When organisations set out to automate, the first candidates nominated are usually the tasks people complain about most: data entry, report preparation, repetitive approvals. These are legitimate targets. They are rarely the ones that pay back fastest.

Look at the handoffs, not the tasks

The expensive inefficiency in most operations is not a person doing a repetitive job well. It is the gap between two systems that do not talk to each other, bridged by a human who exports a file, reformats it, and imports it somewhere else. That work is invisible in process documentation because it belongs to no function, and it is where errors originate.

A practical way to find these gaps: ask where reconciliation happens. Every reconciliation step is evidence that the same information is being maintained in two places.

Quantify the error cost, not just the time saved

Automation business cases are commonly built on hours recovered. That understates the return, because it ignores the cost of the mistakes manual handling produces — an incorrect invoice, a stock figure that triggers the wrong purchase order, a customer commitment made against stale data.

  • Time recovered: straightforward to measure, usually the smaller number.
  • Error remediation: the effort spent detecting and correcting mistakes downstream.
  • Decision quality: the cost of choices made on figures that were hours or days out of date.

In several engagements the third category proved the largest, and it is the one that becomes visible only once accurate data is available in real time.

Sequence by dependency, not by enthusiasm

Automating a process that feeds unreliable data into another manual process moves the bottleneck rather than removing it. Map the dependency chain first and automate upstream, so each step delivers clean inputs to the next.

This also protects the programme politically. Early automation that visibly improves a downstream team's working day generates internal advocates; early automation that creates new exception handling for someone else generates resistance.

Keep a human in the loop where judgement is required

Not every step should be automated end to end. Where a decision involves commercial judgement, exception handling, or reputational risk, the goal is to remove the preparation work and present the decision cleanly — not to remove the decision-maker. Systems designed this way are adopted; systems that automate away accountability tend to be worked around.

The organisations that get the most from automation treat it as an operating discipline rather than a project. They keep a live inventory of manual handoffs, revisit it quarterly, and retire the next one.