Undiscussables are subjects that are simultaneously (a) known by most members of an organization, (b) consequential to the organization's functioning, and (c) structurally prevented from entering the organization's legitimate conversation. The fact that they are undiscussable is itself typically undiscussable — a second-order defensive structure Argyris called undiscussability of undiscussability. In the AI workplace, the undiscussables are extensive: the senior engineer's suspicion that her expertise has lost its market value, the executive's private uncertainty about what the organization is actually for, the junior employee's fear that admitting AI dependence will disqualify her for promotion. None of these can be spoken without social cost, and the cost is sufficient to keep the most consequential information out of the conversations that most need it.
The undiscussables concept is the operational expression of Argyris's theory of defensive routines. Routines protect; what they protect are the undiscussables; the protection is maintained by norms so pervasive they operate invisibly.
The AI transition has dramatically expanded the inventory of undiscussables in most organizations. Topics that would have been discussable five years ago — honest assessment of skill obsolescence, candid discussion of career trajectory, frank evaluation of organizational purpose — have become radioactive in proportion to how consequential they have become. The organizations that most need to discuss these topics are, by Argyris's framework, structurally prevented from doing so.
The silent middle is a population that holds undiscussables privately because they cannot be spoken publicly without penalty. The hallway confession — the honest assessment delivered to a trusted colleague in a private moment — is the canonical form of undiscussable speech. It is speech that knows it cannot survive translation into the legitimate conversation.
Breaking undiscussables requires what Argyris called organizational learning of the second kind: structural changes that alter not merely what can be said but the conditions under which saying it carries acceptable cost. Mere exhortations to openness fail reliably; the norms that sustain undiscussability are not dissolved by encouragement to violate them.
Argyris developed the concept through his work with executive teams in the 1970s and 1980s, where he consistently observed that the most consequential issues in organizational life were precisely the ones that organizational conversations could not accommodate.
His methodological innovation was to make undiscussability itself discussable in structured intervention settings — a move that typically produced intense defensive activity and, occasionally, genuine learning.
Known but unspeakable. Undiscussables are distinguished from secrets by the fact that they are widely known; the constraint is not on information but on legitimacy of speech.
Second-order undiscussability. The fact that a topic is undiscussable is itself typically undiscussable. This double structure is what makes the phenomenon so difficult to interrupt: speaking about what cannot be spoken about is itself a violation.
AI expansion. The AI transition has produced a rapid expansion in the inventory of organizational undiscussables, concentrated in precisely the topics whose discussion would support genuine adaptation.
Structural remedy. Making undiscussables discussable requires structural change — protected channels, altered incentives, modeled behavior from leadership — not merely cultural exhortation.
Some critics argue that the concept pathologizes normal organizational discretion — that some things cannot be said because they genuinely should not be said, not because defensive routines prevent their speaking. Argyris's framework distinguishes between these: undiscussables are specifically topics whose non-discussion has negative consequences for organizational functioning, not topics whose discretion serves legitimate interests.