Information systems were designed to store, classify and retrieve. For decades, that logic gave us databases, pages, folders, forms, tickets and workflows. But AI moves the problem elsewhere: the point is no longer only to file information away. It is to turn that information into active memory — memory that can connect, interpret and reuse what the organisation has already worked out.

The problem is not that information is badly filed

In a lot of companies, people still treat the information problem as a filing problem.

What we need is a better knowledge base. A better tree structure. A better naming scheme. A better search engine. A better documentation tool.

Sometimes that is true.

But it is not the heart of the matter.

The real problem is not only that information is scattered. The real problem is that it usually sits in a shape that does not match the way we think, decide and learn.

A Notion page does not think. A database does not reason. A Jira ticket does not build up Context. A document has no idea it contains a decision you will need again in three months.

These objects store information. They build no memory.

And that is exactly where AI changes the nature of the problem.

We organised information the way machines do

Databases, pages, forms and folders are not natural shapes of thought.

They are technical answers.

We organised information that way because we had no real choice. It had to be stored somewhere. Fields had to be created. Search had to work. Whatever came into the company had to be given a stable shape.

But nobody naturally thinks in tables, columns, filtered views or nested pages.

When you work on a subject, you are not just handling documents. You are handling goals, decisions, risks, contradictions, assumptions, evidence, memories, weak signals, unwritten rules, examples, hunches.

Our heads work through links, priorities and contexts.

The classic information system works through objects.

That gap is becoming more and more visible.

The classic system stores. Context builds up capital.

A classic information system is very good at keeping data.

It can tell you: here is the ticket, here is the customer, here is the date, here is the status, here is the person it is assigned to.

That is useful. But it is not enough.

Because much of the value is not in the object itself. It is in the reasoning around the object.

Why does this ticket matter? Which product tension does it belong to? Which past decision does it echo? Which weak signal does it confirm? Which contradiction does it expose? Which idea could be reused later?

The classic system keeps the event.

Context keeps what the event taught us.

That is a major difference.

When you work with Context, a piece of customer feedback is not just filed under a category. It can become a piece of evidence. An objection. A friction point. An atomic note. A decision to revisit. An example for sales. A signal for product. A wording problem to clear up. A task outside the current focus, to be parked in a todo.

The same piece of information can exist along several axes.

That is what classic systems handle badly.

They often force you to pick one shape too early.

The future of the information system is not a better form

There is a lot of talk about AI inside existing tools.

A chatbot in support. An assistant in the CRM. Automatic generation in the back office. A summary in the ticketing tool.

That is useful.

But it is probably just a staging post.

The future of the information system will not only be a form filled in automatically by an AI. It will not only be a conversational search engine sitting on top of existing databases.

The real change runs deeper: moving from a system organised around applications to a system organised around Contexts.

A product Context. A support Context. A customer Context. A marketing Context. A source code Context. A documentation Context. A standards Context. An opportunity Context.

Each Context holds its own sources, rules, decisions, notes, goals, limits and links.

It does not necessarily replace existing tools. It cuts across them.

It turns what was scattered into memory you can actually draw on.

Memory becomes the asset

The classic way of using AI: you hand over some documents, you ask a question, you get an answer.

Then you start again.

You reload the sources. You re-explain the Context. You restate the constraints. You go back over decisions already made. You lose part of the previous reasoning.

That works well enough for producing a deliverable.

It is weak for building anything up.

The point is not only that the AI answers well today. The point is that it can answer better tomorrow because the Context has grown richer.

That is where memory becomes central.

Not the magic memory of a chatbot. Not a conversation history locked inside an interface. A real working memory: readable, inspectable, editable, versionable, reusable.

Files. Notes. Decisions. Directives. Templates. Logs. Syntheses. Links.

What matters is not only what the AI produces. It is what it lets you keep as capital.

Markdown is not the point. Portability is.

You can use Markdown, JSON, a document store, a graph, a RAG setup, a vector database or something else entirely.

The format is not the heart of the matter.

The heart of the matter is that memory must not be locked inside a tool.

If an organisation's working intelligence is stuck in chat histories, it has not really been built up into anything. It is useful as long as the interface exists, as long as the subscription is live, as long as you can find the right conversation again, as long as you remember where you discussed what.

That is not a company memory.

It is a trace of usage.

Useful memory has to be readable by a human, picked up by another AI, moved, backed up, enriched, audited.

