The retirement date is not the handover plan. In a plant with thin documentation, the dangerous knowledge is usually buried in small judgments: a sound that matters only after a warm restart, a trend shape that deserves a second check, or a reason an experienced technician refuses to take a familiar shortcut. A successor cannot inherit those decisions through a farewell meeting.
A workable two-year expert handover is a controlled transfer of evidence and competence. IAEA knowledge-management guidance It starts with dependency and risk, watches the work as it is really done, captures criteria and exceptions, reconciles them with approved procedures, and then tests whether another qualified person can explain and perform the task. The stages below are a retirement runway for an expert or small group whose absence could leave a plant guessing; this is not generic knowledge management or an extended shift handover.
HSE describes a safe shift handover as preparation by the outgoing person, an exchange of task-relevant information, and cross-checking by the incoming person. That model belongs inside this longer runway, but it cannot replace the months of observation and competence validation described here. HSE shift handover guidance
Treat observations as evidence, not instructions; an expert’s memory does not outrank a controlled procedure. WizeeMind can connect a knowledge unit to an asset, source, revision, and owner, but it cannot approve a setpoint, release equipment, or certify a worker. Fit these recommendations to the site’s risk assessment, competence system, change process, and applicable law.
Start with dependency, not the retirement date
Weeks 1-2 are for a map, not interviews. List the work that depends on one person’s judgment and ask what would happen if that person were unavailable tomorrow. Include operations, maintenance, laboratory, quality, engineering, commissioning, and contractor interfaces. Seniority is only a clue. A less experienced employee may be the only person who understands a rare product change or a legacy control cabinet.
The IAEA’s knowledge-loss method separates the risk in a position from the risk that the person will leave. It uses those dimensions to prioritize knowledge elicitation or another mitigation action, rather than trying to preserve everything equally. IAEA knowledge management for nuclear R&D organizations For a plant, score consequence of loss, knowledge uniqueness, time until likely departure, and effort to create competent cover. Do not call the score a safety calculation unless the local risk process says it is one.
Use a short dependency record for each task or decision:
- asset, functional location, product, or process boundary;
- operating phase, frequency, and conditions that make the judgment appear;
- cue, decision, consequence of delay, and current competent cover;
- controlled procedures, drawings, histories, permits, or training records that should be checked;
- expert, reviewer, owner, next review date, and the reason the dependency was placed in the queue.
The IAEA Corporate OSART report on ČEZ describes a Nuclear Knowledge Transfer and Retention Program using Knowledge at Risk and Expert at Risk lists. It records approximately 80 identified experts at each nuclear power plant, a five-year “look ahead” retirement projection, and fields for knowledge uniqueness, future applicability, the holder, and risk of loss. The IAEA Corporate OSART report on ČEZ lists 212 identified experts, 121 shared experience reports, and no compliance findings related to lack of qualifications or knowledge loss. IAEA Corporate OSART report on ČEZ, pp. 39-40 These are historical results in an IAEA mission record, not a universal benchmark or proof that another plant will get the same outcome.
Mark when the dependency is exposed. A person may be replaceable during steady production but irreplaceable during startup, an infrequent inspection, a product change, a night callout, or a vendor dispute. Mark those conditions. Also mark what is already documented well enough to exclude. Labeling every task critical creates a large spreadsheet and little attention.
For privacy, do not turn this map into a public ranking of named employees. Use role-based access, a clear purpose, and only the personal detail needed to coordinate the transfer. If the site will analyze worker activity or attach recordings to a person, pause for HR, worker-representative, privacy, and legal review. ICO guidance is UK-specific, but its warning about purpose limitation and data minimization is a useful prompt: collect no more information than the stated purpose requires, and do not reuse it quietly for performance management. ICO guidance on monitoring workers Local requirements may differ.
Observe the work that documents miss
Weeks 3-8 are for observation of permitted work, not a request for the expert to recite a career. Choose a safe task, a simulator, or a historical case. The observer records the sequence, the conditions, the comparisons, the pause points, and the moments where the expert asks for another source. Then ask about the surprise: What did you expect? What changed your mind? Which similar-looking condition would invalidate the interpretation? Where would you stop and use the approved escalation route?
