AI Adoption Spreads Through Social Networks, Not Tech Budgets

I kept noticing this pattern in case after case. A company announces an enterprise AI strategy. The same company has one department where adoption takes off, led by a senior manager who talks to the right people, and another department, with the same budget and the same tools, where adoption never happens. The second department is across the building, but socially it might as well be in a different city. The gap has nothing to do with how much money each department spent on AI tools. It has everything to do with who talks to whom.

I do not think this is a failure of planning. It is exactly what diffusion theory would predict, and the prediction was available fifty years before anyone had heard of ChatGPT.

Rogers (1962) built the diffusion of innovations framework around a claim that the technology industry still has not absorbed. Innovations spread through social systems over time. The mechanism is not product quality, not budget allocation, not even executive mandate. It is social communication through channels among members of a social system. Rogers named five perceived attributes that shape adoption, relative advantage, compatibility, complexity, trialability, and observability, but the channel through which those perceptions travel is the social network. A technology can score high on all five attributes and still die in an organization where the right nodes are not connected.

Granovetter (1973) gave us the mechanism that explains why. His strength of weak ties argument says that novel information does not flow through strong ties within a closed group. Strong ties connect people who already know what each other knows. Weak ties, the acquaintances and cross-departmental contacts, are the bridges that bring new information into a cluster. The senior manager who champions AI in one department has weak ties to the IT team, to external vendors, to peers in other companies. Those ties carry information about which tools work, which vendors are reliable, and which use cases generate real returns. The department across the building, where nobody talks to the champion, is not a separate team. It is a structural hole.

Burt (1992) formalized that gap as structural hole theory. A structural hole is the empty space between two groups that do not share information. The person who bridges the hole has brokerage advantage, access to information from both sides that nobody else has. The AI champion is the broker. The champion's department adopts faster not because they are smarter or better funded, but because the champion sits at the bridge and the second department does not. The second department has no broker and no bridge. Rogers would say diffusion requires observability, you need to see someone using the technology to want it yourself. If the champion's use is invisible across the structural hole, observability is zero, and the S-curve never steepens.

This framework explains three patterns I see repeated in enterprise AI adoption.

First, adoption is uneven within firms because the network is uneven. Organizations allocate AI budgets on a rational model: calculate the expected return, distribute funds proportionally, expect proportional results. The actual diffusion pattern follows network centrality, not budget size. The department whose manager has high centrality in the organizational network adopts faster because information, social proof, and implementation support flow through the champion's ties. The department with low centrality, even with identical funding, never gets the same flow. Organizations treat AI adoption as a procurement process. They are managing a supply chain when they should be studying the social graph.

Second, the AI champion model works for a specific theoretical reason. The champion is not just enthusiastic. The champion occupies a brokerage position between IT and the business unit. That is a weak tie bridge across a structural hole. IT understands the technology. The business unit understands the problem. Without a broker, those two knowledge domains do not meet, and adoption stalls at the pilot stage. When organizations ask "how do we find AI champions," they are really asking "where are our structural holes, and who already sits at the bridge?" The answer is never a job description. It is a network position.

Third, copying a competitor's AI strategy fails for a reason that has nothing to do with the strategy itself. The strategy worked at company A because company A had a specific social network. The champion at company A had weak ties to the right groups, brokered the right structural holes, and occupied central positions. Company B buys the same tools, writes the same policy, allocates the same budget, and gets nothing resembling the same result. The strategy is portable. The social network that made it work is not. DiMaggio and Powell (1983) diagnosed this as mimetic isomorphism, organizations copy each other under uncertainty, but my earlier post on why everyone copies everyone else's AI strategy framed the mechanism specifically. Mimetic adoption produces convergent organizational forms without convergent organizational competence. The copy has the visible artifacts of strategy, the press releases, the product pages, the consulting engagements. It does not have the social infrastructure that made the original work.

The implication for organizations is uncomfortable. If you treat AI adoption as a technology procurement decision, you will measure the wrong variables. You will track licensing costs, pilot counts, and user registrations. You will miss the actual mechanism, which is whether the right people are talking to the right other people. Rogers would have told you in 1962 that you need to understand the social system. Granovetter would have told you in 1973 that the bridges are weak ties. Burt would have told you in 1992 that the structural holes are where the action is. The theory has been available for decades. Most organizations still act like adoption is a function of the IT budget line.

Meta description: Rogers, Granovetter, and Burt explain why the department with the AI champion adopts faster while the department across the building with the same budget does not.

Internal links:
- Why everyone copies everyone else's AI strategy for the mimetic isomorphism mechanism that explains why copying competitor AI strategies fails.
- Why AI strategy is a garbage can model for how solutions arrive in the organizational environment before problems are identified.
- The S-curve does not care about your launch date for a deeper unpacking of Rogers' diffusion framework and its five innovation attributes.
- TOE framework and why context beats best practices for the organizational context factors that interact with social network structure.

JSON-LD:

```json
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "AI Adoption Spreads Through Social Networks, Not Tech Budgets",
"author": {
"@type": "Person",
"name": "Ali Safari",
"url": "https://alisafari.space"
},
"datePublished": "2026-05-14",
"description": "Rogers, Granovetter, and Burt explain why the department with the AI champion adopts faster while the department across the building with the same budget does not.",
"keywords": ["diffusion of innovations", "social network theory", "AI adoption", "Granovetter weak ties", "structural holes", "Rogers", "organizational change", "information systems"]
}
```

Open Graph:
```html```