Bought a $2M Platform, Nobody Learned a Thing

The number that surprised me in Cohen and Levinthal (1990) was not the definition. It was the path dependence. Absorptive capacity, the organizational ability to recognize the value of new external information, assimilate it, and apply it to commercial ends, depends on prior related knowledge. That means an organization that has not invested in building knowledge in a domain cannot absorb new knowledge in that domain, no matter how much money it spends acquiring it. This is not a flaw in the platform. It is a structural feature of learning itself.

I kept noticing this pattern everywhere after I read that paper. Organizations buy state-of-the-art analytics platforms and the data sits there. They purchase enterprise knowledge management systems and nobody uses them. They hire consultants to build dashboards and the dashboards become expensive wallpaper. The technology is present. The ability to learn from it is absent. Cohen and Levinthal gave us the mechanism, and it is more uncomfortable than most organizations want to admit. You cannot learn from new knowledge if you have no prior related knowledge to connect it to. The firm that has never made data-driven decisions, never hired analysts who ask questions of data, never built routines for acting on insights, cannot suddenly absorb data intelligence just because a vendor sold them a platform.

Absorptive capacity is cumulative, which makes the problem self-reinforcing. The organizations that already know something learn faster, because their prior knowledge lets them recognize which new information matters, understand it quickly, and apply it effectively. The organizations that know nothing fall further behind. High absorptive capacity enables faster learning, which builds more knowledge, which raises absorptive capacity further. Low absorptive capacity means you cannot even tell what you are missing, so you do not invest in what would help, so you fall further behind. This is the trap, and I have never seen a technology purchase that broke it by itself.

Roberts et al. (2012) reviewed how IS research has used the absorptive capacity construct, and they found something that bothered me. The concept gets applied at contradictory levels of analysis. Some papers treat it as an individual skill, some as a team property, some as an organizational capability, and they rarely specify which. Roberts et al. clarified the IS-specific operationalization through four phases: acquire (recognize and acquire relevant external knowledge), assimilate (analyze, process, interpret, and understand the information), transform (combine prior knowledge with new knowledge), and exploit (apply the transformed knowledge to create competitive advantage). I want to be careful with the wording here because the original Cohen and Levinthal (1990) paper says "recognize the value of new information, assimilate it, and apply it to commercial ends." The "exploit" phrasing and the four-phase decomposition came later through Zahra and George (2002), who reconceptualized absorptive capacity into potential absorptive capacity (acquire and assimilate) and realized absorptive capacity (transform and exploit). That distinction matters more than it sounds like it should.

Potential absorptive capacity is what lets you get the knowledge in the door. Realized absorptive capacity is what lets you do something with it. A company can be strong on potential and weak on realized. It acquires data, buys subscriptions, hires talent, builds data lakes. But if it has no routines for transforming knowledge into action, the knowledge stays in the lake and the company stays in the same place it was before it spent the money. This separation explains something I have seen repeated in enterprise analytics. The organization succeeds at acquisition. The dashboards exist. The reports flow. But nothing changes in decision-making because there is no mechanism to transform insights into organizational routines.

The Zahra and George (2002) reconceptualization is where I think the practical force of absorptive capacity becomes clear. Potential without realized is the most common failure mode I see. The organization has the data, the tools, and maybe even the people, but it lacks the decision routines, the authority structures, and the feedback loops that turn absorbed knowledge into organizational action. You can see this in hospitals that bought analytics platforms during COVID. Some used real-time data to redistribute resources and adjust protocols, which is realized absorptive capacity in action. Others had the same data on the same dashboards and made slower, worse decisions because the organizational routines for acting on data did not exist. The data was not the difference. The routines were.

This connects directly to what I wrote about the productivity paradox. Brynjolfsson named mismanagement as one of four explanations for why IT spending fails to produce measurable productivity gains. Mismanagement, in absorptive capacity terms, is the failure to build realized capacity. You buy the tool. You skip the complementary organizational changes. If you accept that IT business value travels through complementary resources and process change, as Melville et al. (2004) showed, then absorptive capacity is part of the mechanism that determines whether those complementary resources actually get built. The firm with high absorptive capacity builds the complementary resources around its technology investments. The firm with low absorptive capacity buys the same technology and changes nothing.

The same logic applies to AI adoption in 2026. Every organization can access the same large language models. The compute is commodity. The APIs are public. What separates the firms that get value from AI from those that do not is not the model. It is prior knowledge, which determines whether the organization can recognize which AI applications matter, and realized capacity, which determines whether it can transform those applications into changed routines. This is absorptive capacity at work, and it predicts that organizations with no history of data-driven decision-making will struggle to exploit AI no matter how much they invest. The productivity paradox was never about the technology being insufficient. It was about the organization being unable to absorb what the technology offered.

There is also a connection to what theory is and is not. Cohen and Levinthal built a theory with a real mechanism. Prior knowledge enables recognition, recognition enables assimilation, assimilation enables application. The mechanism is why-logic, not just a list of variables. Whetten (1989) would say the paper has the what (absorptive capacity), the how (path dependence and cumulativeness), and the why (prior knowledge as the enabler of each subsequent stage). It also has clear boundary conditions: absorptive capacity is domain-specific, which means investing in marketing knowledge does not automatically help you absorb engineering knowledge. Zahra and George (2002) sharpened the boundary by separating potential from realized, which predicts exactly where organizations get stuck. They acquire but never transform.

I think this is one of the most exam-ready theories in the organizational IS toolkit precisely because it names a mechanism that most organizations ignore. The instinct is to buy the tool. The mechanism says you cannot use the tool without prior knowledge. Every delegation decision about what to hand off to an AI agent presumes someone in the organization can evaluate whether the agent's output is any good. That evaluation requires prior knowledge. Without it, delegation becomes blind trust, which is not delegation at all.

The uncomfortable implication of absorptive capacity is that technology investments are path-dependent and partly irreversible. The organization that never built its data muscles cannot suddenly become data-driven by writing a check. The organization that never developed decision routines around evidence cannot develop them overnight during a crisis. The capacity has to exist before the crisis arrives. The organizations that responded effectively during COVID were the ones that had already been making data-driven decisions, already had analysts asking hard questions, already had routines for converting evidence into action. They had high absorptive capacity before the pandemic made it urgent.

Knowledge acquisition is path-dependent. Organizations that have not invested in prior related knowledge cannot absorb new insights no matter how expensive the technology. This is not a statement about budgets. It is a statement about what budgets can and cannot buy.