An agency automation audit is a scheduled review of every workflow the team depends on: what triggers it, which accounts it touches, who owns it, what happens when it fails and whether it still earns its cost. The goal is to find silent failures before a client does.
This guide is documentation-based. It describes how to run the audit and what to record; it does not report an audit we performed or a measured saving. Adapt the checks to your tools, client agreements and security procedures, and test failures with fictional data in a controlled environment.
Build an inventory first
List every live workflow, including native automations inside a CRM or project tool. For each, record the trigger, connected accounts, data transferred, owner, expected result, frequency and last successful run. Mark workflows that affect a client deadline or send client-facing messages.
| Workflow | Trigger | Owner | Client data | Last checked | Recovery action |
|---|---|---|---|---|---|
| New client intake | Form submitted | [Name] | Contact details | [Date] | Manual project task |
| Report delivery | Schedule | [Name] | Campaign metrics | [Date] | Export and send manually |
| Content approval notice | Status change | [Name] | Draft links | [Date] | Notify approver in the project tool |
An inventory that lives only in one person's head is the first finding of the audit.
Copy the audit checklist
- Every live workflow has an owner and a backup.
- Connected accounts belong to the right organization and have suitable permissions.
- Triggers still fire on representative test data.
- Duplicate events do not create duplicate clients or invoices.
- Invalid or missing fields produce a visible error.
- Alerts reach a person who can act, not an unused inbox.
- Failed runs can be retried safely without duplicate actions.
- Client-facing output has a review step where mistakes would be costly.
- Usage and billing have been checked against recent activity.
- Credentials and tokens are reviewed when staff leave.
- A manual fallback exists for critical workflows.
- Obsolete workflows are disabled and documented.
Test failures deliberately
Use fictional records. Submit the same form twice, omit a required field and disconnect a test integration if you can do so safely. Confirm where the error appears and who receives it.
Do not run destructive tests in a live client account without a controlled test environment. If the only way to test is to risk real data, mark the workflow as untested and schedule a safer test, rather than assuming it works.
Prioritise by impact, not by age
Fix workflows that can lose client requests, send wrong information or alter billing first. Then address silent failures and expensive manual recovery. Unused scenarios and oversized schedules come last.
| Easy to fix | Hard to fix | |
|---|---|---|
| High impact | