Interventions

The four families

Cummings and Worley's standard taxonomy.

1. Human process — relationships, communication, group dynamics.

Process consultation; team building and chartering; third-party conflict resolution; coaching and mentoring; facilitation and dialogue; after action reviews; large-group methods.

Use when the diagnosis says: low trust, unresolved conflict, poor meeting quality, silence in the face of risk, unclear team purpose.

2. Technostructural — structure, workflow, technology, division of labor.

Structural redesign; work design and job enrichment; team boundary design; process improvement (lean, six sigma, value stream mapping, theory of constraints, queueing analysis); sociotechnical redesign; downsizing and reengineering (high risk, frequently misused).

Use when the diagnosis says: handoff delays, unclear decision rights, work requiring constant cross-boundary coordination, a visible bottleneck.

3. Human resource management — attracting, developing, retaining, rewarding.

Goal setting and performance management; reward and recognition redesign; career frameworks; talent development and succession; wellness and workload; diversity, equity, and inclusion.

Use when the diagnosis says: the system rewards the opposite of what it asks for, people cannot see a future here, or capability gaps are real rather than motivational.

4. Strategic — the organization's relationship with its environment.

Mission, vision, values; culture change; merger and acquisition integration; organizational learning; alliances and networks; digital and AI transformation.

Use when the diagnosis says: the environment has changed and the organization has not, or identity and direction are genuinely contested.

Matching intervention to diagnosis

Process consultation

Edgar Schein's stance, and the most under-used skill in the field. The consultant's expertise is in process — how work is done, how people interact, how decisions are made — not in content. The client keeps ownership of the problem.

Three models of helping:

1. Expert — the client defines the problem and buys an answer. Requires three fragile assumptions: the client diagnosed correctly, communicated accurately, and the answer is implementable in their context.
2. Doctor-patient — the client presents symptoms; you diagnose and prescribe. Fails when the client will not reveal the real symptoms, accept the diagnosis, or take the medicine.
3. Process consultation — you help the client diagnose and solve their own problem. Slower to satisfy, more durable, and the only model that builds client capability.

All three are legitimate; the error is defaulting to 1 or 2 when 3 is called for, which happens because 1 and 2 feel like value delivery.

Principles of humble inquiry: always try to be helpful; stay in touch with current reality; access your ignorance — the most useful question is one you genuinely do not know the answer to, and clients can always tell the difference; everything you do is an intervention, so there is no neutral observation; the client owns the problem and the solution; go with the flow; timing is crucial; when in doubt, share the problem.

What to watch while a group works: who speaks, for how long, and after whom; who is interrupted and by whom; how decisions actually get made — consensus, default, silence, or the loudest; what topics change the energy in the room; what is raised and dropped without response; what is discussed in the corridor and not in the room. Feeding these back descriptively, without interpretation, is often the entire intervention.

Health care: a team asks for "communication training." Observe three handovers first, then report what you saw: conducted standing in a corridor, interrupted four times, no structured format, no read-back. That description does more than a course. Software: a team asks for an architecture review; the observable finding is that three architecture decisions have been made and reversed, always after the same person's objection, always outside the meeting. Executive: an executive team asks for a strategy facilitator; the process finding — that the CEO speaks first on every item and discussion converges within ninety seconds — is more valuable than any strategy framework, and much harder to deliver.

See topics/process-consultation.

Large-group methods

Four methods, chosen by what you need.

World café. Tables of four to five, paper tablecloths, three rounds of 20–30 minutes on questions that build, with all but a table host rotating between rounds, then a harvest. Its value is entirely in question design and harvesting; without a named owner for each theme and a stated decision route, do not run it. A standard progression: "what is your experience of X today?" → "what is making X the way it is?" → "what would need to be true for X to be different, and what is one thing you could do?"

Open space. No pre-set agenda; participants create it in the opening session. Four principles (whoever comes are the right people; whenever it starts is the right time; whatever happens is the only thing that could have; when it's over, it's over) and the law of two feet: if you are neither learning nor contributing, go elsewhere. This guarantees every session contains only people who chose to be there. Preconditions: a real question with a real stake, genuine openness to topics leadership did not choose, leadership present as participant rather than chair, and a route to action.

