What happens to company knowledge when an expert leaves?
When an expert leaves, undocumented context, exceptions and workarounds walk out with them, and two weeks of handover cannot capture it. A verification loop that records decisions as signed Q&A preserves what to do, when and why, with owners and dates, so continuity survives departures.
What actually goes missing is not only a set of steps but the reasoning that made those steps safe and fast. Thresholds and limits that were never written down, the order of operations that avoids rework, which exceptions are allowed, who approves them, and how to recover from a partial failure. Vendor names and model numbers live in one head, along with escalation paths, service windows, and the fixes that only work on the older line. Mental checklists vanish, so teams keep the visible forms and lose the invisible guardrails. The two week handover becomes hallway conversation and scattered notes, which creates a brittle memory, easy to forget and impossible to audit. After a departure, work continues, but people rediscover pitfalls the hard way and waste time validating what used to be second nature.
A verification loop changes that by turning knowledge into small, owned units. Each unit is a question and an answer that states the trigger, the action, and the rationale, with links to the canonical source. It records the constraints, edge cases, and the normal path, and it names the owner who signed off and the date. When someone asks a new question, it routes to the right expert, and their answer becomes the next verified entry. Old entries age into review, which signals that the knowledge may be indicative, not final, until someone re-verifies. The loop preserves context and accountability, and it does so at the moment of need, which is when the real detail appears. If an expert later leaves, the Q&A stays, and the next person can see not only what to do but why it was chosen and who to contact if the situation changes.
Say a shift supervisor in a bottling plant leaves after a short notice period. They carried a mental map of how to restart after an unplanned stop, which loads to drop first, and which valve to re-seat to avoid a pressure spike. They knew that the line can restart at half speed for exactly one hour without breaching temperature hold times, that the vendor for the older filler answers a phone call faster than a ticket, and that quality will accept a partial lot if a documented rinse was completed. If this lives only in memory, the next stop forces trial and error, longer downtime, and finger pointing. If the same knowledge is a verified Q&A, the on-call lead searches for the trigger, sees the steps, the one hour constraint, the contact and the fallback, and can act. The entry shows who signed off and when, so if equipment changed after that date, the lead knows to confirm before using it.
Duna handles this by presenting curated company Q&A, verified by your own experts, and by refusing to guess. Every answer shows its source and the expert who signed it off, with a date. Confidence is explainable, not a black box, and results appear as one of three outcomes: Verified in green, Indicative in amber when ageing or awaiting re-verification, or Routed when below the threshold so the question goes to the right expert rather than get a wrong or vague reply. Matching is language-agnostic, so an employee can ask in Polish and get matched to a Dutch curated answer without translation. Hosting is in the EU in Germany, or fully self-hosted. Attachments are deliberately not indexed, so search runs on curated answers only, which keeps results precise and auditable. Unanswered questions feed the flywheel: routing creates the next verified entry as a by-product of real work, which builds continuity before, during, and after departures.
Related questions
What should we capture first if a senior expert has already resigned?
Start with high-frequency and high-risk workflows. Write Q&A that state the trigger, the normal path, the known exception, who owns the decision, and the next review date. Include vendor contacts and links to canonical documents, then mark entries as Indicative until a named expert re-verifies them.
How is this different from long standard operating procedures?
SOPs explain end-to-end processes. Verified Q&A capture the specific decisions and edge cases that people ask about in the flow of work, with owner and date. The two live together: the Q&A points to the SOP for full detail, while preserving the why, the limits, and the escalation path.
What happens if no one can verify an answer yet?
Duna will not present it as certain. The entry is marked Indicative, and new questions on the topic are Routed to the best available expert until someone signs off with a date. The system shows its confidence and state, so users know whether to act, to confirm, or to escalate.
Duna keeps company knowledge as verified question-and-answer pairs, so an answer either carries a named expert and a date, or it does not appear at all. Book a 15-minute demo.