Organising intelligence.
Why we are building Comoto Enterprise OS, and the eight foundations it rests on.
This text describes the vision that guides Comoto Enterprise OS. Part of what it describes is available today, part is being built; we make this clear when scoping each project.
A new capacity for work
Imagine that tomorrow morning, several new colleagues join your team.
One can go through thousands of sources to prepare a piece of research. Another can build the tools your team needs. A third analyses your figures. A fourth checks the others’ work and flags what does not hold.
Each has a speciality. They work in parallel. And all of them can contribute to the same goal.
They do not just advise you. They can use tools, work on your documents, look for the information they are missing, hand a task to one another and carry a piece of work through several steps, without you having to dictate every action.
Except that this new team is not entirely human. These new colleagues are AI agents.
Until now, most of us have discovered artificial intelligence through a chat window. We make a request. The AI answers. Then we go back to work. Agents change this relationship: intelligence starts to take a place within the work process itself.
One person can work with several agents. A team can bring together its members’ agents and the company’s. We no longer only speak of a personal assistant. We start to speak of a new capacity for work for the whole organisation.
Then a question appears: how do you organise all this? Who does what? Who sees what? Who can act? Who checks? Who decides? And what does the organisation keep from all this work?
This is exactly why we are building Comoto Enterprise OS: a shared environment in which people, their AI agents and their tools can work around the same projects, with clearly defined roles, access, missions and levels of autonomy.
Why now?
Because several limits that kept agents from becoming truly useful are receding at the same time.
Models are becoming more capable. Above all, agents are becoming able to work much longer: on some technical, verifiable tasks, what was measured in seconds, then minutes, is starting to be measured in hours of human work. These results do not guarantee the same autonomy in every situation a company faces.
The amount of information these intelligences can work with has changed by an order of magnitude. They can use computers, operate software, write and run code, consult sources and produce real professional deliverables.
At the same time, the cost of some capabilities is falling. Available computing power is growing, hardware is improving, algorithms are becoming more efficient. Capabilities that once required cutting-edge infrastructure are becoming available to more organisations.
But one essential limit remains: reliability. An agent can be extraordinarily capable and still be wrong. We have not reached the moment when artificial intelligence can do everything alone. We are reaching the moment when it can do enough real work that we must learn to organise it.
Technology has started to cross the threshold before organisations. We are building Comoto Enterprise OS to help with that passage.
How do you organise this intelligence?
To build Comoto Enterprise OS, we started from eight simple questions, which appear as soon as humans and agents really start working together.
Identities and responsibilities
Who is who, and who is responsible for what?
In Comoto Enterprise OS, every person and every agent has its own identity. We deliberately separate identity, role, permissions and the intelligence engine used.
A Finance agent can be in charge of checking the figures without having access to every financial document in the company. And producing something, checking it and approving its use remain three different acts.
One agent can prepare. Another can check. A human decides.
Human responsibility does not disappear because more work is delegated. The aim is to create enough structure to delegate more, while keeping a record of contributions, rights and responsibilities.
Rooms and calls
How do they really work together?
Today, we still carry a great deal of context ourselves: a document added to one assistant, a meeting summarised elsewhere, a decision passed on, then everything starts again with the next one.
In Comoto Enterprise OS, humans and agents can work directly in the same spaces. Conversations, meetings, documents, questions and decisions stay linked to the work they belong to.
But sharing a space does not mean sharing all the information.
Every person and every agent receives the context its role and access give it the right to. Context can then be found and shared within the project, according to each one’s access, instead of being rebuilt at every exchange.
Heritage and forge
What does the organisation keep?
When we work with artificial intelligence, we do not only produce documents. We also produce decisions, corrections, analyses, instructions and reusable methods. Step by step, we develop the organisation’s own way of solving its problems.
What the organisation builds with its humans and its agents should be able to become its own heritage.
Deliverables stay linked to their sources, their history and their versions. Methods can evolve too: a correction can reveal a better procedure, a successful mission can give rise to a reusable method, and that improvement can be reviewed and approved before it becomes the new reference.
Work produces experience. Experience improves methods. Methods improve the next missions. And new missions in turn produce new knowledge.
The organisation no longer only does work. It gradually learns to work better. This working memory, these methods and this knowledge form a heritage that stays under its control.
Projects and mini-apps
How do you stay stable without becoming slow?
Comoto Enterprise OS organises work around projects. A project brings together the people, agents, context and capabilities needed for an outcome.
