METHOD · 5 MIN READ
Why AI automates the waste unless you change the workflow first
Bolting agents onto a broken process makes the broken process faster. Redesign the flow, then automate what is left.
LAUNCH AGENTS ·
The faster horse problem
When teams first deploy AI, they usually point it at the most visible manual task. Someone spends hours copying data from emails into the ERP, so an agent does it instead. The task disappears, but the process around it does not change. The data still waits for a three-step approval, still gets checked twice, and still sits in a queue at month-end.
The result is a broken process that runs faster. Costs drop slightly, cycle time barely moves, and leadership concludes that AI is overhyped.
Waste hides in the flow, not the task
Most delay in knowledge work is waiting, not working. A request may take twenty minutes of actual effort and ten days of elapsed time. The ten days are spent in inboxes, in approval queues and in handoffs between teams that use different systems. Automating the twenty minutes saves twenty minutes.
Common sources of waste include:
- Approvals that exist because of one incident years ago and are now applied to everything.
- Re-keying the same data into several systems because they were never connected.
- Expert reviewers checking routine cases that follow clear rules.
- Batching work weekly when it could flow daily.
Change the workflow first
Before building anything, map how the work moves today and ask of every step: does this need to happen, does it need to happen here, and does it need this person? Remove what is not needed. Route routine cases away from experts. Replace batches with flow where you can.
Only then design the agents. In the new workflow, agents do the reading, matching, drafting and routing. People handle exceptions and decisions. Because the waste has already gone, the agent is accelerating work that actually matters.
Why workflow change alone is not enough either
The opposite mistake is a process redesign with no technology behind it. The new way of working depends on discipline, and when attention moves on, people drift back to old habits. Agents make the new workflow the easy path, because the system does the routine work by default.
Workflow change and technology need each other. One without the other either fades or automates the waste. Together they move the number.
Key takeaways
- Most delay in knowledge work is waiting between steps, not the steps themselves.
- Automating a task inside an unchanged workflow saves the task time, not the cycle time.
- Remove unnecessary approvals, re-keying and batching before designing agents.
- Agents make the redesigned workflow stick by doing the routine work by default.
Frequently asked questions
What does it mean to automate the waste?
It means applying AI or automation to steps that should not exist, such as duplicate data entry or unnecessary approvals, so they run faster instead of being removed.
Should we redesign processes before using AI?
Yes. Map the workflow, remove steps that add no value, and route routine work away from experts first. Then deploy agents on the redesigned flow.