The Era We Just Left
For over three decades, we’ve been living in we call the era of classical computing.
It began with the graphical user interface going mainstream, accelerated through the web, exploded with smartphones, and culminated in the app economy. Throughout this entire period, despite extraordinary innovation, one fundamental dynamic remained constant:
The human did all the translation work.
We learned to navigate file systems. We memorised keyboard shortcuts. We understood menu hierarchies. We filled forms correctly. We structured search queries effectively. We downloaded apps for specific tasks. We configured settings. We learned new interfaces every time we adopted new software.
Every interaction required us to take our messy, contextual, human intent and translate it into something the machine could understand. The machine sat passively, waiting for correctly formatted instructions.
The paradigm was simple, even if we rarely named it:
Human interprets own needs → Human translates to machine language → Machine executes
This pattern became so ubiquitous that we stopped noticing it. We called it “using technology.” We labelled people who were good at it “tech savvy.” Entire industries emerged to make the translation easier: user experience design, interface development, interaction patterns, design systems.
But here’s the provocation worth sitting with:
The entire field of user experience design has essentially been compensation for machine stupidity.
We built buttons because machines could not understand intent. We built forms because machines could not extract structured data from natural communication. We built dashboards because machines could not synthesise and prioritise information for us. We built apps, millions of them, because machines could not understand which capability we needed when.
Every pixel of interface has been a bridge across a comprehension gap.
The gap was always on the machine’s side.
What We’re Unlocking
Large language models and the broader emergence of AI have changed something fundamental. Not incrementally. Structurally.
For the first time, machines can interpret unstructured human input: voice, text, images, video, and derive intent without requiring us to format it first. They can hold context. They can reason well enough to be useful. They can connect to systems and take action.
This shift did not arrive because we suddenly imagined a better future. It arrived because long standing constraints finally collapsed.
Classical computing was limited by representation. Machines could only act on what humans explicitly structured for them. Intent had to be flattened into menus, fields, schemas, and commands. That limitation shaped everything downstream, from interfaces to organisations to job roles.
Intelligent computing emerges when representation collapses. When language, images, and behaviour themselves become usable inputs. When understanding no longer depends on predefined structure. At that point, interpretation stops being a feature and becomes the foundation.
This is why the shift feels sudden. It isn’t. It was blocked by capability, not imagination.
This enables an inversion of the paradigm:
Machine interprets human context → Machine translates to execution → Human validates or adjusts
This is not a UX improvement. It is a fundamental restructuring of where the cognitive burden sits.
I call this the inversion of the burden.
We are moving from classical computing, where machines were execution engines waiting for instructions, to intelligent computing, where machines are interpretation engines that derive intent and act on it.
The shift is profound, but it is happening quietly. That quietness is one of the reasons it is still widely misunderstood.
Why This Shift Is Invisible
When action driven technology improved, you could see it. A new button. A faster animation. A cleaner layout. A new app icon on your home screen.
When interpretation driven technology improves, you simply notice things working better. There is no visual artefact of “the system understood what I actually meant.” The experience changes. The screen does not.
This creates three structural blind spots.
First, we are measuring the wrong things. The technology industry still celebrates visible innovation: new apps, new features, new interfaces. Progress is measured through downloads, engagement, and screen time. But interpretation quality, context depth, and anticipation accuracy are largely invisible. There is no leaderboard for “understood me better this week.”
Second, the shift is experienced, not observed. You cannot screenshot it. You cannot demo it cleanly. You simply find friction disappearing. Answers arrive earlier. Work takes fewer steps. Entire categories of effort quietly vanish.
Third, our psychological model lags reality. We have spent decades thinking of technology as a tool we operate. The mental model of “I do things to the computer” is deeply embedded. Even when people interact with interpretation first systems, they often frame them as better search boxes or writing assistants, forcing genuinely new capability into old conceptual containers.
The Three Phases of Intelligent Computing
This shift is not binary. It is unfolding in phases.
