Cross training is presented as straightforward insurance against absence and departure. The mechanism is sound, but the cost lands immediately and the benefit arrives later, which is why programs stall.
Two people are occupied producing one output
During training the experienced person explains rather than works, and the learner produces slowly and with errors that need correction. Capacity falls for the duration.
In a small team that loss is concentrated. There is no bench absorbing it, so the visible effect is a backlog appearing in exactly the function being made more resilient.
Managers respond by pausing training whenever volume rises. Since volume usually rises, the program is postponed indefinitely and the coverage gap remains.
Knowledge is mostly undocumented
The formal procedure is the easy part. What actually takes time is the accumulated set of exceptions: which customer is invoiced differently, which supplier needs a phone call, which system field is used unofficially.
That knowledge surfaces only when it is needed, so training by explanation misses it. Training by doing under supervision catches more, at a higher cost in supervision time.
Attempts to document everything first usually produce a long manual that is out of date within months and that nobody reads during an actual handover.
Skills decay without use
Someone trained in a task they then perform once a quarter is not reliable cover. Systems change, procedures change, and confidence fades faster than knowledge.
This is why rotation matters more than training events. Periodically performing the backup role, including during normal operations, keeps the capability alive.
Rotation carries its own cost in reduced specialization. The firm accepts slightly lower efficiency in exchange for continuity, and that trade should be made deliberately.
Coverage changes bargaining positions
An employee who is the only person able to perform a critical task holds informal leverage. Cross training reduces it, and some people resist for that reason.
The resistance is rarely stated openly. It shows up as unavailability, as training that never quite completes, and as documentation that omits the parts that matter.
Making knowledge sharing an explicit expectation, recognized in how performance is assessed, addresses this more effectively than mandating attendance at sessions.
The payoff is measured in avoided disruption
The return does not appear in productivity figures. It appears the week someone resigns, is hospitalized or takes leave, and the function continues without an emergency.
Because avoided crises are invisible, the investment is hard to defend against short-term output targets. Framing it as a known operational risk with a known cost makes the discussion concrete.
The practical test is simple: for each critical task, name the second person who has performed it recently. Tasks with no answer are the ones the program should address first.