The market is catching up to context

There are moments when a market does not suddenly change direction, but suddenly starts putting words on something that was already happening.

This week feels like one of those moments.

Satya Nadella’s recent message about AI has been discussed widely because it points at a problem many companies are starting to feel: the winning move is not simply to rent the strongest frontier model and call that a strategy.

The model matters.

But the model is not the company.

The company is the work it has done, the decisions it has made, the mistakes it has survived, the knowledge it has accumulated, the context it has preserved, and the way people inside it learn together over time.

That is the part you cannot buy back later from an external model provider.

And this is exactly where the AI conversation is now moving.

Not away from models.

Beyond models.

AI adoption is no longer the interesting question

For the last two years, a lot of the public conversation has been about adoption.

Which model should we use?

Which chatbot should employees have access to?

Which provider is ahead this month?

Which benchmark changed last week?

Those questions still matter, but they are no longer enough.

Europe is now pushing hard to become an AI continent, with large-scale investment in compute, AI factories, data, skills, and adoption across strategic sectors. Norway is moving in the same direction through KI Norge, the AI Act implementation work, and a clear public ambition to support responsible and innovative AI use in both the public and private sector.

This is important.

It means AI is leaving the experiment corner.

It is moving into normal work.

When that happens, the problem changes.

A prototype can live with a clever prompt.

A real organisation cannot.

A prototype can accept that nobody remembers exactly why the AI answered the way it did.

A real organisation eventually needs to know.

A prototype can copy and paste context from documents, tickets, meeting notes, and Slack threads.

A real organisation needs a governed way to decide what context matters, what is safe to use, what is outdated, what was ignored, and what should become part of the record.

That is the shift.

The question is no longer only:

> How do we use AI?

The better question is becoming:

> How do we give AI the right organisational context, and how do we keep control of what happens next?

Europe’s AI opportunity is not only compute

I am happy to see Europe investing in infrastructure.

We need compute. We need sovereign capacity. We need European AI companies. We need startups, researchers, industry, and the public sector to have better access to serious AI capability.

But Europe’s real opportunity is not to imitate Silicon Valley blindly.

Europe has something else.

Complex industries.

Regulated environments.

Strong public institutions.

Engineering-heavy companies.

Deep operational knowledge.

High expectations for trust, accountability, privacy, and documentation.

In other words: Europe has exactly the kind of environments where AI cannot just be a magic text box.

It has to be connected to context.

It has to be explainable enough to be reviewed.

It has to respect boundaries.

It has to produce proposals that humans can understand, edit, approve, reject, verify, and trace later.

That is not a weakness.

That could become Europe’s strength.

The next wave of AI adoption will not only be about who has the biggest model.

It will be about who can connect AI safely to real work.

Norway has the same pattern, just closer to home

Norway is now building its own AI coordination and governance layer through KI Norge.

That matters.

Not because a national initiative magically solves AI adoption.

It does not.

But it signals that AI is becoming serious infrastructure for society, business, and public administration.

And when AI becomes infrastructure, the boring questions become the important ones.

Who is responsible?

Which data was used?

Which rules apply?

What was the purpose?

What changed?

Who approved it?

Can we prove it later?

This is where I think Norway has a genuine opening.

Norway is not going to win by pretending to be the largest AI market in the world.

Norway can win by being very good at responsible, practical, high-trust AI in real industries.

Energy.

Shipping.

Maritime operations.

Aquaculture.

Public services.

Health.

Industrial projects.

Research.

Engineering.

All of these fields have the same underlying problem: the work is complex, the context is distributed, the history matters, and the cost of losing traceability is high.

That is exactly where AI needs something more than a prompt.

This is what Nolta has been positioning itself to become

When I started building Nolta, the first concrete product was a digital thread platform.

The original idea was simple enough:

Help teams understand how real-world projects evolve over time.

What exists.

What changed.

Why it changed.

How decisions, risks, assets, comments, relationships, and history connect.

That was already useful.

But as the product evolved, and especially as AI became more present in real work, the larger shape became clearer.

Nolta is not only about project history.

Nolta is about living context.

A structured, time-aware layer around complex work.

A place where humans, software, and AI can understand not just what is true now, but what changed, why it changed, what context was used, and what should happen next.

That is why Nolta is moving toward Trust & Execution.

Context before execution.

Scoped packages instead of loose prompts.

Structured proposals instead of uncontrolled output.

Human review before action.

Approved actions instead of silent mutation.

Verification after approval.

A timeline that remembers the whole loop.

The principle is simple:

> AI proposes. Humans approve. Nolta applies and remembers.

That sentence has become the spine of the product.

The market is now saying the quiet part out loud

For a while, the AI conversation was dominated by model capability.

Now the next layer is becoming visible.

Companies are asking different questions.

How do we prevent AI from becoming another disconnected tool?

How do we stop losing institutional knowledge into prompts that disappear?

How do we avoid giving external systems the keys to our organisational memory?

How do we keep employees in the learning loop instead of reducing them to data providers and reviewers of mysterious output?

How do we make AI useful without making work less understandable?

This is why Satya’s message landed.

It names something many people were already starting to feel.

AI cannot be treated as a replacement for organisational learning.

It has to become part of the organisation’s learning system.

And for that to work, companies need a way to preserve and govern their own context.

That is the space Nolta has been moving toward.

Not because the trend appeared this week.

Because this week made the trend easier to see.

The missing layer is context infrastructure

Most companies already have fragments of context.

They have documents.

They have tickets.

They have project plans.

They have Teams channels.

They have Confluence pages.

They have emails.

They have meetings.

They have screenshots.

They have spreadsheets.

They have decisions hidden in conversations.

They have risks that were once obvious and later forgotten.

The problem is not that context does not exist.

The problem is that it is not shaped for use.

Especially not for AI.

AI does not know which old decision still matters.

It does not know which comment was superseded.

It does not know which asset is central and which one is peripheral.

It does not know which risk was accepted, which one was mitigated, and which one was simply forgotten.

It does not know what the team would consider enough evidence.

Unless the organisation has a way to tell it.

That is context infrastructure.

Not another chat window.

Not another project management board.

Not another document repository.

A living layer that helps work keep its memory.

This is why I am more convinced than ever

I do not think Nolta’s timing is perfect because AI is fashionable.

Fashion is dangerous.

I think the timing is right because the deeper AI problem is becoming unavoidable.

As AI moves from experiments to operations, companies will need more than access.

They will need control.

They will need memory.

They will need provenance.

They will need review.

They will need verification.

They will need to know what context went into a piece of AI work, what came out, what humans changed, what was approved, and what happened afterwards.

That is not a side feature.

That is the foundation for trustworthy AI in complex work.

The market is starting to see it.

Europe is organizing around AI adoption and sovereignty.

Norway is building national structures for responsible AI.

Global technology leaders are warning that companies must preserve their own learning systems and not become dependent on external intelligence alone.

And Nolta has been quietly moving toward the same conclusion:

The future of AI in real organisations is not just bigger models.

It is better context.

It is governed execution.

It is human-approved action.

It is memory that compounds.

That is what Nolta is being built for.

Context matters.