The organizational context window

Picture a small innovation consultancy, thirty people gathered every Friday morning. Someone has baked. Everyone speaks. Everyone says what they are working on, and knowledge moves freely because nobody in the room is a competitor. The collective intelligence of the group makes everyone in it smarter. Now scale that to 300 people, or 3,000, and the magic dies.

That is the fundamental problem of innovation in a modern organization: how do you keep the conditions that make breakthrough thinking possible when human relationships cannot scale indefinitely? Most innovation frameworks focus on individual creativity or on organizational structure, and miss the layer in between, the collaborative intelligence that appears when the right people can reach the right knowledge at the right time.

Cognitive wholeness, and its limits.

Way back, before the corporate roll-ups and acquisitions, I worked at a small, special place called Doblin Group. Our Friday mornings were not status updates. They were collective sense-making rituals, orchestrated by a standup comedian, and everyone could hold the whole firm's work in their head at once. Call it cognitive wholeness: the ability to see patterns across projects, spot where knowledge could be transferred, and hold a shared standard of quality.

It only works at small scales, however. Research on teams tends to put the sweet spot for collaboration at five to eight people, and for complex coordination rather lower. At Doblin, with about thirty people, we were effectively a team of teams, still small enough for wholeness and large enough for diverse expertise. Past that point the cognitive load becomes impossible. Nobody can track everything, knowledge silos, and innovation slows.

Which is why I have started thinking about what I call the organizational context window: a system that works like a collective intelligence platform, keeping context across all the work in progress while preserving human values and ways of learning. Think of it as an organizational memory that learns and evolves, powered by AI but grounded in human relationships and judgment.

Four parts, one of them hard.

The first is a values core: principles that act as the system's constitution, not mission statements but decision-making rules for how information is processed and shared, re-examined at intervals, because we all change.

The second is a learning layer that captures insights, failed experiments and emerging patterns across every project. Not documentation but living institutional memory that gets smarter over time, built on trusting the insights of the people around you.

The third is a porous interface: boundaries that know when to bring outside information in and when to keep the focus inside. Systems can learn this; people already do, in the form of the slow hunch that nags when something is not quite right.

The fourth, continuous navigation, is the hard one. AI can help people find connections across time, teams and projects. But the real magic of those Fridays was not information sharing. It was Patty's bundt cake, the banter while breakfast was laid out, the people standing in the kitchen. The serendipitous conversation while someone waited for coffee was where insight happened. So the question is not just how to scale information flow but how to recreate the informal moments, the shared rituals and the physical proximity that make collective intelligence possible. Technology has to enhance those moments, not replace them.

Innovation under constraint: the real test.

Theory is one thing. At my current startup, Atamai, we are building exactly this kind of system in one of the harder environments there is: agriculture. Farmers face the innovation paradox in its purest form. They need to adopt sustainable practices to cope with a changing climate, changing conditions and changing markets, and they do it on thin margins with no time to experiment. A failed crop is not a learning opportunity; it can be bankruptcy. Traditional innovation frameworks assume the user has slack to experiment with. What if they do not, and the stakes are too high for individual risk?

Our answer is a farmer-owned data ecosystem in which collective intelligence reduces individual risk. A farmer who wants to try regenerative soil practices does not have to take a leap of faith. She can see what peers in similar conditions found, what the universities know, what suppliers are offering and what financial support exists, through one interface that keeps context across the whole agricultural value chain.

The insight underneath is that most innovation fails not because the idea does not work but because the systems for transferring knowledge and sharing risk are broken. People keep doing what they know because learning something new costs too much. A context window changes that arithmetic: AI carries the complexity so people do not have to; peers and institutions validate before anyone adopts; the outcomes that matter become visible; and the technology fits existing ways of working rather than replacing human judgment.

Innovation was always collective.

Edison's genius came out of Menlo Park, a collaborative laboratory that looked a great deal like a scaled-up Doblin Friday. The Wright brothers were part of a worldwide network of aviation enthusiasts trading notes across continents. Einstein's breakthroughs were built on the liquid networks of ideas moving between physicists across Europe, and as Steven Johnson shows in Where Good Ideas Come From, breakthroughs usually are: existing ideas colliding, insights accumulating across neighboring fields, the right minds meeting the right problems at the right time.

The difference now is not that innovation needs to become more collective. It is that we finally have technology that can scale the human connections that have always driven it. We are not entering a new age of collective intelligence so much as returning to innovation's natural state, with tools that can preserve the human dynamics at a scale they never had. The companies that thrive will be the ones that build what I would call collaborative competitive advantage, where individual success depends on collective learning, across stakeholders who have traditionally been fragmented.

This is not about knowledge management or AI tools. It is about rethinking how organizations learn, adapt and innovate together, and building systems that keep the magic of human-scale collaboration while working at the scale modern problems demand. The organizational context window is a blueprint for innovation that can outlive founders, individual genius and the limits of human memory and attention. The future belongs to organizations that can think collectively while acting individually. The tools are arriving. The question is who will be brave enough to use them.