II · The method

Reflective practice

Every serious professional knows that each intervention changes the next. What distinguishes a method from a collection of isolated projects is that it makes that learning explicit: what repeats, what fails, which heuristic deserves to move up a category. This chapter is the bridge’s meta-method —not philosophical ornament, but the circuit by which practice corrects doctrine.

A method that does not learn from its own interventions is a static manual. In consulting that is expensive: the same mistakes recur with different clients and doctrine drifts from the field. Reflective practice avoids that drift: it names what happened in a project with precision and decides whether it deserves to move from private anecdote to heuristic, from heuristic to core concept.

Log and field memory

Granular tracking lives in the repository’s internal log —which node was expanded, what descended to the book or site, what remained pending. That log is operational, but in a public professional key it feeds three things:

Recurring failure patterns, with observable signal and method response.
Decision heuristics refined in the field, to abort, replan or scale.
Instrument adjustments —IMIA scoring, gates— when several diagnoses show an indicator lies or a signal is missing.

The log does not replace statistical validation of the instrument; it complements it. Until there is a sufficient sample for empirical recalibration, it is where accumulated judgment lives that orients the next diagnosis.

Failure patterns

Typical ways projects drift —useful as an early-warning checklist:

Pilot for innovators, death in the early majority. The pilot thrills those who tolerate friction; when scaling, no one uses it. Built for the tip of Rogers’ curve, not for those who need it to work without heroism. Response: design for the early majority from the fourth link.
Automate before cleaning the data. The dashboard contradicts operators; AI scales the lie. Method rule: go back from the third link to the first before modeling or deploying.
Sponsorship without operations. Leadership buys; the team passively sabotages. Usually signals an incomplete actor map in the diagnosis.
Post-hoc governance. Agentic AI is deployed; rules are written after the incident. The IMIA D7/D6 ratio would have flagged alert earlier.
Handoff without judgment. The tool is transferred, not the reasoning —guaranteed sociotechnical debt. Response: transfer of judgment and hygiene designed from day one.
Unclear roles after automation. Duplication: some on the tool, some by hand. Missing sociotechnical RACI before scaling (Amanollahnejad et al., 2026).
Champion dependency. Progress dies with succession of the sole referent —episodic leadership.
Scale misfit. Enterprise template or large-firm case worsens operations.
Training with talent leakage. Seasonal turnover destroys upskilling —design for low friction.
Symbolic hype paralysis. Board blocked by narrative, not evidence —micro-pilot with agreed metric.
Strategy on PowerPoint, culture on the floor. The deck promises innovation; operations punish error —culture-strategy misalignment (see Own concepts). Response: incentive map; link 1 down to tacit assumptions (Schein).

Decision heuristics

Operational rules —a construct of practice, not a theorem— I use to abort or replan:

If the area leader does not take part in the demo, the pilot is not ready to scale.
I never automate a process whose owner cannot explain why it is done that way.
If the IMIA agentic dimension is high and governance low, it is not opportunity: it is risk —the D7/D6 ratio is a warning signal, not a vanity score.
If resistance brings a valid operational reason, it is a requirement, not an obstacle.
If three people describe the same process inconsistently, there is no process: sensemaking is pending before modeling.
If digital momentum dies when one person leaves, there is champion dependency —do not scale without operational backup.
If the critical flow depends on manual export / re-keying, measure bricolage before agentic AI.
Which number would you use for an irreversible decision? If no one agrees, there is epistemic uncertainty —not a modeling project yet.
If strategy declares “innovate with AI” but punishing error is the enacted norm, trigger pattern 11.

Back to the corpus

Every heuristic that repeats across three distinct interventions deserves to move up to the core or to Concepts of our own. Every failure pattern can become a piece or a talk. Reflective practice closes the loop between field and doctrine —the same movement The four links prescribes within a project, but at career scale. Without that circuit, the corpus becomes a museum; with it, it remains a living book.