Marketing Runs on IS It Cannot Name

Recent industry reports project martech spending exceeding $20B annually, and I keep thinking about what that number hides. The dashboards, the customer data platforms, the attribution models, the pipeline trackers, the email automation sequences, the AI-generated copy tools. Marketing departments now operate some of the most technically complex information systems in any organization, and almost none of that complexity is studied through an IS theoretical lens. The IS discipline has the theories. Marketing departments are living inside the phenomena those theories describe. The gap is not in the theory. The gap is in who knows the theory exists.

I noticed this while reading Davis (1989) for the third time this semester. TAM gives you two constructs: perceived usefulness and perceived ease of use. You point them at a CRM rollout, run your survey, and you can predict whether a sales team will adopt Salesforce. The model works. It has worked across hundreds of studies. But there is a question TAM cannot answer, and it is the question that actually matters in a marketing department. Why does the same Salesforce instance produce radically different outcomes in different marketing teams? One team builds elaborate opportunity tracking, cross-references campaign data with pipeline stages, and generates attribution reports that reshape budget allocation. Another team enters the minimum required data, ignores the dashboard, and treats the system as a glorified contact list. TAM predicts both teams will adopt. It says nothing about what happens after the login screen.

This is not a TAM problem. It is a TAM boundary condition. Davis built TAM to explain acceptance, which is a binary gate. You either use the system or you do not. But once usage crosses that threshold, the interesting variation is no longer whether people use the tool. It is how they use it, what structures they enact through it, and what those structures then make possible or impossible. The tool is constant. The outcome varies. The productivity paradox is still alive precisely because adoption does not guarantee value. Value emerges from structuration.

Orlikowski (1992) gave IS the vocabulary for this. She argued that technology is both a product of human action, designed and built through organizational processes, and a medium of human action, shaping what people can do and how they do it. This duality means the same technology enacted in different structural contexts produces different technologies-in-practice. A CRM system is not a fixed artifact with fixed effects. It becomes what each team makes of it through recurrent use. The sales team that turns Salesforce into an attribution engine enacts a different technology-in-practice than the team that uses it as a digital Rolodex. The software is identical. The structural outcomes are not. This is exactly what I wrote about when I covered how the same tool produces different outcomes in different departments. Marketing departments are dense with examples of this, and the IS research community barely studies them.

DeSanctis and Poole (1994) sharpened the mechanism. They introduced Adaptive Structuration Theory, which says that technologies come with structural features, the rules and capabilities built into the system, and a spirit, the designer's intended use pattern. Groups then appropriate those structures, and appropriation can be faithful or unfaithful. When a marketing team uses a CDP exactly as its designers intended, building unified customer profiles and activating segments across channels, that is faithful appropriation. When a team buys the same CDP and uses it only to store email lists because the organizational structure does not support cross-channel orchestration, that is unfaithful appropriation. The structural features are there. The spirit is available. The appropriation diverges because the surrounding structures of meaning, power, and legitimacy push use toward the path of least resistance.

I think this is the single most important theoretical gap in how organizations understand their martech stacks. Marketing leaders spend enormous sums on features and integration. They evaluate vendors on capability lists. They compare platforms by counting what each one can do. But the feature list is Orlikowski's structural features, and structural features do not determine outcomes. The spirit of the technology does not determine outcomes. The appropriation determines outcomes, and appropriation depends on interpretive schemes, resource distributions, and normative structures that live inside the marketing organization, not inside the software.

The gap gets wider when you consider what marketing teams are actually doing with AI tools in 2026. When a marketing operations team configures an AI-driven attribution model, are they using the model, or are they delegating a judgment task to an agentic system? Baird and Maruping (2021) made this distinction precise. They argued that agentic IS artifacts, systems with autonomy that can initiate action and accept responsibility for tasks under uncertainty, require a different theoretical lens than traditional use. Delegation, not use, is the right construct. The delegation framework identifies three mechanisms: appraisal, whether the agent can perform the task; distribution, how subtasks are allocated between human and system; and coordination, how interdependencies between human and agent actions are managed. These mechanisms are structurally absent from TAM. You cannot reduce appraisal, distribution, and coordination to a perceived usefulness item without losing the causal story entirely.

This matters for marketing because the field has moved aggressively into agentic AI without knowing it. Marketing attribution platforms now make autonomous decisions about which touchpoints receive credit for conversions. AI copy generation tools produce content and push it through approval workflows with decreasing human oversight. Customer segmentation algorithms determine which audiences see which messages. When a marketer sets up an attribution window, they are not using a tool in the TAM sense. They are configuring delegation parameters. They are deciding how much uncertainty the algorithm will handle, where the human override happens, and what feedback loops connect model output to future model behavior. Baird and Maruping's four delegation archetypes, reflexive, supervisory, anticipatory, and prescriptive, map directly onto what marketing teams are doing. Reflexive delegation is the automated A/B test that runs without human judgment. Supervisory delegation is the analyst monitoring model performance dashboards. Anticipatory delegation is the predictive audience model that pre-segments before the campaign brief is finished. Prescriptive delegation is the AI that generates and sends email variants within guardrails. Stop counting users and start measuring delegation, because the construct has fundamentally shifted.

Marketing departments do not know they are doing structuration. They do not know they are managing delegation. They call it "tool adoption" or "tech stack optimization" or "digital transformation," which is an umbrella term so broad it explains nothing. The IS discipline has spent forty years building the theoretical apparatus to understand exactly what happens when marketing teams appropriate, deform, or faithfully enact their martech stacks. The structurational model of technology explains why the same CRM produces different outcomes in different departments. AST explains why faithful and unfaithful appropriation diverge. Delegation theory explains why "use" is the wrong construct for describing what happens when a marketer hands a judgment task to an algorithm. And the TOE framework explains why some marketing organizations are structurally ready for these tools and others are not, even when they buy the same ones.

The irony is that marketing practitioners sense this gap. They write about "tech stack bloat" and "tool fatigue" and "why our CRM investment is not paying off." They know something is wrong. They describe the symptoms precisely. What they lack is the theoretical vocabulary that would let them name the mechanism, and that vocabulary already exists. Orlikowski's duality of technology, DeSanctis and Poole's faithful and unfaithful appropriation, Baird and Maruping's delegation mechanisms. These are not abstract frameworks with no application. They are diagnostic tools waiting for the domain that needs them most. Marketing runs on information systems that no one in the marketing department knows how to theorize. The IS field has the theory. The marketing world has the phenomenon. Someone needs to close the distance between them.

The next time a CMO asks why their $2M martech stack is underperforming, the answer is not in the feature comparison. It is in the appropriation.