Building a Cross-Functional AI Steering Committee That Doesn’t Stall

Why Most AI Steering Committees Stall

The pattern of steering committee stall in pharma is recognizable and consistent. Three structural failures account for nearly all observed stalls.

The first is membership inflation. The committee starts with eight to ten members from the most affected functions, but quickly expands to twenty or more as additional functions request representation. At twenty members, decisions cannot be made in the room; the conversation defaults to broadcasting status and gathering input, with decisions deferred to smaller follow-up forums. The committee becomes a status-broadcasting body rather than a decision-making body, which is precisely the failure mode that defines stall.

The second is decision rights ambiguity. The committee is stood up without an explicit charter that articulates what decisions it can make, what decisions require executive sponsor sign-off, and what decisions are simply being reviewed for awareness. In the absence of explicit decision rights, every potentially contentious decision drifts upward to the executive sponsor or sideways to a different forum, leaving the committee with the easy decisions that did not need a committee in the first place. The committee becomes operationally optional, and members lose investment in attending.

The third is artifact poverty. The committee operates on slide decks generated for each meeting, with no persistent artifacts that capture the AI portfolio status, the decisions made, and the open questions. Without persistent artifacts, the meetings repeat themselves: the same questions arise in successive sessions, the same decisions are revisited, and the cumulative direction of the work is invisible to participants. The committee becomes an exhausting time sink that participants begin to skip.

The combined effect of these three failures is the nine-to-twelve-month stall pattern that industry research has documented across cross-functional governance contexts. As Harvard Business Review’s coverage of governance stall has emphasized, the failure mode is structural rather than topical: it would manifest in an AI committee, a digital transformation committee, or a data governance committee equally, because the structural design choices that produce stall are the same.

The Design Pattern That Doesn’t Stall

The design pattern that does not stall is recognizable in steering committees that have been operating effectively for two years or more. It has four elements.

Tight membership with named alternates. The committee has eight to ten voting members, each representing a specific function or domain. Each voting member has a named alternate who attends when the primary cannot, with the same authority. The membership is reviewed annually, but expansion requires explicit executive sponsor approval and replacement of an existing seat. This discipline prevents the membership inflation that produces stall.

Explicit decision rights. The charter articulates the categories of decision the committee makes (typically: AI use case prioritization, governance framework decisions, resource allocation across the AI portfolio, escalation paths from operational forums) and the categories of decision that are made elsewhere (typically: individual use case execution decisions, vendor procurement transactions, technical architecture choices). Explicit decision rights eliminate the ambiguity that produces decision drift.

Disciplined cadence. The committee meets monthly, with a fixed 90-minute slot, a standing agenda template, and a strict closing decision summary. Quarterly, the committee runs a longer (3-hour) portfolio review session. The cadence is not optional; members who cannot attend send their named alternate. The discipline of the cadence is what makes the committee operationally serious.

Living artifact set. The committee operates on three to five persistent artifacts that are updated continuously between meetings, not regenerated for each session. The portfolio dashboard, the decision log, the open question register, and the framework documents persist across meetings and provide the continuity that prevents repetition.

These four elements are mutually reinforcing. Tight membership enables real decisions. Explicit decision rights make the decisions defensible. Disciplined cadence makes the work predictable. Living artifacts make the work cumulative. Remove any one of the four, and the design degrades toward the stall pattern.

Membership and Decision Rights

The membership of a pharma AI steering committee should map to the functions whose work is materially affected by AI portfolio decisions. The composition varies by organization, but a typical eight-to-ten-member committee includes:

The executive sponsor’s role is distinct from the chair. The chair runs the committee; the executive sponsor provides the upward path for escalations and the organizational credibility that allows the committee’s decisions to be respected. In some organizations, the executive sponsor attends every meeting; in others, the executive sponsor attends quarterly and is briefed monthly by the chair. Either pattern works, provided the sponsor’s role is explicit.

The decision rights of the committee should be articulated in the charter, with the categories of decision the committee makes versus the categories it does not. A clean articulation looks like this:

The committee makes: AI portfolio prioritization decisions; governance framework decisions including tiering criteria and credibility expectations; cross-functional resource allocation for AI initiatives; framework-level escalations from operational forums.

The committee does not make: Individual use case validation decisions (delegated to use case owners); technical architecture decisions (delegated to IT/Data leadership); vendor procurement transactions (delegated to procurement with committee framework guidance); operational change control decisions for AI components (delegated to function-specific quality forums under committee framework guidance).

The Charter That Holds Up Under Pressure

The committee charter is the foundational artifact. A charter that holds up under pressure has six elements.

Purpose statement. A one-paragraph statement of why the committee exists and what its outputs are. The purpose statement should be specific enough that it could not be substituted into a different committee’s charter without modification.

Scope. An explicit articulation of what is in scope (typically: AI use cases across R&D, clinical, manufacturing, regulatory, and operational functions) and what is out of scope (typically: classical IT systems without AI components, individual research project decisions).

Membership. The voting members and their functional representations, the chair, the executive sponsor, and the rules for membership changes.

Decision rights. The categories of decision the committee makes, the categories made elsewhere, and the escalation path for ambiguous cases.

Cadence. The meeting cadence, the attendance expectations, and the rules for alternates and proxies.

Review. The annual review of the charter, including the trigger for charter modifications and the process for approving them.

