which assumptions have simply never been questioned because nobody thought to.

Here’s the part that took me a while to really see: two companies can hold identical data, identical AI capability, identical semantic models — and still build entirely different futures, because they represent the world differently. One recruitment company sees itself competing on faster hiring software. Another sees itself redesigning how organisations discover talent in the first place. Same industry, same tools available to both, completely different trajectory — because the difference was never in the data. It was in what each company believed it was actually for.

This is the part strategy work keeps missing, in my experience. Most differentiation exercises start by asking companies to describe their market more precisely. But precision inside the wrong frame just makes you more efficient at the wrong thing. The real leverage point is earlier: what does this organisation believe is possible, and is that belief actually still true?

Representation isn’t a soft add-on to context — it’s the thing that decides what your context is even for. Data gravity organises information. Contextual gravity organises meaning. Representational gravity organises possibility itself — which futures feel available to an organisation before a single strategy document gets written.

Operational context stops AI from misunderstanding the present. Representational thinking is what stops an organisation from getting trapped by it. Get your context perfectly right and you’ll simply become world-class at executing yesterday’s idea of your business.

This is where I believe NAP belongs.

As AI reshapes how organisations operate, the challenge goes well past implementing models or redesigning screens. Organisations need a way to keep constructing and evolving their operational context — while also continually questioning the representation sitting underneath it.

That’s the layer we build for: a living operational model of how a company works and learns, paired with the harder, slower work of surfacing the assumptions a company doesn’t know it’s making. For startups, that means building this before complexity piles up. For enterprises, it means reconciling conflicting definitions across teams and asking, occasionally, whether the market they think they’re in is still the right one.

The organisations that lead in the AI era won’t just understand themselves precisely. They’ll keep rewriting what they believe is possible.

That’s the future we’re building NAP towards.