Phase 1: Interpretation Systems understand unstructured input. Humans still initiate, but no longer need to speak machine. This is where most organisations are today.
Phase 2: Anticipation Systems persist context. They prepare before being asked. Research appears ahead of meetings. Anomalies surface before humans think to look. Initiation begins to shift.
Phase 3: Agency Systems act within defined boundaries. They do not just prepare. They execute. Humans become directors and validators rather than operators.
We are firmly in Phase 1, with Phase 2 emerging quickly in more sophisticated environments. Phase 3 is arriving faster than most organisations are psychologically or structurally prepared for.
What This Means for Builders
If you are building technology, or deciding where to invest, this shift changes the calculus entirely.
In classical computing, value accrued to the interface. The best apps won. The slickest UX won. The interface was the product.
In intelligent computing, value accrues to the intelligence layer. Context, understanding, and anticipation become the product. The interface becomes thin and increasingly interchangeable.
Many organisations are responding by adding intelligence to interfaces rather than removing interfaces in favour of intelligence.
They embed AI into existing workflows, screens, and forms. The system feels smarter, but the human still navigates, initiates, and translates.
This creates a dangerous illusion. The product demos well. The organisation feels modern. But the burden has not actually moved.
That is not intelligent computing. It is classical computing with better autocomplete.
The Business Case Reframed
Most leadership teams are still evaluating intelligent computing with classical logic.
They ask what the ROI is. What it replaces. What the feature set looks like.
These questions are not wrong. They are anachronistic.
Automation optimises known processes. Agency creates capability under uncertainty.
Most ROI frameworks are designed for automation. Intelligent computing delivers its real value through agency.
This is why intelligent systems often feel hard to justify early. You are not replacing a role. You are creating capability that previously did not exist or was too expensive to sustain.
Intelligent computing does not make organisations smarter by default. It exposes whether they ever were.
What This Means for Roles and Work
For over three decades, professional value has been closely tied to translation capability.
As the burden inverts, the nature of valuable work changes.
What decreases in value:
Translation
Navigation
Execution
Manual information gathering
What increases in value:
Direction setting
Judgment
Context provision
Trust calibration
Meaning making
A simple diagnostic helps clarify this shift.
If most of your time is spent translating information between systems or executing predefined steps, the inversion will compress your role.
If most of your time is spent deciding direction, exercising judgment, and interpreting outcomes, the inversion will amplify it.
That ratio matters more than your job title.
This does not mean mass unemployment. Organisational redesign is the cost of adopting intelligent systems.
The Psychological Resistance
There is a deeper reason this shift creates discomfort.
For decades, action driven systems reinforced a particular story. Humans were in control. Technology served. Every click was a small assertion of agency.
Interpretation driven systems challenge that story. If systems understand, anticipate, and act, what is left for us?
Direction. Judgment. Meaning.
These are not minor contributions. They may be the most essentially human ones. But they are harder to narrate to ourselves than “I pressed the button.”
Some of the resistance we see is not technical. It is existential.
The Questions That Matter Now
Most organisations are still asking which AI tools they should buy.
The question that matters is what intelligence architecture they are building.
Models are commoditising. GPT-5, Claude, Gemini will all be capable and widely available. Access is not the advantage.
The advantage is:
Context
Integration
Orchestration
Governance
These are organisational capabilities, not software purchases.
For individuals, the question is not how to use AI tools.
It is what your irreducibly human contribution is, and how you develop it.
Where We Are
We are in the early stages of a shift from classical to intelligent computing.
The translation burden is inverting. Machines are becoming interpretation engines.
This changes what is valuable to build. This changes how businesses create value. This changes the nature of work.
The shift is quiet, but it is not slow.
The burden has inverted.
The question is whether you have noticed, and what you are going to do about it.
Craig Hepburn is an AI Strategist & Builder, Perplexity Fellow and Former Chief Digital Officer, Art Basel & UEFA, helping organisations build useful intelligence for the age of intelligent computing