You have probably had the meeting by now. Someone demonstrates a tool, the output is genuinely impressive, and the question on the table becomes: what else can we hand over? It is a reasonable question, and the pressure behind it is real — the people asking it are usually the ones being asked, one level up, why the team has not moved faster.
There is an older question that comes first, and it is more useful.
Taiichi Ohno spent three decades at Toyota assembling a way of finding waste in work. He ranked seven kinds — overproduction, waiting, transport, over-processing, inventory, movement, defects — and he put overproduction first. Not because making too much is the most expensive error, but because it is the one that hides the others. A stack of finished work sitting between two steps is evidence that something downstream cannot keep up, and it pays for you not to notice.
That ranking is worth holding against a tool that has just made producing more nearly free.
The direction of the question
Applied here, his approach would not start with which tasks to automate. It starts somewhere less comfortable: which of the things your team already produces did anyone downstream actually ask for?
Most organisations cannot answer that, because the measures point the other way. His objection to local efficiency was blunt — the fastest route to a good departmental number is to keep making things. A team measured on output will always look better producing more, and a tool that makes producing more almost costless is, in that light, not a productivity gain. It is an amplifier pointed at whatever the team was already doing, including the parts nobody reads.
You have probably seen the early shape of this. The weekly report that got longer. The specification that grew a background section. The four options presented where two would have done. None of it is wrong, exactly. It is work being made ahead of a real request, which is where his account says to look first.
Watch out for
The risk is not that the automation fails. It is that it works, and the slack it frees is spent before anyone finds out what the slack was holding.
This is the part of Ohno's system with the worst record, and it is fair to say so plainly. Removing the cushion is the mechanism — take out the buffer and the fault announces itself as a stoppage somebody is forced to fix. But the variability that buffer absorbed does not disappear. It lands on whoever is left at the station. Satoshi Kamata worked six months on a Toyota line in 1972–73 and published an account of exhaustion and regimentation; union researchers in the 1980s called the design management by stress. Ohno's own writing engages with none of it.
The version of that in an office is quiet and familiar. Output per person rises, the headcount plan is adjusted to match, and the capacity that used to absorb a bad week is gone. The sign it is happening is not a drop in output — output looks fine, which is the whole problem. It is that recovery stops. A missed handoff that used to cost an afternoon starts costing a week, because nothing in the system has room in it any more.
If you cut the cushion, cut it deliberately and watch what stops. That is the honest version of the method. Cutting it by accident, as a side effect of a productivity gain you did not plan around, gets you the cost without the finding.
Answer this next
In the last month, what did your team produce that nobody downstream asked for?
You can answer that tonight, from your own week, without a survey. It is also the answer that decides whether more output is worth buying.
