Redeployment Is Not a Plan: The Subtraction Nobody Performs
Redeployment is the standard answer to what happens after the workflow ships, and it is rarely checked against the volume. A system that absorbs the…
Every automation engagement produces the same question from the team whose work is being automated, and it is usually answered by someone who will not be in the room when the answer lands. The standard answer is redeployment. Nobody loses a job, the volume goes to the system, the judgment stays with the people. That answer is humane. It is almost never checked against the arithmetic. Redeployment offered without a number is a reassurance. A reassurance is not a plan, and the specification depends on the difference. Start with the subtraction, because it is rarely performed. A confidence threshold sorts. It does not sample. The cases that clear on their own are the easy ones, so the residual is selected on exactly the reasons work takes time, and a system that absorbs the routine cases leaves a residual smaller in volume and harder per case. The consequence nobody carries forward from that is a headcount consequence. Fewer cases, each demanding more judgment, adds up to fewer roles at higher seniority. That is not a forecast about the labor market. It is subtraction. To test redeployment, separate observed workload from assumptions. Measure routine volume, exception volume, and actual handling effort. Do not infer staffing needs from an unverified auto-resolution claim. Exception cases can take more judgment than routine cases did. Measure that effort rather than sizing the new queue from the old average. The people with calendar space are not necessarily the people qualified to review it. The first condition is growing demand. Look for rationed intake, an unserved backlog, or work deferred because capacity is unavailable. Growth must be evidenced, not forecast to justify a staffing promise. The second condition is work spread across several roles. Replacing routine meeting preparation and routing can return capacity without removing a whole job. Map the responsibilities that remain before calling that redeployment. The specification has to come from the people who run the process now, which makes the arithmetic an operational problem and not only an ethical one. The parts worth capturing are the parts nobody wrote down: the improvised check before a file goes out, the case routed by instinct to one reviewer, the rule with an unstated exception everyone honors. A team that correctly suspects the system will reduce its headcount describes its policy accurately and its own behavior inaccurately. We have put that gap down to judgment that was never said out loud. Judgment is one cause. Incentive is the larger one. Handing over undocumented judgment is difficult to distinguish from writing the specification for your own replacement. The failure is quiet and late. The specification arrives on schedule, the fields are complete, and a developer who attended no meetings could build from it. The build proceeds. Then in week eleven the escalations arrive in a shape nobody described, and they are not edge cases. They are ordinary cases the documented process never covered, arriving at the volume that proves they were always ordinary. Each one is a rule somebody knew and did not say. The cost is a rebuild against a fixed date, and the team you need for it is the team you spent eleven weeks unsettling. The first is the exception and threshold work, and it is genuine. Somebody has to adjudicate what the system escalates, notice when two escalations that look unrelated are the same missing rule, and decide whether a confidently wrong case is a defect or a policy that changed without telling the system. That work absorbs fewer people than it displaces, and it selects for the strongest judgment, not the most available capacity. The fastest reviewer is not the best adjudicator. Name the number of seats it holds. Offering it for six when it holds two is worse than not offering it at all. The second is the next build. The promise is real when that build is funded, named, and scheduled, and a stalling tactic when it is none of those. Say which workflow, which quarter, and who owns it. The third is reduction, a legitimate business decision and the only one of the three that has to be said by a person who can be held to it. Said in week one it costs a difficult meeting and some attrition, a real cost and a bounded one. Left unsaid it costs the specification, because a team told nothing does not conclude nothing. It concludes the worst. Cooperative in the meeting, protective in the detail. Timing is what makes any of the three work. The answer has to be settled and stated before the specification interviews start, because the quality of what those interviews produce depends on it. An operator who knows the plan can afford to be precise about the exception they invented three years ago and never mentioned. An operator who does not know the plan cannot afford it, and no amount of rapport in the room changes that calculation. A team told the headcount will fall loses its strongest people first, because the people with the best outside options move fastest, and they are frequently the ones carrying the judgment the specification needs. Say it anyway. Ambiguity holds their presence and not their cooperation, and a guarded expert in the room for sixteen weeks produces a thinner document than a candid one who leaves in week six. Settled means specific. Which roles change, how many seats the residual work holds, what happens to the difference, and by when. It goes in writing, from the person with authority to make it true, before anyone walks a stranger through the last ten cases they handled. A promise revised in week nine costs more than the same promise made plainly in week one, and the reason is not sentiment. It is structure. The revision reprices every answer already given. In week one this is a decision. In week eleven it is a discovery. Engagements run 8 to 16 weeks and end with the system operating inside the business, which means the client team lives with the arithmetic as well as the system after we leave. We settle this question before the specification interviews rather than after them, in writing, from the person who can make it true.