IT Myths: Automation Fixes Bad Processes
The myth
“Let’s automate it — that’ll sort things out.” Automation promises speed, fewer errors and less human effort. So if a process isn’t running smoothly, surely automating it will help… right?
The reality
Automation doesn’t fix broken processes — it amplifies them. If your service requests bounce between teams, if change approvals rely on guesswork, or if handoffs are inconsistent, automation won’t resolve that. It’ll make the confusion happen faster and more consistently. You can’t automate clarity, ownership or accountability.
Badly designed automation locks in bad habits. It can make errors invisible, turn workarounds into “how we do things,” and hide operational debt behind a clean dashboard. When inputs are unclear, data is missing or steps vary, automation will repeat those flaws endlessly — until someone stops to ask why.
What gets better with automation (when used well)
- Less admin: meeting notes, status roll-ups, action extraction and reminders.
- Better visibility: trend-spotting across incidents, tickets, OKRs and feedback.
- Faster handovers: consistent briefs, runbooks and checklists.
- Decision support: scenario outlines, risks and options — not the decision itself.
What never becomes automation’s job
- Setting direction and outcomes.
- Making calls when information is incomplete.
- Building trust, giving feedback and growing people.
- Owning results — you cannot outsource accountability to a tool.
The fix
- Standardise before you systemise. Map how work actually flows. Remove ambiguity, duplicate steps and unclear hand-offs before scripting anything.
- Start with outcomes, not tools. Be explicit about the target: speed, accuracy, customer experience, risk. Then choose what to automate.
- Use automation to illuminate. Build in checkpoints and telemetry so you can see exceptions, retries and failure patterns.
- Keep humans in control. Automate routine steps, but keep oversight, judgment and escalation paths firmly human.
- Treat automation as a service. Give it owners, documentation, version control and a CSI cadence — just like any production system.
First steps this week
- Pick one workflow with repeatable pain (e.g., new user setup). Write the current steps, inputs and outputs on a single page.
- Define success (e.g., “from request to usable account in < 2 hours, zero rework”). Remove one ambiguous hand-off.
- Automate one sub-step only (e.g., account creation or access assignment), add logging, and review the data in 7 days.
Metrics that matter
- Lead time per request (before vs after).
- Rework/rollback rate for automated steps.
- Percent of exceptions handled without human escalation.
- Customer satisfaction or SLA adherence for the automated workflow.
Watch outs
- Automating vague or ownerless steps (you’ll just hide the debt).
- Opaque prompts or hidden data use in AI helpers — if it isn’t explainable, it isn’t acceptable.
- Letting tools drive scope — automation should follow process maturity, not lead it.
Bottom line
Automation is transformative when it builds on strong, well-designed processes. When it doesn’t, it’s faster chaos with better branding.
If you want to talk more, I can help. Let’s have a chat.