Weick's sensemaking theory says people act first and make sense later, preferring plausibility over accuracy. That explains everything from the Twitter rebrand to EU AI regulation.
The Twitter rebrand to X happened in a single weekend in July 2023. Logo swapped, domain changed, brand evaporated. No consultation, no rollout plan, no stakeholder alignment. Overnight, thousands of employees had to explain to advertisers, users, and each other why a platform they had built careers around was suddenly called something else.
I kept thinking about that moment while reading Weick (1995). Not because it was chaotic, obviously it was, but because of what happened next. People inside the company and outside it started making sense of what had happened after the fact. They constructed narratives about why this made sense, what the strategy was, what it meant for the future. The decision came first. The meaning came second. That is the core of Weick's sensemaking theory, and I think it is the most underused powerful lens in information systems.
Weick (1995) defines sensemaking through three subprocesses: enactment, selection, and retention. Enactment flips the rational decision-making model on its head. People do not decide then act. They act, and their actions create the situations they then have to understand. When an organization rolls out a new technology, the people inside are not sitting down with a clean decision matrix. They are reacting, doing things that seemed right in the moment, and then retroactively building an explanation for why those things made sense. Selection means that when faced with ambiguity, people choose from available frames and schemas to interpret what happened. The interpretation that gets selected has to be plausible enough to guide further action. Retention means those selected interpretations get stored in organizational memory and shape what will be enacted next time.
Weick also specifies seven properties of sensemaking. It is grounded in identity construction, retrospective, enacted through action, social, ongoing, focused on extracted cues, and driven by plausibility over accuracy. That last one is the property that hit me hardest. In sensemaking, people prefer a plausible interpretation that enables action over an accurate interpretation that is unattainable or paralyzing. Organizations do not wait for perfect information. They build stories that are good enough to move forward. That is not irrational. It is how sensemaking works.
This is where Maitlis (2005) becomes useful. Her framework identifies four forms of organizational sensemaking varying along two dimensions: leader engagement and issue diversity. Guided sensemaking has high leader engagement and focused issues. Restricted sensemaking has high leader engagement but diverse issues, so leaders dominate interpretation and the range of acceptable narratives stays narrow. Fragmented sensemaking has low leader engagement and diverse issues, producing multiple local interpretations that never cohere. Minimal sensemaking has low leader engagement and focused issues, where sensemaking happens in narrow pockets without organizational visibility.
The Twitter rebrand fits restricted sensemaking because leadership drove the change with high engagement, but the issues were diverse. Advertising revenue, user identity, platform governance, legal compliance, brand equity. All of these needed interpretation, and the interpretations were constrained at the top. People inside the organization did not get to make collective sense of what happened. They were told, and then they built plausible stories around the top-down narrative. Some of those stories were about innovation. Some were about destruction. The point is that the sensemaking was restricted, not because people were irrational, but because the structure of the organization channeled interpretation in a particular direction.
Zoom's 2020 privacy scandal is a different pattern. Zoom had a massive adoption spike when COVID hit, and almost immediately security researchers discovered that Zoom was routing calls through Chinese servers, misrepresenting end-to-end encryption, and leaking data to Facebook through its SDK. Different groups inside and outside Zoom interpreted these events in fundamentally different ways. Security researchers saw negligence. Executives saw a scaling crisis. Users saw a betrayal. Regulators saw a compliance failure. This is Maitlis's fragmented sensemaking: low leader engagement combined with diverse issues, producing multiple local interpretations that never cohere into a shared organizational narrative. Zoom's leadership did not immediately control the sensemaking, and the interpretations multiplied uncontrollably before the company could impose a coherent story.
