Scenarios by Balane. The images are concept images by Bria.
Nobody builds apps (2029, Software)
Today, teams write code for months. The day after tomorrow, software takes shape in conversation, in minutes. The machine takes over the typing. Not the deciding, and running it does not get cheaper.
Today
Code by description: Developers let AI assistants write functions and only review them. Cursor, Copilot and Claude Code are part of everyday work in many teams.
Prototypes the same day: What used to be a sprint now stands as a clickable draft the same afternoon. Whether it holds up in operation shows later.
Experts build for themselves: People without a computer science degree build their own internal tools without waiting for IT. Who maintains them once they leave is usually written down nowhere.
What changes
You describe instead of build. An app emerges from a conversation about the problem. The software asks back, proposes, builds and tests itself.
Being wrong gets cheap. A tool for one week pays off because it costs almost nothing. Whatever does not fit is regenerated. What has to keep running stays expensive: maintenance, security, liability.
Deciding becomes the bottleneck. No machine can automate a decision nobody ever made. Whoever does not know what they want gets the wrong thing very quickly.
The story, step by step
A developer floor, today. Code is running on every screen.
On the wall, the backlog. Every card is a wish, and new ones keep coming.
AI already writes code alongside people, who review it. Still, every wish waits for a team.
The day after tomorrow. The desks are gone. A person says what they need.
The software listens, asks back and makes a suggestion.
Then it builds itself. Piece by piece, in seconds.
Before anyone uses it, it tests itself. Every test green, checked against what it understood.
Around every person, whatever they need right now takes shape.
Building now means describing. The machine takes over the typing. Not the deciding.
A tool for one week pays off. Whatever falls apart is regenerated. Whatever has to keep running still costs.
Deciding becomes the bottleneck. If you don't know what you want, you get the wrong thing very fast.
Who still understands your software when the machine that built it fails?
Who still understands your software when the machine that built it fails?
The objection: Maybe running software becomes as cheap as building it. Then the only bottleneck left is knowing what you want. That was always the most expensive part.
Your colleague is an agent (2030, AI)
Today you call and wait on hold. The day after tomorrow, an agent settles the request before anyone picks up the phone. Friendly, fast and patient. For the company that pays it.
Today
Agents in customer service: First companies let AI agents answer requests, look things up in their systems and act on their own.
AI with tools: AI models operate browsers, fill in forms and book appointments.
Gains mostly for newcomers: Studies of AI in customer service find the biggest gains among beginners. For experienced staff, little changes.
What changes
Hold music dies. An agent takes the request, checks contract, stock and calendar, and resolves it in the same moment.
Goodwill becomes a setting. The agent is polite and has time. How far it accommodates you was decided beforehand by someone. Hardly anyone asks about that.
Trust becomes work. Someone has to define what an agent may do, check what it did, and answer for it. That is the new job on the team.
The story, step by step
An open-plan office, today. A person at every desk, a screen on every desk.
Up front, the hold queue. Every light is a call that is waiting.
Work moves as fast as people type.
Evening. One by one, people go home.
Then the lights come back on.
Agents pick up requests the moment they arrive.
They check contract, stock and calendar and settle it in the same moment. Someone set how generous they are.
The hold queue is empty. No human is waiting anymore.
One person stays, not to type but to decide.
Whatever an agent may not do alone lands on this desk. Where that line sits is the real decision.
Trust becomes work: deciding what an agent may do, checking what it did, and answering for it.
What may your agent decide without asking you?
What may your agent decide without asking you?
The objection: An agent working for the company may still beat a hold queue that works for nobody. Today the bar is very low.
The company without an org chart (2031, Company)
Today, a sheet of paper decides who reports to whom. The day after tomorrow, teams form around tasks instead of boxes. Power does not disappear. It just is no longer on paper.
Today
Teams by task: Many companies work in project teams across departments. The org chart barely reflects this anymore.
Knowledge in tools: Who knows what increasingly lives in chats and documents rather than in a manager's head.
Small teams, big impact: With AI, small teams do work that used to require entire departments.
