There is a well-known rule in operations: automate an efficient process and you magnify the efficiency; automate an inefficient one and you magnify the inefficiency. With AI and workflow tools now easier than ever to deploy, that rule matters more, not less.
The temptation to skip ahead
It is easy to see a repetitive task and reach straight for automation. Invoices are keyed in manually — so automate data entry. Approvals take days — so add an automated reminder. But often the deeper question goes unasked: why does this step exist at all, and why does it take so long?
Map how work actually flows
A simple swimlane map — one lane per role or department, boxes for each step, arrows for each handoff — exposes problems that are invisible in day-to-day work. Typical findings include:
- Approvals that add no real control because the approver lacks the information to say no.
- The same data captured in two or three places.
- Work waiting in queues far longer than it takes to process.
- Rework loops caused by incomplete information at the start.
Lean Six Sigma gives this a vocabulary — the eight wastes, value-added versus non-value-added time, root-cause analysis — but the core habit is simply to look carefully before acting.
Remove the step first. Simplify what remains. Then automate.
A three-pass approach
Eliminate. Remove steps that add no value to the customer or to control. These are often the easiest savings.
Simplify and standardise. Combine handoffs, clarify ownership, and write the standard operating procedure for the improved process.
Automate. Now automation targets a stable, well-understood workflow. Rules are clear, exceptions are known, and the result is measurable.
Where AI fits
AI earns its place at points of judgement that follow learnable patterns: classifying incoming requests, extracting fields from documents, drafting first responses, or flagging anomalies for a human to review. It works best inside a well-designed process, with clear rules about when a person must decide.
Measure the difference
Capture a baseline before you change anything — cycle time, error rate, cost per transaction. Without it, even a successful improvement is just an anecdote. With it, you have a business case for the next one.
Want to apply this in your organisation?
Talk to us about where to start — we’ll give you an honest view of the quickest win.
Book a Consultation
BPA Dynamics Insights TeamConsultants, analysts and engineers at BPA Dynamics writing about what works in practice.