Both cases illustrate something I think IS research underappreciates. Adoption models assume beliefs form before action. Sensemaking says the sequence runs the other way. This is also why institutional pressure and sensemaking reinforce each other: when organizations adopt because everyone else is adopting, the adoption is not a rational evaluation but a plausible story constructed under pressure (see the pattern I wrote about in how everyone copied everyone else's AI strategy). People act, then they construct beliefs that justify those actions. The first question is not "what do we think about this?" It is "what are we doing with this?" and only later "what does it mean?" If sensemaking precedes decision-making, then organizations that try to manage sensemaking after a disruption are doing exactly what they should be doing. But most change management frameworks still assume you can decide your way into clarity. You cannot. You can only enact your way toward a plausible interpretation.
Now, most sensemaking theory is retrospective: people make sense of things that have already happened. Seidel, Frick, and vom Brocke (2025) introduce prospective sensemaking, which flips the temporal direction. Prospective sensemaking asks how organizations make sense of things that have not fully happened yet, specifically emerging technologies. Their context is regulation. The Collingridge dilemma says that when a technology is new, it is easy to change but hard to understand; by the time you understand it, it is too entrenched to change. Seidel and colleagues argue that regulators can resolve this dilemma through two complementary sensemaking processes: abstraction and elaboration. Abstraction extracts essential properties from a technology to write technology-neutral regulation that stays flexible as the technology evolves. Elaboration specifies detailed requirements for legal certainty and user protection. The interplay between abstraction and elaboration, along with three reconceptualization levels of technology, use, and roles, allows regulation to be simultaneously flexible and specific. This is prospective sensemaking because regulators are building interpretations of a future they have not yet experienced, and they need those interpretations to be plausible enough to act on now.
The EU AI Act is a live case of prospective sensemaking at work. Regulators had to write rules for AI systems that did not exist yet. They used abstraction by defining risk categories rather than naming specific technologies. High-risk AI systems are defined by their use case and impact, not by whether they use a particular model architecture. At the same time, they used elaboration by specifying concrete requirements: transparency obligations, human oversight mechanisms, conformity assessments. The abstraction keeps the regulation flexible as AI evolves. The elaboration gives companies and courts something specific to work with. The tension between these two is exactly the Collingridge dilemma playing out in real time, and sensemaking theory gives us the vocabulary to see it.
I think sensemaking is the most underused powerful theory in IS. It explains why technology adoption is never just about evaluation: if people enact before they interpret, then post-adoption sensemaking is where the real action is. The interesting question is not what organizations believe before they adopt, but what stories they tell after they have already committed. This aligns with the emergent view of technology and organizations, where outcomes arise from interaction over time rather than from any single cause (I wrote about that in why technology neither determines nor submits). Maitlis's four forms also give us a diagnostic. When a technology implementation is failing, I can ask: is this guided, restricted, fragmented, or minimal sensemaking? The answer tells me where the intervention should go. And Seidel and colleagues show that sensemaking is not only retrospective. Regulators making rules for AI systems that do not yet exist are doing prospective sensemaking, and the abstraction-elaboration distinction gives us a way to understand how they balance flexibility with specificity.
The trap, and I have seen this in comps answers, is confusing sensemaking with decision-making. Weick (1995) explicitly distinguishes the two. Decision-making assumes a set of alternatives and a rational process of choosing among them. Sensemaking assumes ambiguity and retrospective construction of meaning. Sensemaking precedes and enables decision-making. It does not replace it. This distinction maps onto the variance versus process theory split that Mohr (1982) identified: sensemaking is process logic, not variance logic, because the sequence of enactment, selection, and retention matters more than the strength of any single variable (if that distinction sounds familiar, I wrote about it in why your regression cannot explain a divorce). But if you skip sensemaking and jump straight to decision-making, you are building decisions on interpretations that nobody has had time to construct. That is what happened at Twitter. That is what happens whenever an organization adopts a technology without leaving room for people to make sense of what they are actually doing with it.
The next time you see an organization explain why it adopted something, ask yourself: did they decide, then act, then explain? Or did they enact, then select a plausible story, then retain it as if it had been the plan all along?
About the author
Share
More notes
Related notes