simmistoward organizational intelligence.
Organizations become intelligent with the right structure. It does not become intelligent merely because it has intelligent members. Without the right setup, a room full of brilliant people can still forget, duplicate work, and pull the organization in opposite directions. Implementing a fleet of agents will only replicate the same destructive behaviour.
Intelligence at organizational scale appears when the whole can combine what its members perceive, preserve what it learns, coordinate parallel action, and change how it operates. Biology repeatedly reached new levels of capability this way. Mitochondrial endosymbiosis integrated a formerly independent bacterium into the machinery of the complex cell. Multicellularity added communication, specialization, shared reproduction, and control over conflicts between cells. Nervous systems combined enormous numbers of locally acting neurons through distributed, hierarchical, and recurrent networks.
These systems combine local autonomy with nested control. Their parts act locally and in parallel, while the whole maintains boundaries, memory, communication, and control. We are convinced that organizational intelligence requires an analogous architecture.
That analogous architecture is still largely missing. Simmis is building the structure through which individual intelligence synergize, using shared state, controlled change, and a programmable operating environment. Workers may come and go. The organization must remember, coordinate, and learn.
ONE FOUNDATION. A WIDER FIELD OF ACTION.
the shoulders of giants.
Lynn Margulis, publishing as Lynn Sagan, argued for the endosymbiotic origin of cellular organelles in On the Origin of Mitosing Cells. Nick Lane and William Martin’s The Energetics of Genome Complexity examines how mitochondrial integration may have enabled greater cellular complexity. Katrin Hammerschmidt, Caroline Rose, Benjamin Kerr, and Paul Rainey’s Life Cycles, Fitness Decoupling and the Evolution of Multicellularity studies how cooperating cells become a new level of individuality.
Laurence Hunt and Benjamin Hayden’s A Distributed, Hierarchical and Recurrent Framework for Reward-Based Choice and Edward Bullmore and Olaf Sporns’s Complex Brain Networks describe cognition emerging through distributed processing, hierarchy, recurrence, modules, and hubs.
Alan Turing’s Intelligent Machinery and Computing Machinery and Intelligence treat intelligence as something a machine may acquire through education and experience. John von Neumann’s The Computer and the Brain and Theory of Self-Reproducing Automata examine how computation, memory, and complex organization can arise from arrangements of simpler parts.
Warren McCulloch’s A Heterarchy of Values Determined by the Topology of Nervous Nets, Herbert Simon’s The Architecture of Complexity, and William Ross Ashby’s An Introduction to Cybernetics examine contextual authority, semi-autonomous systems, feedback, and the conditions under which complexity remains governable.
George Huber’s Organizational Learning, James March’s Exploration and Exploitation in Organizational Learning, and James Walsh and Gerardo Ungson’s Organizational Memory distinguish individual insight from learning that changes organizational memory and future action.
Anita Williams Woolley, Christopher Chabris, Alex Pentland, Nada Hashmi, and Thomas Malone’s Evidence for a Collective Intelligence Factor in the Performance of Human Groups shows that group performance cannot be reduced to the average or highest individual intelligence. Thomas Malone and Michael Bernstein’s Handbook of Collective Intelligence extends the question to systems made of people, computational agents, and organizations.
These works do not describe Simmis. They establish the biological, computational, and organizational questions Simmis is trying to answer with present-day software.
direct outcomes, not every step.
AI agents make work easier to start and harder to coordinate. Context splinters across chats, several attempts collide, and people are left carrying information between systems.
Simmis’s immediate horizon is a shared workspace where people and agents begin from the same organizational context. Work should proceed locally and in parallel inside a common operating boundary, without immediately changing the accepted state others rely on.
Human attention should go where authority, uncertainty, or consequences still require it. As agent activity grows, the interface should summarize what matters instead of asking people to read every transcript. People should not become the integration layer.
Organizations should own their memory, provenance, policies, and decisions even when models and peripheral services change. They should own the institution and rent the intelligence they need.
make better ways of working part of the organization.
An organization learns whenever someone finds a better process, builds a useful tool, or discovers how to make a recurring decision. Today, those gains often leave with a person, stay buried in a document, or become another application to maintain.
A programmable organization can keep the improvement. People and agents should be able to change the tools, procedures, and interfaces through which work happens while sharing one organizational basis.
Organizational learning begins when experience changes what the organization can do next. Simmis is building a programming model through which accepted work can improve shared memory, procedures, tools, and interfaces instead of disappearing with the worker who produced it.
Simmis should arrive as an assembled product, not an infrastructure assignment. It should also remain malleable beneath the interface. The organization can then gain capabilities without adding another disconnected layer every time it learns.
A new organization should be able to establish this operating core from the beginning. An existing one should be able to start with one meaningful operation alongside current systems and expand as related work begins to share context.
compare possible futures before making one real.
Consequential decisions depend on assumptions about what happens next. A product launch, market entry, hiring plan, or policy change draws on evidence that is usually scattered across people, documents, and spreadsheets.
An organization that remembers its decisions and outcomes could maintain those assumptions as a model. It could branch them, explore consequences under explicit conditions, compare alternatives against its purposes, act, and learn when reality disagrees.
The purpose is better judgment, not synthetic certainty. Useful organizational simulation must expose its assumptions, remain grounded in evidence, and learn from the difference between projection and outcome.
Rigorous scenario planning is costly and often requires specialist teams. AI may make it useful to smaller organizations, provided the system makes uncertainty visible instead of manufacturing confidence.
Human oversight is not necessarily the destination. As AI becomes more capable, people should not remain trapped inside every operational loop merely to certify that a human was present. Machines may eventually run much of an organization.
Operational control may increasingly pass to machines. The organization must still remain answerable to the people whose lives it affects. Its purposes, measures of success, and avenues of recourse cannot be reduced to whatever an optimizer happens to pursue.
Organizational intelligence is what makes that transition possible. An intelligent organization remembers the conditions under which work was attempted, observes what followed, learns which methods succeeded, and turns those lessons into better tools, rules, and ways of operating. Its autonomy grows not because a model claims competence, but because the organization can demonstrate, retain, and correct its own performance.
Delegation should expand as the organization demonstrates that it can carry more responsibility, retain evidence, detect failure, and correct itself.
The product surface and research path are still being formed. The roadmap states what exists now.
Put more AI to work today while building the organizational memory and control required for greater autonomy tomorrow.