What changes
The boxes dissolve. Teams form around a task and disband afterwards. Structure follows the work.
Leadership becomes a role. Whoever is needed right now leads. Who decides who is needed sits in no box anymore.
Risk moves outside. Fewer permanent staff, more partners and agents. The company becomes agile. The uncertainty is now carried by those who can be let go at any time.
The story, step by step
A company, today. Every team has its box, every box its place.
A question goes through the chain of command. All the way up, and back down.
The answer sits next door. But no line leads there.
Then the lines snap.
The boxes dissolve. Everyone stands on one level.
The question becomes a task. Whoever can help comes over.
Solved. The team breaks up, each to the next task.
Everywhere, teams form and dissolve again. Structure follows the work.
Whoever is needed right now leads. Who decides that is written down nowhere anymore.
Partners join from the fog, agents work alongside. They stay as long as the task lasts.
The edge of the company blurs. With it blurs who carries the risk and who can still say no.
Whom do you contradict when there are no boxes anymore, and what does it cost you?
Whom do you contradict when there are no boxes anymore, and what does it cost you?
The objection: The box also protected people. Knowing whom you report to meant knowing whom you do not. Without structure a company may get faster. Or just harder to see through.
The last screen (2033, Interface)
Today we tap colorful icons. The day after tomorrow, software listens and acts, and the interface dissolves. With it goes the choice you could see.
Today
Answers instead of links: Search engines increasingly give the answer directly instead of a list of pages. Which pages there would have been, you no longer see.
Devices without screens: Glasses, earbuds and pins try to take the phone out of our hands.
The screen as shop window: In summer 2026, BMW showed a Spider-Man animation on the displays of cars it had already sold. Whoever controls the screen decides what appears on it.
What changes
The app disappears. You say what you want. Which software handles it, someone else decides.
The default turns invisible. If a view is needed after all, it is generated in that moment. What is missing from it, nobody notices. Every default is a decision someone made for others.
Calm becomes a luxury. Whoever does not constantly look at a device gains time. Like every luxury, it gets sold.
The story, step by step
A square, today. Lots of people, every one of them looking down.
Above every head, something wants attention. Every alert asks for a glance.
An app for every task. You search, tap, switch, and search again.
The day after tomorrow. One screen after another goes dark.
You say what you want. The software listens and gets it done.
The app disappears. Nobody needs to know which software handles it. Someone still decided.
When you do need a view, it is made in that very moment.
Once you have seen it, it is thrown away.
The square falls quiet. People look at each other and out into the distance.
Calm becomes a luxury. And like every luxury, it gets sold.
One screen remains. Whoever walks up to it wants to see what is otherwise chosen for them.
Who chooses for you when you no longer see a choice?
Who chooses for you when you no longer see a choice?
The objection: Most interfaces are well meant, and less screen is a real gain for many people. The open question is whether you can still change the default.
Prices that negotiate (2035, Markets)
Today, the price is on the tag. The day after tomorrow, two agents negotiate in milliseconds, and every price holds for a single moment. Whoever has the better agent pays less.
Today
Demand-based pricing: Flights, hotels and ride services adjust their prices continuously to demand.
Digital price tags: In retail, electronic labels are replacing paper. Prices change at the push of a button.
Agents as buyers: First services let AI assistants search, compare and buy on behalf of their users.
What changes
Price becomes a conversation. The buyer's agent and the seller's agent agree before any human sees the offer.
Every price is used once. Quantity, timing and loyalty flow into every single price. It can no longer be compared with anything.
Lock-in becomes the answer. An agent switches to the cheapest offer without hesitation. Sellers answer with lock-in: subscriptions, accounts and perks you lose when you leave.
The story, step by step
A shop, today. Long shelves, and a small glowing label on every product.
One push of a button at head office, and every price changes. The same for everyone.
When the shop fills up, prices rise. The label follows demand.
Soon the customer no longer goes in person. She stays home and sends her agent.
The shop answers with an agent of its own. It steps out of the shelf.
