Written by: David Carneal – Digital Efficiency Consulting Group (DECG)
Read Time: 3 min
Automation fear usually comes from experience. Someone rolled out a “solution,” it broke, and now the team trusts nothing that has a workflow diagram. Fair.
The cure is boring structure: a decision screen (filters) and safety rails (guardrails). If a task fails a filter, it’s not a forever no. It’s a not yet.
The 5 Filters
Use these to decide whether a task is a good automation candidate right now. The goal is not perfection. The goal is contained risk.
- Volume: does it happen often enough to matter?
- Rules clarity: can the decision be explained plainly?
- Data quality: are inputs consistent enough to trust?
- Exceptions: are unusual cases easy to spot and route?
- Risk: if it goes wrong, is the impact contained?
How to interpret each filter
- Volume: if it’s rare, automate later. Start with high-frequency pain.
- Rules clarity: if the rule changes depending on who’s asked, fix the rule first.
- Data quality: if inputs are messy, automation will fail faster, not smarter.
- Exceptions: if you can’t name them, build an exception queue and learn.
- Risk: if the blast radius is big, require approvals and add a kill switch.
Guardrails That Build Trust
Guardrails keep humans responsible while reducing fatigue. They also prevent the classic failure mode: “It worked in testing” followed by “Why is Finance on fire?”
- Approvals for high-risk actions (money, customer impact, compliance).
- Exception queue with a named owner and response expectations.
- Audit trail: what happened, when, and why.
- Sampling early: spot-check until confidence is earned.
- Kill switch: pause fast if something looks off.
Minimum Guardrails by Risk Level
Not every workflow needs the same safety rails. Match guardrails to risk so you don’t over-engineer and stall out.
When in doubt, start with a low-risk workflow, prove stability, and then move up the risk ladder. Trust grows by repetition, not by a single heroic launch.
- Low risk (internal notifications, reminders): audit trail + kill switch.
- Medium risk (task creation, routing changes): audit trail + sampling + exception owner.
- High risk (money, customer impact, compliance): approvals + audit trail + exception queue + kill switch.
A One-Page Pilot Template
If you can’t fit your pilot plan on one page, you’re probably trying to automate too much at once. Use this format:
- Task (one sentence).
- Trigger and required inputs.
- Routing rule (who owns it).
- Validation checks (what gets rejected or flagged).
- Exception path (who owns the weird cases).
- Guardrails (at least one).
- Metric and review date.
Micro case: the exception queue saves the day
Teams often try to automate exceptions and end up automating confusion. An exception queue flips that. Normal cases flow fast. Weird cases land with a human who has context. The workflow gets better over time because the exceptions become design inputs instead of surprise attacks.
If you want one “trust builder” to start with, pick the exception queue. It tells the team, out loud, that you expect edge cases and you’re planning for them instead of pretending they don’t exist.
CTA: Print the 5 Filters and score your top task. If it’s green, pilot with guardrails. If it’s yellow, tighten intake and ownership first. If it’s red, keep it human and pick a safer win.