GenAI adoption does not spread evenly across a workplace. Some employees try it early, some wait for social proof, some avoid it, and some use it quietly when guidance is unclear. This simulation uses Threshold Theory to explain cascades and Segregation Theory to explain workplace clustering.
This project keeps the academic theory at the center. AdoptAI (working concept) is the workplace context used to make Segregation and Threshold Theory visible, concrete, and relatable.
In Dynamic Models of Segregation, Thomas C. Schelling showed how local choices can produce large patterns of separation. In this workplace version, employees do not need strong or negative intentions to form groups; similar habits, comfort, and learning networks can gradually create clusters of AI users, cautious users, and non-users.
Stay if enough nearby coworkers are similar
In Threshold Models of Collective Behavior, Mark Granovetter explained how people may join a behavior only after enough others have already joined. In GenAI adoption, one employee may adopt after a few trusted examples, while another may wait until adoption is already common.
Adopt if nearby AI users ≥ personal threshold
One starter can fail. Two starters can trigger a small chain reaction. Enough starters can cross the threshold and produce a full cascade.
Broken cascade → critical point → full cascade
A workplace can have high AI use but weak visibility. Without clear policy, training, and review, adoption may become hidden rather than guided. The purpose is not to blame employees, but to understand the conditions that shape responsible adoption.
More adoption ≠ safer adoption
Each employee reads the local environment: coworkers, manager signals, policy clarity, training, and visible success stories. If the signal crosses their threshold, they adopt. If the surrounding behavior feels too different, clustering can appear. The result is not only adoption, but adoption patterns.
Everyone may have access to the same technology, but not everyone adopts at the same time. Some employees are curious. Some wait for evidence. Some are cautious because of quality, ethics, privacy, or policy concerns.
This is the first academic point: individual thresholds are uneven.
Early adopters create the first visible signal. But a signal is not yet a cascade. If surrounding employees need stronger proof or clearer guidance, adoption remains isolated.
green cells are safe adopters. yellow cells are waiting for stronger proof.
An employee adopts when the local signal reaches their personal threshold. The signal may come from trusted teammates, manager support, training, or repeated successful examples.
If nearby AI adopters ≥ my threshold, I adopt.
A broken cascade happens when adoption starts but cannot continue. The first adopters are not enough to activate the next layer of employees.
Workplace meaning: people may try GenAI once, but without training, trust, manager support, or clear policy, the behavior may not become part of daily work.
At the critical point, the system becomes sensitive. One manager endorsement, one credible use case, one training session, or one clear policy can determine whether adoption spreads or stalls.
This is why threshold models are useful: they explain sudden change without assuming everyone changes at once.
Once enough people adopt, GenAI becomes self-reinforcing. Employees learn from each other, share workflows, refine prompts, and begin treating AI as part of ordinary work.
This is the cascade: local adoption produces collective behavior.
Schelling’s insight now enters the story. Even mild local preferences can produce visible separation. Confident AI users may cluster together. Cautious employees may remain with cautious peers. Non-users may become isolated from the learning network.
Segregation is not only about place. It can also describe behavioral clustering.
The result is not simply “users versus non-users.” A workplace can develop multiple zones: safe adopters, cautious employees, non-users, and hidden or informal AI users.
safe users cautious users non-users shadow users
The concern is not only low adoption. A workplace may appear cautious because employees do not openly report their AI use. If policy is unclear, AI adoption may spread informally and remain difficult to see.
That is shadow AI: adoption without enough visibility, disclosure, approved tools, or human review.
Governance does not simply block adoption. It changes the path. Training reduces uncertainty. Clear policy improves visibility. Approved tools and human review move the system toward safer adoption.
The goal is not just more adoption. The goal is visible, responsible, governed, and inclusive adoption.
AdoptAI is a workplace application, but the intellectual center remains Segregation and Threshold Theory.
Because thresholds create tipping behavior. Once enough people join, they activate others whose thresholds are now satisfied. What looked like isolated use becomes a cascade.
Because segregation dynamics create clusters. People learn, share, and normalize behavior locally. If teams are disconnected, one group may become AI-powered while another remains cautious or excluded.
Governance changes model parameters: it lowers uncertainty, increases visibility, reduces hidden use, and helps convert risky adoption into safer adoption. In the model, governance is not decoration — it changes the trajectory.
A workplace does not adopt GenAI one employee at a time in a smooth line. It crosses thresholds. Below the threshold, adoption remains isolated. Near the threshold, small interventions matter most. Above the threshold, adoption becomes self-sustaining.
But when adoption combines with segregation, the workplace can split into safe users, cautious users, non-users, and shadow users. The deeper goal is not simply adoption. It is safe, visible, responsible, governed, and inclusive adoption.