The IAEA lists observation, elicitation interviews, storytelling, concept mapping, and process mapping as ways to recover experience-based knowledge. Fit the method to the work: conversation may explain why a reading matters; observation may reveal a check of maintenance history before touching a control screen. IAEA guide to knowledge management strategies and approaches
NASA’s public resources make a similar separation. They include a knowledge-transfer guide, pause-and-learn reviews, continuity guides for supervisors, incoming and departing employees, and a task-documentation template. These are prompts, not a plant qualification system. NASA Knowledge Management Borrow the structure and keep site authority boundaries.
Run one focused observation each week, with a short debrief while the details are fresh. Use a two-person team when possible: one person follows the work and one asks questions. A checklist can include:
- what the expert saw or heard, without turning the observation into a diagnosis;
- what the expert expected, compared, or ruled out;
- which operating mode, asset identity, and document revision set the boundary;
- what evidence would change the judgment;
- what must be escalated, stopped, or checked by another role.
Do not film by default. If audio, video, photographs, screen capture, or wearable data would materially improve a particular capture, explain the purpose before the session, obtain the approvals and consent that actually apply, limit the frame, set access and retention, and offer a non-recorded route when feasible. The ICO says continuous audio or video monitoring is highly intrusive in most worker settings and recommends considering less intrusive means; that guidance is UK-specific and is not legal advice for Spain, the EU, or another jurisdiction. ICO guidance on monitoring methods
Treat the exchange as designed communication, not a private conversation. The Energy Institute briefing asks teams to make responsibilities, information needs, training, guidance, monitoring, and audit explicit, then check that important information was understood. Energy Institute communications briefing Use separate observer, successor, document-owner, and assessor roles.
Hitachi’s official accounts describe a brazing skills training support system for air-conditioner manufacturing at Daikin’s Shiga Plant. The system uses cameras and sensors to collect time-series data on hand motions, torch angle, wire feeding, material orientation, and temperature; a virtual comparison shows how a trainee’s work differs from a meister’s. Hitachi Review says Daikin identified 18 evaluation criteria and that the system had started operation at Shiga, with plans to roll it out to other plants. Hitachi skill-transfer press release Hitachi Review account
The method is useful, but the partners’ plans are not independent evidence for your plant. Record what the expert observed and under which conditions, not a single sound, gesture, or preference as a plant-wide rule.
Turn judgment into reviewable evidence
Months 3-6 are for converting observation notes into knowledge units and reconciling them with the current SOP, drawing, training material, and asset configuration. A transcript is an archive. A knowledge unit is a bounded record another person can challenge.
Use one unit for one cue, exception, or diagnostic distinction. Include the asset and task, operating phase, condition, expected pattern, reasoning, alternatives, exceptions, evidence, procedure identifier and revision, contributor, reviewer, status, and review date. Keep the expert’s words attributed. Separate observation from interpretation. If the cause was not established, say so.
The IAEA distinguishes explicit knowledge in procedures, drawings, and specifications from experience-based knowledge that can be difficult to reveal. Its guidance treats knowledge capture as material that may feed procedures, job guides, experience reports, and training, but still requires review and integration through the organization’s controlled processes. IAEA nuclear industry operating-organization guidance
Use statuses that make disagreement visible: observed, corroborated, supplementary, candidate change, conflict, historical, and unverified. A unit marked conflict must not be used as an operating instruction. Put it into the site’s management-of-change or document-review route with the responsible owner. Our guide to management-of-change evidence covers the evidence packet around a proposed plant change; this article stops at the handover boundary.
Reconcile each unit with the document owner and a competent peer. Check asset, configuration, operating mode, limits, terminology, revision, effective date, and approval status. Decide whether it matches the SOP, adds context, preserves history, or warrants change. Do not rewrite a surprising account until it sounds compliant; the surprise may show that a document or configuration needs attention.
If the review exposes a competence gap rather than a document gap, route it through the site’s competence system before proposing a procedure change. HSE’s guidance ties competence assessment to the task and its foreseeable conditions. HSE training and competence
This is also where procedure version control matters. A remembered workaround may belong in a controlled lesson, require engineering evaluation, or be unsafe to repeat. Link the unit to the exact revision and keep provenance visible. Our guide to procedure version control for plant AI explains why an assistant should retrieve approved versions and surface drift without silently rewriting them.
Keep technology modest: a structured form, revision-controlled repository, read-only asset link, and review queue are enough to start. WizeeMind can make units findable by asset, task, source, and status; the qualified document owner decides whether anything enters the controlled procedure. This protects against treating the expert as infallible or dismissing useful context because it is not an SOP.
Use teach-back to test competence
Months 7-12 are for teach-back and validation. A successor repeating the expert’s words is weak evidence. Ask the successor to explain the cue, the conditions that make it relevant, the plausible alternatives, the uncertainty, the controlling procedure, and the stop or escalation point. Then use a permitted demonstration, simulator, tabletop, annotated trend, photograph, or historical scenario.
Change one condition in the scenario. Move from startup to steady state. Remove the corroborating trend. Present a similar sound on a different asset. The successor should be able to say when the captured judgment no longer applies and locate the approved next step. Do not ask the exercise to produce a new operating threshold. Its purpose is to test bounded use of the captured knowledge.
HSE competence guidance says training alone is not enough. Competence combines skills, knowledge, and experience, and assessment should match the task, hazard, and foreseeable conditions, including infrequent work and upsets. It also calls for assessment to be validated, evaluated, and recorded. HSE training and competence The assessor and standard must come from the site’s competence system, especially for safety-critical work.
Build a small evidence pack for each priority unit:
- teach-back notes showing what the successor understood and where they were uncertain;
- the scenario, task, or demonstration used, including the relevant document revision;
- assessor, date, criteria, supervision level, outcome, and follow-up;
- failed, ambiguous, or disputed points linked back to the unit;
- a reassessment trigger for a configuration change, long absence, or procedure revision.
If coaching remains necessary, set an assessment point in a structured mentor relationship. IAEA mentoring and coaching for knowledge management
Do not treat one pass as permanent competence. It means the person met criteria in one setting; it does not authorize work beyond qualification, remove supervision, or make a rare event routine. HSE says assessment must fit the task and foreseeable conditions, so define reassessment and supervision rather than treating a pass as a license. HSE training and competence If the unit cannot be tested safely, mark the gap and use simulation, peer review, or controlled training.
The incoming person should also be able to retrieve the source without the expert’s help. That is a practical check on the record itself. A clear explanation that cannot be found again is still a dependency.
Build redundancy before you need it
Months 13-18 are for coverage, not a ceremonial shadowing period. Assign at least one second competent person to every high-risk capability, then test the coverage across shifts, holidays, callouts, and contractor boundaries. A name on a matrix is not coverage if that person has never performed or explained the task.
Use controlled rotation where the site’s competence system permits it. Let the successor observe, explain, perform under supervision, perform with peer challenge, and eventually support another learner. The expert should deliberately hand over questions as well as answers. A short rotation through maintenance history, control-room context, laboratory evidence, or quality disposition can expose the interfaces that a single job description hides.
The manufacturer’s own description of its expert-teaching system shows a similar institutional move. Experienced technicians pass on quality-relevant skills such as brazing to younger technicians, and the company uses designated experts to train people beyond the Sakai plant. Manufacturer skill-transfer initiative It is a company example, not a target for every plant. The transferable idea is to make teaching a named responsibility with a route to more than one learner.
The IAEA’s mentoring and coaching guidance describes structured relationships that continue until a newcomer can apply a process correctly without direct supervision, with evaluation of the knowledge and skills transferred. IAEA mentoring and coaching for knowledge management Use that as a design principle, then let the local competence standard set the bar.
Review the map monthly. Show each high-risk unit, owner, second competent person, last teach-back, next reassessment, and unresolved conflict. Add a third contact where capability is rare or the plant operates continuously. Three names do not eliminate risk: coverage exists only when all can find, explain, and perform the work within authority.
Measure leading signals rather than pretending they predict safety outcomes: current owners per priority unit, completed teach-backs, successful independent retrieval, overdue reviews, unresolved conflicts, and coverage by shift. Keep the reasons for a failed or delayed transfer. A low count can mean the queue was small, or that the plant stopped recording.
Simulate the departure and own the gaps
Months 19-24 are the departure test. Run a tabletop or controlled exercise with the expert unavailable: first on the planned date, then early, with no questions answered. Do not disrupt production or ask an unqualified person to improvise. Use a simulator, historical case, or approved low-risk task; keep existing procedures and escalation routes in force.
Give it an owner, observer, stop condition, and evidence pack. Ask the team to locate the unit, controlling document, cue and limits, qualified assessor, and response to conflicting evidence. Record source-search time, calls, unanswered questions, stale links, missing approvals, and reliance on memory.
Use HSE’s preparation, exchange, and cross-checking model for this responsibility transfer. HSE shift handover guidance Energy Institute communications briefing
The exit packet should name four owners: the knowledge-unit owner, the controlled-document or change owner, the competence assessor, and the contingency owner. Confirm access removal and transfer, open work orders, vendor contacts, training records, permits, and review dates through the site’s normal offboarding controls. Do not copy personal files or private messages into the plant repository just because the expert is leaving.
If the expert leaves early, triage rather than pretending to finish the original schedule. Freeze low-risk capture. Protect the top dependency records, schedule an independent review, pair the remaining learner with a qualified peer or approved external support, and mark every unvalidated unit as a gap. If no competent cover exists, use the site’s existing safe operating, maintenance, and escalation rules. A knowledge plan cannot create authority that the plant does not have.
The final decision should be explicit: ready for planned departure, ready with named restrictions, or not ready with a contingency and owner. A red status is useful only when it is visible. A visible gap can be staffed, escalated, or included in a change decision. An anonymous gap waits until a person is standing in front of an unfamiliar condition.
This plan does not replace qualification, controlled SOPs, permits, management of change, record-retention policy, or legal advice on employment and privacy. ČEZ, the Japanese manufacturer, Hitachi, NASA, IAEA, HSE, and ICO evidence comes from particular organizations or jurisdictions. Use their methods as design prompts, not guarantees.
Start with one expert-dependent task. Create its dependency record and schedule the first observation with the expert, successor, document owner, and privacy or worker-representative contact. If the team cannot agree what the task, approved owner, or competent coverage would be, that disagreement is the first finding.
Frequently asked questions
These questions are about planning, not authorization. HSE says training alone does not establish competence, and the IAEA’s risk method is designed to prioritize knowledge-loss work. Apply those ideas through the site’s own assessors, document owners, worker representatives, and legal or privacy review. Local requirements remain controlling in every case. HSE training and competence IAEA knowledge management for nuclear R&D organizations
What should happen first in a retirement handover?
Start with a dependency and risk map for work that would be difficult to perform, explain, or verify if the expert became unavailable. Do not begin by asking for a complete autobiography or by ranking people by age.
Can this plan work if retirement is less than two years away?
Yes, but compress the sequence around the highest-risk work and add an early-exit contingency instead of pretending every task can receive equal coverage. A shorter runway changes the queue and the supervision plan; it does not justify skipping validation.
Should we record an expert on video or audio?
Usually begin with written observation and consent-based capture; record only what the purpose requires and confirm local privacy, labour, and site rules first. Audio, video, photographs, and screen capture can expose more personal or operational information than the team needs.
How do we know the successor is competent?
Use teach-back, a permitted demonstration or scenario, and an assessor-recorded decision against the task’s existing competence standard. Attendance, a signed interview, or a fluent explanation without performance evidence is not enough for safety-critical work.
Can an AI assistant replace the retiring expert or approve a procedure?
No. An assistant can connect evidence and expose gaps, but a qualified person and the site change process still own competence, approval, and operating decisions. Keep the source, revision, uncertainty, and escalation boundary visible.
Sources
- IAEA: Guide to Knowledge Management Strategies and Approaches in Nuclear Energy Organizations and Facilities
- IAEA: Mentoring and Coaching for Knowledge Management in Nuclear Organizations
- IAEA: Knowledge Management for Nuclear Research and Development Organizations
- IAEA: Knowledge Management for Nuclear Industry Operating Organizations
- IAEA: Corporate OSART report on ČEZ
- Hitachi: Daikin and Hitachi skill-transfer press release
- Hitachi Review: Daikin brazing skill-transfer project
- Manufacturer: Skill transfer initiative
- NASA: Knowledge Management
- HSE: Training and competence
- HSE: Shift handover
- Energy Institute: Human factors briefing note no. 10 - Communications
- ICO: Data protection and monitoring workers