A platform evolves at the pace of the features its vendor builds. The team’s needs, and what AI makes possible, can change faster.
So Comoto Enterprise OS lets each project host its own mini-apps and capabilities. A new capability can have an interface for humans and expose functions that agents can use, within the permissions granted.
The result is a tool-building capability built into the work: a stable base for identities, rights, projects and heritage, on which new mini-apps can be built as needs change.
Agents and AI engines
What happens when the best model changes?
In Comoto Enterprise OS, we do not want to define an agent by the provider that powers it. An agent has an identity, a role and a place in the organisation. Only then comes the intelligence it uses.
Its runtime environment, called a harness, organises the work and the use of tools. The model provides the intelligence. The account used determines access to that model.
It is not a Claude, Codex or Gemini agent. It is your agent.
Today, one model may suit a mission better. Tomorrow, another may become better, cheaper, faster or able to run locally. The organisation should not have to be rebuilt each time the market changes.
Models will change. Harnesses will change. Capabilities will change. The agent stays. The intelligence behind it can change.
Missions and methods
How do you really hand it work?
With a chatbot, we make a request. With an agent, the unit of work can become different: we hand it a mission.
A mission starts with an expected result: for example, a proposal ready on Friday, with the figures checked, the sources needed and precise criteria.
Then we define the resources it can reach, the limits, the methods available and the decisions that need a human.
The agent can then look for what it is missing, use a tool, apply a method, ask another agent for help, have its result checked, correct and carry on. When it meets a decision beyond its mandate, it comes back to the human.
Continuity also means keeping the state of the mission when it stops: what has been done, what remains and what is blocking.
We no longer simply ask AI to help us. We organise the work we can really hand to it.
Connectors and autonomy
How far can it act?
To take on real work, an agent sometimes has to step outside Comoto Enterprise OS: send an email, change a calendar, use a CRM, file a document or trigger an operation.
Connectors let it use the company’s real services. But capability and authorisation are two different things.
The agent can simply prepare, ask for approval before acting or, for some actions, act within a pre-authorised frame: duration, recipients, budget, frequency or resources.
Autonomy is not a simple switch. It is gradual and depends on context.
And when an action is carried out, the system tells apart what had been authorised from what really happened. An authorisation is not proof of execution.
The aim is not to build agents that can do everything alone. It is to decide precisely what we are ready to hand them, then let them really do it.
Infrastructure and continuity
Where does this organisation live?
So that work can carry on when you close your computer, Comoto Enterprise OS relies on a shared base controlled by the company.
People reach it from their devices, while shared services and authorised missions can keep running.
You can almost picture this infrastructure as an extra room in the company. A room that stays open when you go home.
You hand over a mission. You close your computer. And the next day, you do not simply find a new answer: you find a file that has moved forward.
Storage, running the agents and the place where models run their inference remain separate choices. The company decides what it keeps, the services it uses and what it agrees to pass on to them.
Continuity is not simply keeping a machine switched on. It is making sure the organisation can go on existing and stay under the control of those it belongs to.
What if the company itself changed?
For a long time, increasing an organisation’s capacity mainly meant adding people, capital and software. Today a new variable appears: the amount of intelligence and working capacity each person can organise around them.
Automating an existing task is not the same as rethinking how work should be done when humans and agents can work together from the start.
Software itself is changing too: agents are becoming users of software. They consult data, call functions, use capabilities and act directly in systems.
Two companies may have exactly the same model and still draw radically different value from it. One will discover its capabilities on the day it is released. The other will already have learned to structure its data, methods, permissions, evaluations and teams to work with this kind of intelligence.
The real question becomes: how fast can we become an organisation that truly knows how to work with it?
What we choose to do with it
The most interesting question may not be how far this intelligence can go. It is what we will choose to do with it.
A company can use it simply to produce more. It can also use it to become more ambitious: letting a small team attempt things that used to require a much larger organisation, letting an expert spend more time on the problems where their expertise really counts.
The goal is not to build an organisation in which humans become useless as agents get better. It is to build an organisation in which humans can become more ambitious as their tools become more capable.
That this intelligence can change without us losing our work. Learn without us giving up our memory. Act without us giving up our responsibility. And grow more powerful without us losing control of what we build with it.
Not a company without humans. Not a company filled with AI. An organisation able to bring together the best of both, and to let each person go further than they could alone.
Comoto Enterprise OS. Designed in Neuchâtel by Meotis.
