The [YOU] on AI cycle uses the left-hand column as a frame for the interior experience that the transition generates but organizational life rarely surfaces. When a senior marketing director reads an AI analysis that identifies an opportunity she missed in her own assessment, the right-hand column says: “This is a good starting point but needs refinement.” The left-hand column says: the tool saw what I could not see, and endorsing it fully would mean admitting I was wrong in last month’s strategy meeting, and the framing I am choosing right now is shaped by the need to reassert my authority rather than by any analytical assessment of the output’s quality. The left-hand column is not dishonesty. It is the structure of professional self-presentation in organizations where governing variables around competence, control, and identity operate continuously below the surface of every professional interaction.
The exercise becomes particularly clarifying when applied to the interpersonal dynamics that AI tools create between senior and junior professionals. The senior developer who reviews a junior colleague’s AI-augmented code and finds it excellent—better structured, more thoroughly tested than anything the junior previously produced—may write in the right-hand column: “Nice work. The architecture here is solid. A few minor suggestions.” The left-hand column records a governing-variable crisis: this is better than what I would have produced, and if the tool can enable a junior colleague to produce work at this level, what is the value of twenty years of experience? The crisis is real and the left-hand column is where it lives, shaping subsequent behavior—less willingness to assign complex work to the junior, more skepticism toward AI tools in general—without ever entering the organizational conversation where it might be examined and resolved.
Argyris developed the left-hand column technique as part of his broader action science methodology, which required that research produce information that is directly observable, publicly testable, and independent of the defensive interests of the people producing it. The technique operationalizes this requirement: it forces the practitioner to make visible the private reasoning that shapes public behavior, in a form concrete enough to be examined, tested, and if necessary revised.
The exercise was not designed to produce emotional venting or personal disclosure for its own sake. It was designed to generate the specific kind of information that Argyris called actionable knowledge: knowledge that is precise enough to guide behavior, testable against observable evidence, and honest enough to acknowledge the gap between what is known and what is assumed. The left-hand column, properly analyzed, reveals the governing variables that are operating in a specific interaction—the assumptions about competence, identity, and worth that shape behavior without being available for examination—and that revelaton is the precondition for double-loop learning.
The information that does not travel. The left-hand column is not a private supplement to public discourse; it is the primary location of the valid information organizations need. Professionals who work with AI tools carry detailed private assessments of what the tools can and cannot do, how they change the relationship between experience and output quality, which professional activities add genuine value in the AI-augmented environment and which are sustained by tradition and defensive routine. This information stays in the left-hand column because the organizational norms that reward confident competence make its disclosure dangerous. The result is a collective information deficit precisely where the organization most needs information.
The organizational left-hand column. Argyris extended the concept from individual to collective: organizations have left-hand columns too. The undiscussables—the topics everyone recognizes as important but that organizational norms forbid raising—are the organizational left-hand column. In the AI transition, the undiscussable list is long and consequential: that AI tools can produce work of comparable quality to experienced professionals, that junior employees often adopt AI more effectively than senior ones, that current evaluation criteria may reward the wrong skills, that the business model may be incompatible with the efficient use of AI. Each undiscussable is a left-hand column entry at the organizational level.
The conditions for surfacing. Bringing left-hand-column material into conversation requires three specific conditions that most organizations lack: psychological safety sufficient to permit honest disclosure without professional risk, analytical rigor sufficient to transform the disclosure into productive inquiry rather than emotional venting, and organizational commitment sufficient to sustain the practice over time rather than treating it as a one-time exercise. These conditions are structural—they must be built into evaluation systems, leadership behavior, and meeting formats—and they are the organizational equivalent of the orange pill: you cannot unsee what the left-hand column reveals.