What if the data doesn't exist?

Much of what we see in Enterprise AI focuses on automation of existing workflows and information synthesis. Organizations are exploring how AI can help employees complete tasks more efficiently, automate workflows, and improve access to information. The resulting solutions often help people retrieve knowledge buried in traditional systems, generate reports, answer questions, and execute processes that already exist.

These developments are meaningful and will undoubtedly create substantial value. However, they also share a common assumption: the work itself has already been defined. The workflow exists. The data is already modeled. The process is already there. The organizational function just needs to be optimized. AI is being applied to improve the efficiency with which that work is performed.

What often isn't discussed is the possibility that AI may enable entirely new categories of organizational capabilities that organizations have historically been unable to operate; not because they lacked importance, but because they were too costly or complex to maintain.

In enterprise software, most systems were built around things that could be represented and managed economically. Customer relationships became CRM systems. Projects became project management systems. Support interactions became ticketing systems. Financial transactions became ERP systems. In each case, software emerged around a category of information that could be sufficiently structured, maintained, and governed to justify the investment. Intentionally around artifacts that were already maintained and processed by humans. Moving to a database was a more effective, digital filing system.

The result was a generation of systems designed primarily around refiling existing organizational artifacts: projects, tasks, tickets, documents, records, and transactions. These systems became indispensable because they helped organizations coordinate work at scale. Yet they also reflected an important limitation. They captured the outputs of existing organizational activity far more effectively than they captured the organizational dynamics that ultimately shaped those outputs.

Many of the most important functions within an organization exist at a different level of abstraction. For example, dependency management is not merely the act of recording dependencies between projects. It's the broader organizational capability of understanding how priorities, team structures, resource allocation decisions, and execution patterns create coordination costs across the enterprise. For capacity planning, where the challenge is not simply estimating available resources, it's continuously understanding how changing priorities, staffing decisions, constraints, and strategic objectives affect an organization's ability to execute. And portfolio management isn't about getting the dashboard just right. It's where leaders must evaluate competing investments, assess tradeoffs, understand downstream consequences, and continuously adapt decisions as conditions change.

These are not discrete workflows. They are ongoing organizational functions. They require continuous interpretation, synthesis, evaluation, and adaptation. Historically, organizations have addressed them through combinations of planning rituals, operational reviews, ad hoc spreadsheets, governance processes, consultants, PMOs, docs, and decks. Even then, these activities are often periodic rather than continuous because maintaining an accurate understanding of organizational reality is extraordinarily expensive.

The challenge has never been that organizations lacked the desire to understand these dynamics. The challenge has been that understanding them requires a system that actively maintains the context. And if that system has to be dependent on human data entry, it quickly becomes untenable.

The most significant contribution of AI may not be its ability to execute tasks, but its ability to reduce the cost of creating and maintaining these systems in support of these incredibly costly capabilities. For the first time, organizations have access to systems capable of continuously reconciling information across fragmented sources, extracting meaning from unstructured interactions, maintaining relationships between concepts, and updating representations of reality as conditions evolve.

What if the opportunity is not building better interfaces for existing systems, but operationalizing organizational capabilities that were previously uneconomical.

Rather than beginning with a workflow, what if organizations began with the capability they wish to support? If the goal is to continuously evaluate strategic investments, what information is required? What decisions must be modeled? What relationships influence outcomes? If the goal is to understand organizational capacity, what constraints, resources, objectives, and dependencies must be represented? If the goal is to improve coordination across teams, what aspects of organizational behavior must be continuously observed and understood?

Answering these questions inevitably leads to the need for an operating ontology: a structured representation of the organization and the relationships that define it. An operating ontology provides the foundation upon which higher-order capabilities can operate. It captures not only the existence of things like projects, teams, initiatives, and objectives, but also the relationships between them, their properties, how they interact, and their purpose. Most importantly, it's expected to adapt over time.

Historically, maintaining such a model would have required enormous amounts of manual effort. Organizations would have needed system administrators to curate data, analysts to reconcile inconsistencies, and governance processes to keep information current and up to date. The cost of maintaining the system often exceeded the value generated by the resulting insights. Consequently, organizations accepted simplified representations of reality and limited themselves to capabilities that could be supported by those representations.

AI changes the physics of this problem. When systems become capable of continuously collecting, curating, maintaining, enriching, and updating themselves, entirely new possibilities emerge. Organizations can begin to operate functions that previously required dedicated teams or could only be performed periodically. Dependency management can become a continuously running capability rather than an occasional planning exercise. Capacity planning can evolve from static forecasting into an adaptive understanding of organizational readiness. Portfolio management can become a dynamic process of evaluating tradeoffs and reallocating resources in response to changing conditions. Strategic planning can incorporate simulation, scenario analysis, and continuous reassessment rather than relying solely on periodic review cycles.

The significance of this shift is that it moves software beyond its traditional role as a system of record or a system of workflow. Instead, software begins to function as a system of action. Its purpose is to continuously perform forms of higher-order organizational work that were previously too expensive, too complex, or too disruptive to operationalize.

The most important AI systems of the coming decade may not resemble the categories of software we recognize today. They may not be defined primarily by search, automation, or productivity. Instead, they may be defined by the organizational capabilities they enable. The jobs they do. Their value will be measured not by the number of tasks they automate, but by their ability to help organizations sense-make, coordinate effectively, make better decisions, adapt to changing conditions, and self-improve. And ultimately, their ability to continuously run organizational capabilities that have never before been practical to operate at scale.