MELLORCATalk to Mellorca
Patterns

Operations & Process

Why Does This Task Still Need a Person Every Time?

Explores the persistent need for manual, person-in-the-loop handling of routine tasks, the frictions it creates, and how to reframe improvements beyond mere automation hype.

Mellorca Patterns2026-09-08

Observable condition

In many operations, a seemingly simple task still demands a person at every cycle. Someone enters the data, routes the request, or checks that a status has changed before the next step can begin. The operational rhythm assumes that only a human can judge when and how to act, even when the underlying data and rules are stable and the answer is almost always the same.

The result is predictable friction. Work that should be routine becomes a bottleneck because it cannot proceed until a person signs it off. The pattern shows up in checklists that require manual approvals, in handoffs that invite delay, and in dashboards that rely on someone to read and interpret them rather than on signals that trigger the next action.

An illustrative example

The following scenario is illustrative and does not describe a specific organisation. A distributor receives supplier delivery notes by email. Each morning a coordinator opens them, checks the quantities against the purchase order, keys the receipt into the stock system and forwards any discrepancy to purchasing. In most weeks, nearly every note matches. Yet when the coordinator is on leave, receipts queue up, stock appears unavailable when it is physically on the shelf, and sales staff start promising later dates to customers. The task looks small. Its dependency on one person's attention is not.

Warning signs

  • A step is described as "just a quick check", but work visibly waits for it.
  • Throughput drops noticeably when particular people are absent, busy or on holiday.
  • The person doing the task can explain the rule they follow in a sentence or two.
  • Most items pass straight through; only a few genuinely need a decision.
  • Escalations and reminders are a routine part of getting the task done.

Realisation

The real cost of this pattern is not only time. It is the risk of error, fatigue and misalignment between teams. When a task must be touched by a person at every iteration, capacity planning becomes uncertain, because the bottleneck shifts with personnel availability rather than with what the process actually needs.

The organisation also accrues hidden costs: context switching, cognitive load and small delays that compound across the value stream. Each individual touch may take a minute. The waiting before and after it may take hours. In that sense the condition is less about a single task and more about a systemic reliance on manual gatekeeping to keep work moving.

Realisation prompt. If you removed the person from this step for a week, would the work stop because a genuine judgement was missing, or because nobody else was told to move it?

Why the condition persists

Manual steps are often introduced for good reasons: a past error, an audit finding or a new customer requirement. Once in place, they rarely get revisited. The person doing the task becomes skilled at it, the cost is absorbed into "how we work", and nobody owns the question of whether the check still earns its place. Over time, a control designed for exceptions becomes a toll on every transaction.

Identification

The class of issue is operational friction arising from manual intervention in routine workflows. It is identifiable through persistent escalation frequencies, repeated rework and dependence on specific individuals for routine execution. The symptom is not only delay but a creeping loss of scalability, as headcount growth outpaces process redesign.

Diagnosis

The mechanism is a combination of outdated process design, insufficient automation coverage and trust in human judgement over reliable rules. In many cases, teams over-check or over-interpret data, which leads to decision fatigue and inconsistent outcomes. The diagnosis is structural: a process designed around people rather than around robust, observable signals and automated pathways.

It helps to separate three kinds of human touch, because they call for different responses:

  • Transcription: moving information from one place to another. This is rarely judgement at all.
  • Rule application: checking a value against a threshold, a list or a matching record. If the rule can be written down, it can usually be applied consistently by a system.
  • Genuine judgement: weighing incomplete or conflicting information, or making a decision with significant consequences for a person or the business. This is where people add real value.

Most recurring manual tasks are dominated by the first two, with occasional moments of the third. The process, however, treats every instance as if it were the third.

The technology categories reflect the same distinction. Microsoft's description of robotic process automation points to rule-based, non-subjective processes in areas such as finance, procurement and customer service as typical candidates, and distinguishes attended automation, which works alongside a person, from unattended automation that runs on triggers and schedules. IBM's overview of RPA similarly describes it as rule-based software that follows processes defined by people, as distinct from AI approaches that learn from data. Neither category removes the need to understand which part of the task is a rule and which is a judgement.

Commercial impact

The commercial value at stake is capacity, productivity, speed, customer experience and risk. Reducing unnecessary manual touches can deliver faster cycle times, higher throughput and more consistent outcomes, which in turn support customer experience and lower operating risk. It frees capacity for higher-value work, reduces the opportunity for human error and helps teams learn faster, because rule-based steps produce data about what happened.

The cost of inaction is a persistent drag on productivity, and the opportunity cost of not evolving the process. As volume grows, every routine touch scales linearly with it, so growth arrives as additional workload rather than additional margin.

Common misidentification

Teams often assume the task must stay manual because data quality is imperfect or because "human judgement is essential". Both can be true for a minority of cases while being untrue for the majority. Many tasks can be codified into rules or routed through automated decisioning with supervisory oversight, so the human role shifts from gatekeeper to handler of exceptions.

The opposite mistake also occurs: automating a step wholesale without asking where judgement genuinely matters. Where a decision about a person is based solely on automated processing and has a legal or similarly significant effect, the UK GDPR's automated decision-making provisions apply. The ICO's guidance on automated decision-making sets out the safeguards involved, including the ability for people to obtain human intervention and contest decisions. That is a reason to design human oversight deliberately, not a reason to keep every routine step manual.

Possibility

A better state rearchitects the workflow to rely on signals rather than manual checks. Lightweight automation handles routing and validation; clear, accountable escalation paths preserve control without stalling momentum. A thoughtful mix of automation, decision support and human-in-the-loop review can substantially reduce cycle time and error rates.

In that state, routine items flow through without waiting for anyone. Items that fail a rule, fall outside a tolerance or carry higher risk are routed to the right person with the context needed to decide. People spend their attention on the cases that need it, and the business can see how many cases needed it and why.

Intervention

Start by mapping the end-to-end flow around the task, including the waiting before and after it, not just the touch itself. Then work through each decision point:

  1. Write down the rule the person actually applies, in their words.
  2. Sort recent cases into those that followed the rule and those that needed real judgement.
  3. Check whether the data the rule needs is structured, available and trustworthy at the moment the decision is made.
  4. Automate the rule-following path with clear triggers and an audit trail, and route exceptions to a named owner.
  5. Monitor the exception rate and review the rule when the business changes.

Avoid over-automation. Preserve human oversight where it adds real value, and accompany the change with governance that documents ownership, service levels and measurable impact. Automations need owners and failure paths of their own; see why every automation needs a failure path and how to reduce human error without removing human judgement.

Practical diagnostic questions

  • Which steps in the flow require human sign-off, and why?
  • Are the data inputs structured and reliable enough for automated routing?
  • What is the average cycle time with and without manual intervention?
  • What proportion of cases does the person actually change, reject or escalate?
  • What happens to the work when that person is unavailable?
  • What risks emerge if automation handles the routine steps with appropriate controls, and which decisions must remain with a person?

Bottom line

A task that needs a person every time is often a rule that has never been written down, wrapped around a small number of genuine decisions. Replacing one-off manual interventions with rules-based automation and clear escalation paths can unlock capacity, reduce errors and accelerate delivery. The goal is not to remove people but to position them where they add the most value: on the exceptions, and on improving the rules themselves.

Sources and further reading

Related Mellorca reading

Article summaryExplores the persistent need for manual, person-in-the-loop handling of routine tasks, the frictions it creates, and how to reframe improvements beyond mere automation hype.
Explore our solutionsStart a conversation