Appreciative inquiry. The 4-D cycle: Discovery (paired appreciative interviews about peak experiences), Dream (what might be), Design (provocative propositions in the present tense, grounded in real examples), and Destiny (self-organized action). Often preceded by Define — choosing the affirmative topic, which is the most consequential decision. Topics are stated as what you want more of, not the problem inverted: not "reduce handover failures" but "handovers that leave the next team fully equipped."

Its serious criticism deserves stating: participants with genuine grievances can experience AI as enforced positivity, and facilitators can use "stay appreciative" to silence dissent. Skilled practice makes room for difficulty inside an appreciative frame.

Future search. Whole system in the room — all stakeholders with Authority, Resources, Expertise, Information, and Need. Five tasks over three days: review the past; explore external trends; own our own actions ("prouds and sorries" — each stakeholder group states publicly what it is not proud of, which does the emotional work that makes day three possible and is what sponsors most want to cut); create ideal future scenarios; identify common ground and plan action. Differences are recorded and explicitly not worked on — groups act only on what they agree about.

Health care: future search suits pathway design across hospital, primary care, social care, ambulance, commissioners, voluntary sector, and patients, because no single organization controls the pathway. World café suits whole-department conversations across professions and grades. Software: internal open space days are frequently the most productive engineering event of the quarter — pair with an explicit commitment session and named owners. Executive: open space is valuable when the executive team genuinely does not know the answer, and destructive when used to manufacture buy-in for a decision already taken. Participants detect the difference immediately and remember it for years.

See topics/world-cafe, topics/open-space-technology, topics/appreciative-inquiry, and topics/future-search.

After action review

The cheapest organizational learning mechanism available, and the one most often degraded into a status report. Four questions:

1. What was supposed to happen?
2. What actually happened?
3. Why were there differences?
4. What will we do differently next time?

Establishing the intent before discussing the outcome is what prevents hindsight from rewriting what people meant to do.

Ground rules: no blame; everyone participates and rank is left outside; facilitated rather than chaired, preferably not by the person who led the event; soon after, while memory is fresh; findings recorded and shared, or the learning stays in the room.

Facilitation guidance: ask the most junior people first, because seniority anchors the account; establish the timeline factually before discussing causes; ask "what made that the reasonable thing to do at the time?" rather than "why did you do that?"; beware hindsight bias, since with the outcome known the correct action looks obvious and it was not; and distinguish counterfactuals ("they should have noticed the alarm") from causes ("the alarm was one of forty active, and eleven were routine"). Counterfactuals feel like findings and are not.

Variants: blameless postmortem (software operations), hot wash (immediate, on-site, minutes), morbidity and mortality conference (clinical), retrospective (agile, on a cadence), and pre-mortem — run before the event: "it is a year from now and this failed completely; write down why." Pre-mortems surface risks people will not raise as objections but will happily raise as fiction.

Why AARs decay: the senior person speaks first; it becomes a defense of decisions already made; actions are recorded and never done; only failures are reviewed, so the process becomes associated with blame; and the same findings recur for years unaddressed — which is itself the most important finding and is never named.

Countermeasure: audit your last ten AARs for repeated findings. Repetition means the mechanism is producing documents, not change.

Health care: a five-minute corridor debrief after a cardiac arrest call is a hot wash and is good practice. A full AAR for a serious incident must be separated in time and personnel from any disciplinary process, and everyone must be told which one they are in. Software: publishing post-incident reviews internally in full — including the embarrassing parts — is the differentiator, because teams learn from incidents they did not experience. Executive: executive teams almost never run AARs on their own decisions. Reviewing a major decision twelve months later against what was predicted at the time is uncomfortable, cheap, and the fastest route to better executive judgment.

See topics/after-action-review.

Dosage and sequencing

• Start where credibility is cheapest to earn. A visible, fast win the sponsor did not have to force.
• One major intervention per group at a time.
• Sequence structural clarity before cultural work when decision rights are genuinely ambiguous. People cannot be candid about a game whose rules are undefined.
• Sequence trust work before structural work when the last three restructures failed. Nobody will engage honestly otherwise.
• Do not intervene in a working team. The urge to improve something that is fine is a real and expensive failure mode.

See topics/interventions-catalog.