That is why simple formats are worth so much. Not because they are elegant. Because they make knowledge portable.

The AI can change. The model can change. The interface can change. The Context has to stay.

The information system becomes a network of Contexts

Tomorrow's information system probably will not look like one huge central brain.

It will look more like a network of specialised Contexts.

Each Context will have its own logic.

Support does not need the same memory as product. Marketing does not need the same breakdown as engineering. The leadership team does not need the same level of detail as the people delivering. A customer Context does not do the same job as a standards Context or a source code Context.

But these Contexts will have to talk to each other.

A product Context will be able to draw on signals from support. A marketing Context will be able to pull in objections from sales. A training Context will be able to lean on the documentation and on how the product actually behaves. A customer Context will be able to connect tickets, history, usage, risks and commitments. A source code Context will be able to serve as the source of truth for regenerating documentation.

The point is no longer for each team to own its own store of information.

The point is for each team to contribute to a shared memory without losing its own professional angle.

That is an important difference.

The goal is not to merge everything. The goal is to make Contexts connectable.

AI does not make data matter less

You might think AI will make information systems less structured.

I think the opposite.

The more we use AI, the more the quality of the Context will decide the outcome.

An AI plugged into a pile of documents produces fragile answers. It summarises what it finds, but it does not always know what is true, out of date, urgent, approved, speculative or contradictory.

An AI plugged into a structured Context can do far better.

It can know which sources are reliable. It can find past decisions. It can apply directives. It can tell raw material from interpretation. It can connect several signals. It can produce an answer that fits the current focus.

So AI does not do away with the governance of information.

It makes it matter more.

Sources of truth matter more than ever. Naming rules matter more than ever. Explicit decisions matter more than ever. The limits of a Context matter more than ever.

Documentary chaos does not turn intelligent because you add a language model to it.

It just gets faster at hallucinating.

People become gardeners of Context

In this model, people are no longer just users filling in fields.

They become gardeners of Context.

They bring in raw input. They correct interpretations. They validate sources. They set the rules. They create templates. They set the focus. They flag contradictions. They decide what is worth keeping.

This is not a side role.

It is probably one of the core skills of the next few years.

Knowing how to work with an AI will not just mean writing good prompts. It will mean knowing how to build an environment for thinking around it.

Good Context lowers the cost of the next decisions. Bad Context produces bad summaries faster.

The difference will not lie in the tool you use. It will lie in the quality of the memory you build around the tool.

The information system does not disappear. It moves.

Saying that databases and pages will matter less does not mean they will disappear as technology.

There will always be databases. There will always be applications. There will always be permissions, workflows, states, business objects.

But for the user, the centre of gravity is going to shift.

Today we still mostly ask people to understand the structure of the system: where to look, in which tool, with which filter, using which vocabulary, on which page, in which ticket.

Tomorrow we will expect more of the system: that it understands the context of the question.

Not just the words. The context.

What I am trying to do. What we already know. What has been decided. What is reliable. What is uncertain. What is connected. What sits outside the focus but must not be lost.

That is where the information system moves: from the interface to the memory.

The real change is cultural

The hardest part probably will not be technical.

We know how to connect tools. We know how to index documents. We know how to generate notes. We know how to build graphs. We know how to build assistants.

The hardest part will be cultural.

Accepting that the deliverable is not always the capital. Accepting that a document is sometimes an output, not the source. Accepting that raw information has to be turned into reusable ideas. Accepting that an organisation's memory cannot stay in a few people's heads. Accepting that Context has to be built, maintained and governed.

It is a change of stance.

For a long time, we asked information systems to keep a record of what the company did.

Tomorrow we will also ask them to keep what the company understands.

Conclusion: the future of the information system is active memory

The information system can no longer be thought of only as a store.

It has to become active memory.

A memory that keeps the sources, and the links between them. A memory that keeps the documents, and the decisions. A memory that files information away, but above all makes it reusable. A memory that does not just answer a question, but improves the quality of the next ones.

That may be the real shift AI brings.

Not just producing faster. Not just automating tasks. Not just talking to your tools.

But turning scattered information into living Context.

The companies that come out ahead will not just be the ones that plugged an AI into their data. They will be the ones that worked out that raw data is not yet memory, that memory is not yet understanding, and that understanding is where the real capital is.

Tomorrow's information system will not just be the place where a company stores what it knows.

It will be the infrastructure that helps it remember, connect and decide.