AI can generate answers. The organization still has to determine which answer is meaningful, acceptable and feasible in its own circumstances.
Visual summary of the article
Explore the article’s key ideas through diagrams and visual explanations.
AI has become a versatile organizational tool. It can summarize documents, analyze data, develop scenarios, identify anomalies, generate ideas, compare alternatives and propose action plans.
This raises an understandable question: as AI becomes more capable, will organizations still need complex collective decision processes? Could they simply formulate the right question and ask AI for the best answer?
The difficulty is that making an organizational decision is fundamentally different from processing information.
A decision begins before the search for an answer
The hardest part of an organizational problem is often understanding what the question really is.
Consider declining sales. Possible explanations include product quality, pricing, marketing, the sales team’s capabilities, changing customer needs, competition or internal processes.
AI can generate all of these hypotheses. But knowledge about which factors matter most in a particular organization is often spread across different people:
- A sales representative sees how customers respond.
- An engineer understands the product’s limitations.
- A project manager sees workflow bottlenecks.
- A finance specialist understands the cost structure.
- A senior leader knows the strategic constraints.
- A customer sees things that nobody inside the organization sees.
This knowledge does not necessarily exist in one database. It is distributed among people. The first decision challenge may therefore be fragmented knowledge, even when information is plentiful.
AI may have extensive information without access to the organization’s lived experience
Large language models draw on broad bodies of general knowledge. Yet the most important knowledge for a particular decision may never have been published or recorded.
Why did a specific customer really leave? Why does a procedure fail even though it looks logical on paper? Why do employees avoid a new system? Which informal workarounds keep the operation running? What problem are people reluctant to raise with their manager?
Much of this local, experiential and tacit knowledge develops through working in a particular situation. AI can help process it once it has been expressed and made available. The organization first needs a way to elicit it and bring it together.
An organizational decision involves values as well as optimization
If the objective and evaluation criteria were fully specified, many decisions could be treated as optimization problems. In organizations, however, people often disagree about the objective itself.
One department may prioritize lower costs, another higher quality. Senior management may focus on growth, while employees emphasize working conditions. A city administration may prioritize efficiency, while residents place greater weight on access to services. Investors may favor near-term returns, while the organization also needs long-term resilience.
AI can help explore the consequences of these choices. Data and analysis alone cannot establish which value the organization ought to place first. That is also a social and normative judgment.
An AI recommendation is not yet an organizational decision
Suppose AI analyzes all the available data and recommends what it considers the best restructuring option. Has the organization now made its decision? No.
Several questions remain. Who knows something that was missing from the data? Who will experience the consequences? Which risks do different people perceive? Is the proposal workable within the organization’s culture? Do the people responsible for implementation understand the reasoning behind it?
AI can contribute analysis and proposals to a decision process. Its recommendation does not, by itself, resolve these questions or assign responsibility for the action taken.
A major risk is automating a poorly designed decision process
Organizations often introduce AI into an existing hierarchical system. Previously, a leader might consult a few direct reports and make a decision. With AI, the leader may consult those same people, ask AI as well, and then decide.
The capacity to process information has increased. The underlying decision architecture may remain unchanged.
The same weaknesses can persist:
- Information is filtered as it moves up the hierarchy.
- The views of higher-status people carry disproportionate weight.
- Employees adjust their responses to the leader’s position.
- Uncomfortable knowledge remains unspoken.
- Job titles become substitutes for evidence of expertise.
- The people shaping decisions are separated from those who experience their consequences.
Making a poorly designed decision process faster does not necessarily make it better.
Where collective intelligence contributes
Collective intelligence addresses a distinct challenge: how to bring many people’s distributed knowledge into a shared decision process.
A survey, meeting, brainstorming session or simple vote is not sufficient by itself. In the approach described here, collective intelligence emerges when a sufficiently large group interacts in the same structured information environment. We call this process collective decidement.
Participants first evaluate earlier proposals. If they have something new to add, they submit one clear, original, non-repetitive idea. Submission and evaluation are interwoven, and participants can contribute asynchronously within the agreed project period. Their evaluations are integrated algorithmically into a collective result.
An important principle is that a proposal is evaluated independently of its author’s status. Authors and interim aggregate results remain hidden during evaluation. This supports independent judgment and reduces pressure to agree with a senior manager or the most outspoken person.
Anonymity is accountable: the platform preserves the connection between each participant and their actions, so contributions remain traceable and measurable even though participants do not see one another’s authorship.
Why a sufficiently large group matters
Having several people involved does not automatically create collective intelligence. The process needs a useful range of perspectives and enough evaluation activity to compare proposals meaningfully.
For this method, we recommend at least eight active participants in each project. A group of 15–20 or more can provide a broader range of expertise when members contribute relevant, complementary knowledge.
These are practical guidelines for this approach, not universal thresholds. Neither eight participants nor a larger group guarantees a good outcome. The task, the relevance of the participants’ knowledge and the quality of interaction all matter.
A larger group can improve the chances of finding perspectives that a small team would miss. If it is impossible to assemble a suitable participant group for a particular task, another decision process may be more appropriate.
A cluster is the wider participant pool. Each project addresses one clearly defined task, and different members can participate in different projects. A complex problem may require several linked projects, with the findings of one helping to frame the next.
What if the organization lacks the necessary expertise?
Collective intelligence cannot create missing expertise simply by increasing participation. If nobody understands the relevant technology, market or process, adding more people with the same knowledge gap will not resolve it.
The organization can use external expertise sourcing: bringing external experts, customers, partners, researchers or practitioners from other organizations into the same collective decidement space.
This allows the organization to supplement its knowledge while retaining responsibility for its own decision. External expertise becomes part of the shared process.
AI and collective intelligence address different limitations
Artificial intelligence expands analytical capacity. It can process information, generate hypotheses, explore scenarios, critique alternatives and suggest additional possibilities.
Collective intelligence improves how human knowledge enters the decision process. It connects distributed experience, reduces the influence of status, reveals convergence and polarization, and can involve people who will experience the consequences.
The two can complement one another. Their combined value depends on how the process is designed and how the resulting information is used.
From artificial intelligence to symbiotic intelligence
An organization can use AI to strengthen collective decidement throughout the process.
People help identify the problem. AI assists with analysis of its context. Participants evaluate and develop proposals, while AI can suggest additional hypotheses or identify missing information. People assess practical and social value. AI can assist with analysis of the evaluation patterns. The collective process identifies promising directions, and AI helps develop scenarios and action plans for consideration.
These activities can feed back into one another as understanding develops. Idea generation and evaluation remain interwoven within the collective process.
This creates a hybrid decision system that can be described as symbiotic intelligence: human collective judgment and machine-assisted analysis working together.
The organization remains responsible for choosing what to implement, organizing that implementation and assessing the resulting change.
The organizational AI revolution may also be a revolution in decision architecture
Many organizations ask: “How can we use AI?” A deeper strategic question is: “How should our decision architecture change as analytical capabilities become more widely available?”
As generating answers becomes less costly, competitive advantage may increasingly depend on the ability to identify the right problem, bring together relevant expertise, evaluate alternatives and turn distributed knowledge into a sound organizational decision.
AI makes that challenge more visible. It also creates new opportunities to address it through collective intelligence.
Saulius Norvaišas
Developer of the collective intelligence methodology and platform