The charter should be one to two pages. Charters that run longer typically encode aspirations rather than operating rules, and the additional length is a sign that the committee’s role is not yet clear. The discipline of constraining the charter to two pages forces the design conversation to surface the structural choices that matter.

Cadence and Operating Rhythm

The cadence that works for most pharma AI steering committees is monthly 90-minute sessions, with quarterly 3-hour portfolio reviews. The monthly cadence is fast enough to keep decisions moving and slow enough that members can complete the work between sessions. The 90-minute slot is long enough for substantive discussion and short enough that members do not begin to skip.

The standing agenda template for the monthly meeting:

• Minutes 0-10: Review of decisions from prior meeting and status of action items
• Minutes 10-30: Portfolio dashboard review (status of major use cases, exception highlights)
• Minutes 30-70: Two to three decision items with discussion and decision
• Minutes 70-85: Escalations from operational forums requiring committee guidance
• Minutes 85-90: Closing decision summary, action items, and confirmation

The discipline of the closing decision summary is critical. In the last five minutes, the chair walks through every decision made in the meeting, every action item assigned with owner and due date, and every open question carried to the next meeting. The chair confirms each item verbally with the responsible member. This discipline prevents the recurring pattern of “we thought we decided X, but the next meeting reveals we didn’t really decide anything.” The summary becomes the basis for the decision log artifact.

The quarterly 3-hour portfolio review session is structured differently. It opens with a comprehensive portfolio dashboard review (one hour), proceeds to a focused review of the framework documents and any proposed modifications (one hour), and closes with strategic conversation about the next quarter’s priorities (one hour). The quarterly session is where the committee zooms out from the monthly operational decisions to confirm the direction of travel.

The Artifacts That Power the Committee

The committee operates on a small set of persistent artifacts that are maintained continuously between meetings. Five artifacts cover the typical pharma AI steering committee.

Portfolio dashboard. A single document (typically maintained in a shared collaborative tool) that shows the status of every AI use case in the portfolio. Standard columns include: use case name, owner, tier, current phase, credibility evidence status, next milestone, blockers. The dashboard is updated continuously and reviewed at every meeting. It is the single most important committee artifact.

Decision log. A running log of every decision the committee has made, with date, decision, owner, and rationale. The decision log is consulted whenever a question arises about whether a particular decision has been made and what it was. It prevents the recurrent revisiting of settled questions.

Open question register. A list of questions the committee has flagged for resolution but has not yet decided. Each question has an owner, a target resolution date, and a brief description of the substance. The register prevents questions from disappearing between meetings.

Framework documents. The current versions of the governance framework, tiering criteria, credibility expectations, and similar foundational documents. These are versioned and changed only through committee decisions.

Meeting minutes. Brief minutes (one page) for each meeting, capturing attendance, agenda topics, decisions, and action items. Minutes are circulated within 24 hours of each meeting and confirmed at the start of the next.

The artifacts work together. The dashboard shows current state; the decision log shows what has been decided; the open question register shows what is pending; the framework documents articulate the operating principles; the minutes provide the meeting-by-meeting record. As McKinsey Quarterly coverage of effective governance has emphasized, the discipline of maintaining persistent artifacts is what distinguishes governance bodies that produce decisions from governance bodies that produce meetings.

Anti-Patterns and How to Avoid Them

Six anti-patterns recur across stalled committees. Quality leaders and AI leaders setting up or reviving a committee should design against these explicitly.

The everyone-is-a-member anti-pattern. The committee accepts every function that requests representation, growing to twenty or thirty members. The fix is to define a tight membership in the charter and require executive sponsor approval for expansion. Functions that want visibility without decision rights should be served through other mechanisms (newsletters, quarterly forums) rather than through committee membership.

The status-update-only anti-pattern. The committee meetings become rounds of status updates with no decisions made. The fix is the standing agenda template that allocates time explicitly for decisions, plus the closing decision summary that requires every meeting to produce confirmable decisions.

The slide-deck-treadmill anti-pattern. Each meeting generates fresh slide decks that no one reads outside the meeting. The fix is the living artifact set that persists between meetings and is updated continuously rather than regenerated.

The ambiguous-decision-rights anti-pattern. The committee never quite has the authority to decide anything important. The fix is the explicit decision rights articulation in the charter, with the executive sponsor’s authority backing the committee’s decisions within scope.

The chair-without-mandate anti-pattern. The chair runs meetings but does not have the authority to drive the committee’s work between meetings. The fix is an explicit chair designation with named authority for decision summary, agenda setting, and artifact maintenance.

The executive-sponsor-as-spectator anti-pattern. The executive sponsor is named but does not engage substantively. The fix is an explicit role for the executive sponsor including quarterly attendance, monthly briefing, and named role in escalations.

The design pattern articulated in this article is not novel. The elements are well-established in the broader literature on cross-functional governance. What is specific to pharma AI is the calibration: the membership maps to pharma functions, the decision rights respect the regulatory environment, the artifacts connect to credibility evidence, and the executive sponsor is positioned to engage with regulators and the board on the committee’s behalf. The pattern works when it is applied with discipline; it does not work when elements are skipped or compromised. Quality leaders and AI leaders setting up new committees, or reviving stalled committees, should treat the four elements (tight membership, explicit decision rights, disciplined cadence, living artifacts) as non-negotiable. The committee’s effectiveness over multi-year time horizons depends on the structural integrity of these elements being preserved through leadership changes, organizational restructurings, and the inevitable temptation to relax the discipline when other priorities compete for attention.

References & Sources