I Read DiMaggio and Powell While Watching Every Company Announce an AI Strategy

I was on my third read of DiMaggio and Powell (1983) and still getting the trap wrong in my notes. Coercive must be external. I wrote it in the margin. I wrote it on a sticky note. I wrote it in Persian so it would stick differently. I was not reading for insight anymore. I was reading to memorize the boundary condition that my professor had warned would cost me points on the comps exam, the one where students use coercive isomorphism to describe an executive mandate when coercive isomorphism specifically requires external regulatory pressure.

And then I opened a news tab and saw that a company I had never heard of had announced its AI-first strategy. The press release had a quote from the CEO about transformation. It listed a partnership with an API provider. It promised that AI would be embedded across all products by next quarter. I had seen the same press release maybe forty times in the previous six months, from enterprise software companies and logistics firms and healthcare systems and banks. The logos changed. The wording did not. DiMaggio and Powell wrote about this in 1983. They called it mimetic isomorphism, and what I was looking at was the purest example I will ever see in real time.

I closed the news tab and opened another. Same press release, different logo. And then I understood something about studying theory that nobody had told me. A well-built theoretical lens does not stay in the exam room. It follows you out into the world and changes how ordinary things look. I could not read a corporate announcement anymore without seeing the isomorphism types. I could not watch a regulatory hearing about AI without mapping every argument onto Scott's three institutional pillars. The theory that I had been treating as a dry framework for answering exam questions about organizational adoption had become the only thing I could see when I looked at the technology industry.

The mimetic pressure was the easiest to spot, because it was everywhere. Every boardroom in 2024 and 2025 faced the same question: what is our AI strategy? Nobody had a reliable answer, because the technology was evolving faster than anyone could evaluate it, and under genuine uncertainty the safest organizational move is to copy whoever looks like they know what they are doing. So companies copied. They copied the press release structure. They copied the organizational structure, announcing a Chief AI Officer because their competitor had one. They copied the product roadmap, bolting a chatbot onto whatever they already sold and calling it AI transformation. DiMaggio and Powell would recognize this pattern instantly. Mimetic isomorphism is not about irrationality. It is about legitimacy under uncertainty. When nobody knows the right answer, the right answer is whatever the perceived leader is doing. The press releases looked identical because the organizations were not independently reasoning toward an AI strategy. They were reasoning toward the appearance of one.

The coercive pressure was slower to arrive but harder to ignore. The EU AI Act, GDPR enforcement actions that started reaching into AI training data practices, sector-specific regulations from financial and healthcare authorities. These are the external regulatory forces that DiMaggio and Powell meant when they wrote about coercive isomorphism. And my professor's trap was vindicated: when I heard executives say "we are adopting AI governance because the board demands it," that was not coercive isomorphism. That was an internal mandate. The EU AI Act imposing fines on a company that deploys a high-risk system without a conformity assessment, that is coercive isomorphism. The difference matters because internal mandates can be reversed by the same executive who issued them. External regulatory pressure persists regardless of who is in the corner office. Every time I see a consulting firm's webinar titled "Preparing for AI Regulation," I see the coercive mechanism propagating through the field, not just at the level of individual firms but at the level of the whole industry structure that Scott described through the regulative pillar.

The normative pressure is the one I think will be the most durable, because it operates through the mechanism that is hardest to resist: professional identity. Consulting firms are building AI maturity models. Industry associations are publishing responsible AI frameworks. Professional certifications for AI governance are emerging. Over time, these standards become what a competent AI professional recommends, and because the professionals who hold these certifications circulate through the same conferences and hire from the same talent pool, their recommendations converge. Organizations that hire these professionals get the same advice regardless of context. That is normative isomorphism. Scott's second pillar captures it perfectly. The logic is not expedience, as it is with regulation. It is social appropriateness. You adopt the practice not because you must, but because a professional in your position is expected to.

The cultural-cognitive pillar is the one I keep coming back to because it is the hardest to see from inside. Scott described it as the layer of taken-for-granted shared understandings that operate below conscious awareness. The logic is not expedience or appropriateness. It is ontological reality. This is just how things are done. I saw it happening in real time with AI strategy. At the beginning of 2023, a company could reasonably decide not to have an AI strategy. By late 2025, the absence of an AI strategy was itself a statement requiring justification. The question had shifted from "should we adopt AI?" to "which AI vendor should we use?" That is the cultural-cognitive pillar closing. The first question is the strategic one. The second question is a procurement decision that assumes the first question was already answered. And the first question was answered not through analysis, but through the accumulation of mimetic, coercive, and normative pressure that made the alternative inconceivable.

I wrote about some of this before. In my post on institutional isomorphism and AI adoption, I worked through the three isomorphic pressures as an analytical framework applied to industry patterns. And in the post about the EU AI Act hitting all three of Scott's pillars, I traced how a single piece of regulation operates simultaneously at the regulative, normative, and cultural-cognitive levels. Those posts came from the same theoretical foundation, but they approached it as analysis. This post is different. It is about what happens to a person who studies theory seriously enough that theory rewires perception.

I did not expect that to happen. I expected comps preparation to be about memorizing definitions and practicing comparisons. I did not expect to walk away from my study desk, open Twitter, and see a screenshot of an AI press release that I could diagnose like a clinical case. I did not expect to watch a congressional hearing about AI safety and map every senators question onto institutional pillars faster than I could process what the witness was saying. The theory did not stay on the page. It became part of how I see.

This is what I think the comps process is actually for, beyond the exam. It forces you to hold a theoretical lens long enough and close enough that the lens becomes part of your visual system. You stop looking at the theory and start looking through it. And once that happens, the theory is tested every time you read a news article, every time you watch a company announce something, every time you see an industry converge on a practice that nobody can quite justify. The theory either holds up under that pressure or it does not. Institutional theory held up for me. I saw convergence happening in plain sight, and the framework explained why without forcing the evidence.

The question that stays with me is what happens after the isomorphism settles. DiMaggio and Powell told us why organizations converge. Scott told us what makes convergence durable. But neither framework tells us what happens to a specific organization after it has adopted AI because everyone else did. That question needs structuration theory to trace how AI and organizational practice co-evolve after adoption. It needs affordance theory to specify what AI enables for which actor with which goal, because the affordances that emerge in the post-isomorphic organization are not guaranteed by the isomorphic adoption that preceded them. I am not sure I have the answer to that yet. But I know the question would not occur to me at all if I had not spent the last few months reading DiMaggio and Powell until the trap finally stuck, coercive must be external, and then looked up from the page to find the theory everywhere.