The person everybody calls when a machine “doesn’t sound right” may hold knowledge that no shift log can name. They notice a short rattle after a warm restart, know which reading settles late, or ask one question when a batch behaves oddly. If that person changes shift, moves role, or retires, the plant can keep every document and still lose part of its operating memory.
Knowledge transfer between shifts should make expert judgement observable and traceable. Find work that depends on one person’s reasoning. Watch it, ask about the conditions and exceptions, then record small knowledge units against the relevant asset, task, approved document, and source. A successor tests the transfer through explanation and performance. Each unit also needs an owner and review date.
This process supplements controlled procedures, competence assurance, and accountable supervision. It cannot authorize an operating action. The HSE handover model requires preparation, exchange, and incoming cross-checking; the IAEA knowledge management guide separates difficult-to-document tacit knowledge from explicit material such as procedures and drawings. Together, those ideas turn “ask Sam” into a transfer process without letting experience outrank approved instructions.
Find the work that depends on one person’s judgement
Do not begin with the oldest employee list. Begin with operational dependency. Someone with thirty years on site may share well-documented work with several peers. Someone with five may be the only operator who can explain a recurring transition or the technician every crew calls during an infrequent upset. Dependency, not seniority, is the reason to act.
The IAEA knowledge management guide describes two ways to find exposure. One maps competencies against employees to identify important positions and experts at risk. The other combines attrition likelihood with the importance of the knowledge and the effort needed to replace it. This is nuclear-sector guidance, not a universal scoring standard. Its useful test is whether losing the knowledge would leave a consequential task without enough competent coverage.
Build the map from evidence people already produce. Review callouts, troubleshooting escalations, repeat work orders, abnormal-operation debriefs, overtime patterns, training exceptions, and questions deferred until a particular shift returns. Ask supervisors and peers who they rely on when the procedure no longer answers the diagnostic question. The earlier IAEA guide describes experts as people who notice violated expectations, subtle cues, or rare-event lessons that others miss. Treat those descriptions as prompts to investigate, not proof that a nominated person is always right.
Create one dependency record for each exposed task. Name the asset or process, operating phase, cue-bearing decision, competent coverage, consequence if knowledge is unavailable, and evidence that revealed the dependency. Include contractor-held knowledge where the site may lose the capacity to question the work. The result is a capture queue, not a personality ranking.
Map when the dependency appears. Knowledge may be shared during steady production yet collapse during startup, product change, unusual maintenance, or night coverage, when the support network is absent at the exact point competent cover becomes difficult to obtain.
A useful decision rule is: capture first where one person’s absence would make a qualified team slower to recognize a meaningful exception or unable to explain why a familiar judgement was made. Do not capture routine preference merely because it is distinctive. HSE competence guidance defines competence as a combination of skills, experience, and knowledge, and links assurance to responsibilities and tasks identified in risk assessment. The dependency map should follow that risk context.
Watch the task, then ask about the surprise
Experts often omit the most useful parts when asked, “How do you do this?” Familiar work has compressed into recognition. A broad interview tends to produce the official sequence or tips without context. Observation gives the interviewer something concrete to probe: a pause, a comparison, an unexpected trend shape, or a choice not to continue.
Choose a permitted task or safe simulation. Brief everyone on the purpose and respect the site’s normal controls. The observer records what happened without turning the session into an improvised procedure. Afterwards, replay the decisions: What did you expect? What changed your attention? Which cue mattered? When would that interpretation fail? What would make you stop and consult the approved route? Ask for a similar-looking case with a different explanation.
A second observer can help. They may notice that the expert checks a control-room trend before going to the field, or asks maintenance about recent work before interpreting the cue. Sequence often carries the reasoning that an interview misses.
The IAEA knowledge management guide lists task observation, elicitation interviews, storytelling, concept mapping, and process mapping among methods for recovering experience-based knowledge; the tool should fit the knowledge being transferred. Its 2006 IAEA guide case material combines an iterative portfolio, generic interview, specialist interview, and peer review. One clear story is a lead to verify, not organizational truth.
Hypothetical example: during a permitted pump inspection, an experienced operator pauses after a warm restart and compares a brief casing sound with the suction-pressure trend. The capture is not “a rattle means cavitation.” It records the observed sound, the restart phase, the comparison the operator made, alternative explanations they considered, and the approved inspection or escalation source they would consult. The example creates no threshold and authorizes no action.