The two negotiate. Bid and counterbid swing back and forth until they meet.
Only once both agree does the offer reach the customer. All she has to do is say yes.
Quantity, timing, loyalty: everything flows into this one price.
Every price is valid only once. After the sale it dissolves, and the next one is already different.
Soon the whole shop is negotiating. Pairs everywhere, each with a price of its own.
The price list dies. No label tells you what anything costs. You can only compare what the agents show you.
Someone has to decide what machines may know and exploit. Otherwise whoever has the better agent pays less.
Who owns the agent that negotiates for you?
Who owns the agent that negotiates for you?
The objection: Maybe it simply gets cheaper for most people. If every agent knows every price, sellers have less room, not more. What matters is who builds the agents.
The state as an interface (2038, State)
Today, files pile up under a stamp. The day after tomorrow, administration is an interface that checks the application while you file it. Then what counts is who writes the rule, and what happens when it stops running.
Today
Digital offices: Estonia handles almost all dealings with authorities online. The country shows it can be done.
Promised is not delivered: Germany's Online Access Act was meant to bring hundreds of public services online by the end of 2022. Three years after the deadline, only part of them were there. There were hardly any consequences.
When the network falls, the service falls: After the attack on two Berlin senate departments in August 2026, housing benefit payments for tens of thousands of households were at risk.
What changes
The application files itself. The state knows what you are entitled to and offers it before you ask.
Law becomes code. Laws are written so that machines can check them. Whoever translates them writes along, and every default becomes procedural law.
A person for the rest. A person remains for the cases that need empathy. And for the day the system goes down.
The story, step by step
A government corridor, today. Door after door, and someone waiting outside each one.
The display changes. Nobody can tell you when your turn will come.
The same form for every door, filled in once more. Files travel from stamp to stamp.
Then the paper folds into itself. What the state needs to know sits in one place. Once.
A woman walks in. Before she reaches a door, the offer comes to her.
The application files itself. She says yes and walks past every door.
The law on the wall becomes a rule a machine can check. Whoever translates it writes along.
The stamp becomes an interface. Decisions flow without anyone touching them.
The chairs are no longer needed. There are no waiting numbers anymore.
The walls fall. The corridor becomes an open space where everything runs quietly. An error now runs quietly too, the same in every decision.
One person remains. For the cases that need empathy, and for the day the system goes down.
Whom do you turn to when the interface says no?
Whom do you turn to when the interface says no?
The objection: A state that delivers wins twice: once in results, once in trust. The risk lies less in the machine than in nobody practicing how to work without it.
2045: Not yet released. This scenario is still being thought through.
The future, painted.
N° 01 · Software · 2029
Nobody builds apps
01 · Today · Signs that it has begun
Code by description
Developers let AI assistants write functions and only review them. Cursor, Copilot and Claude Code are part of everyday work in many teams.
Prototypes the same day
What used to be a sprint now stands as a clickable draft the same afternoon. Whether it holds up in operation shows later.
Experts build for themselves
People without a computer science degree build their own internal tools without waiting for IT. Who maintains them once they leave is usually written down nowhere.
Draft. Sources to follow.
02 · Tomorrow · 2029 · Concept image
Today, teams write code for months. The day after tomorrow, software takes shape in conversation, in minutes. The machine takes over the typing. Not the deciding, and running it does not get cheaper.
03 · What changes
01
You describe instead of build.
An app emerges from a conversation about the problem. The software asks back, proposes, builds and tests itself.
02
Being wrong gets cheap.
A tool for one week pays off because it costs almost nothing. Whatever does not fit is regenerated. What has to keep running stays expensive: maintenance, security, liability.
03
Deciding becomes the bottleneck.
No machine can automate a decision nobody ever made. Whoever does not know what they want gets the wrong thing very quickly.
04 · The question
Who still understands your software when the machine that built it fails?
The objection Maybe running software becomes as cheap as building it. Then the only bottleneck left is knowing what you want. That was always the most expensive part.
Keep scrolling: N° 02 · Your colleague is an agent