Use the incoming operator as an active interviewer. Ronny Lardner’s effective shift handover review frames handover as a managed human-factors process: preparation, reliable exchange, and incoming-shift cross-checking belong in the routine. The Energy Institute briefing says feedback should show that a message was received and understood, and reports weaknesses in responsibilities, information needs, training, guidance, monitoring, and audit. Questions expose where an explanation still depends on shared history.
Write knowledge units that can be checked
Long interview transcripts preserve words but make poor handover material. A useful capture is smaller: one cue, exception, or diagnostic distinction with enough context to test it. Call this a knowledge unit. It states what someone noticed, where and when it applies, how the expert interpreted it, what could make that interpretation wrong, and which approved source governs any response.
Give every unit a stable structure:
- asset, system, product, or process boundary;
- task and operating phase;
- observed cue or condition, using measurable detail where available;
- expert reasoning and alternatives considered;
- known exceptions and uncertainty;
- linked procedure, drawing, work order, trend, lesson, or training source;
- contributor, reviewer, capture date, and status.
The IAEA knowledge management guide distinguishes explicit knowledge in procedures, drawings, and specifications from experience-based knowledge that can be hard to reveal or document. Keep that distinction visible. A procedure has controlled authority within its scope. An operator’s cue is an attributed observation. A maintenance history records work; a trend records measurements. One card can connect them without treating them as interchangeable.
Avoid universal statements. “A metallic tick after restart means X” hides conditions and confidence. Prefer: “Contributor noticed a short metallic tick on P-204 during two warm restarts and compared it with trend Y before consulting procedure Z; cause was not established.” The second form can survive challenge. It tells the next person what was observed and where the reasoning came from without inventing an operating rule.
IAEA guide says captured material may feed improved procedures, job guides, and experience reports. Its checklist asks whether content is reviewed, integrated into configuration management or training, and maintained through expert review. Keep the unit under review until the document owner decides what belongs in a controlled artefact.
WizeeMind can connect the unit to the exact asset, procedure revision, source record, and related experience. It preserves attribution and status: communicated, corroborated, refuted, approved in a controlled document, or expired. The assistant does not approve the unit, amend the procedure, or turn a remembered workaround into an instruction. Our guide to industrial data context explains why source, time, asset, and decision relevance travel with a record.
Reconcile experience with approved documentation
Sooner or later, practice and documentation will disagree. Do not smooth over the gap. A cue may add context that a concise procedure does not need. It may reveal a changed configuration, local drift, or a practice that should stop. Only the accountable site process can decide which.
Run reconciliation as a separate review. Put the unit beside the current procedure, drawing, training material, operating-experience record, and asset configuration. Check identifiers, revision, effective date, scope, operating mode, limits, terminology, and approval status. Ask the document owner and a competent peer whether the experience is consistent, supplementary, obsolete, or conflicting. Record the decision and evidence. Never edit the expert account merely to make it look compliant.
IAEA guide calls procedures a logical home for captured knowledge, but warns that explanation should not impair reliable procedure use. It also says procedures, drawings, and training plans should follow configuration changes. HSE competence guidance links structured on-the-job training to risk assessments, controls, and procedures, with stronger assessment for safety-critical tasks. Neither source lets an interview alter an approved method.
Use plain statuses. “Consistent” adds context without changing the approved route. “Candidate improvement” enters the site’s change process. “Conflict” must not be used as guidance until the responsible authority resolves it. “Historical” no longer matches the current asset or process. “Unverified” still needs a source or scope check.
HSE shift handover guidance says handover should use verbal and written communication, match incoming staff needs, and include cross-checking as responsibility transfers. The Energy Institute briefing says information should reach the right person in the right form and time, with feedback confirming understanding. A disputed unit should arrive as a visible dispute, with the controlling source and escalation owner attached.
WizeeMind retrieves and connects approved documentation with traceable experience. It can flag an old drawing or show that two shifts describe one exception differently, then route the discrepancy to an owner. It never approves an instruction, chooses an operating limit, releases equipment, or authorizes an operational action. Our procedure version drift guide covers the document-control side.
Prove the transfer through performance
A successor repeating the expert’s words is weak evidence of transfer. They may remember the label but miss the cue or the moment when the pattern stops applying. Validation should expose the reasoning and match the task’s hazard, frequency, and complexity.
Start with teach-back. Ask the successor to explain what was observed, which context matters, what remains uncertain, and which approved document controls the response. Then use a permitted demonstration, simulator, tabletop scenario, annotated trend, photograph, or historical case. Change one condition. Does the interpretation still hold? What evidence comes next? The assessor records explanation and performance against defined criteria, not attendance.
HSE competence guidance states that training alone is not sufficient for competence; people consolidate knowledge and skill through practice. Assessment should suit the hazard, cover foreseeable conditions such as infrequent work and upsets, and be validated, evaluated, and recorded. The IAEA mentoring and coaching guide describes relationships that continue until a newcomer can apply a process correctly without direct supervision, with evaluation of the knowledge and skills transferred.
Teach-back checks explanation. Demonstration checks performance under approved conditions. A scenario tests a changed context; peer challenge may catch an unsupported leap. None proves permanent competence alone. The site’s competence system decides the assessor, standard, supervision, reassessment interval, and authorization.
Hypothetical validation: the incoming operator reviews a historical warm-restart trend and explains why the captured sound cue was noteworthy only in that phase. A second scenario shows the same sound during a different mode with no matching trend. The operator should reject the shortcut, state the uncertainty, and locate the approved inspection route. Passing does not mean diagnosing the pump correctly from a sound; it means using the captured experience within its boundary.
Record the result against the unit: person assessed, scenario or task, criteria, evidence reviewed, assessor, date, outcome, gaps, and follow-up. Link failed or ambiguous demonstrations back to the unit. They may reveal poor transfer or weak original reasoning.
The incoming shift now tests operational memory instead of merely receiving it. The maintenance troubleshooting evidence guide uses the same discipline: separate symptom, hypothesis, verification, and approved safe-work route. WizeeMind can retrieve the evidence and record the trace. A qualified person remains responsible for assessment and any operational decision.
Keep each unit findable, owned, and temporary
Captured knowledge ages. Assets are modified, procedures are revised, an old cue proves misleading, or a successor finds better evidence. Without review dates, attributed experience gradually becomes anonymous folklore.
Retrieval begins with work metadata: asset and functional location, task, operating mode, product or material, cue, procedure identifier and revision, source type, contributor, validation status, and date. Add relationships, not a pile of keywords. A pump cue should connect to the pump, restart phase, controlled inspection route, supporting trends, related work, and authorized reviewers.
The IAEA knowledge management guide says knowledge portals can connect technical documents, expert directories, operating experience, work management, and configuration management; efficient search is essential. IAEA guide asks whether captured knowledge is retrievable, searchable, maintained through its life cycle, and routinely reviewed. Those are design checks, not proof that a portal makes content correct.
When a search returns several units, rank current, reviewed material above older experience and show why each result matched. Do not hide expired or conflicting records; move them out of the default answer and expose them when a reviewer needs the history behind a decision.
Assign one accountable owner by role, close to the asset, procedure, training programme, or operating discipline. Set the next review from change triggers and risk. Trigger it when a linked document changes, the asset is modified, conflicting evidence appears, validation fails, the contributor leaves, or the scheduled date arrives. If review does not happen, mark the unit expired and remove it from default operational retrieval. Preserve its history.
Measure whether dependency fell. Track successors who demonstrated the bounded knowledge, units reconciled with current sources, unresolved conflicts, and searches that end at one unavailable person. The IAEA knowledge management guide proposes tracking successors who reach a minimum competence level and continuing reliance on individual experts. These measures show transfer progress; they do not prove safe performance.
Start with one asset or recurring transition, not the whole plant. Choose a dependency with a willing expert and an identified successor. Observe one permitted task, build three to five knowledge units, reconcile each against current documents, and run a teach-back plus one scenario. Give every unit an owner and review trigger. The WizeeMind operations workspace can connect approved documents, asset context, and attributed experience so crews can retrieve the trail. Site roles still approve documents, assess competence, and decide what happens in the plant.
Sources
- Health and Safety Executive: Shift handover
- Health and Safety Executive: Training and competence
- IAEA: Guide to Knowledge Management Strategies and Approaches
- IAEA: Knowledge Management for Nuclear Industry Operating Organizations
- IAEA: Mentoring and Coaching for Knowledge Management
- Energy Institute: Human factors briefing note no. 10
- Ronny Lardner: Effective shift handover literature review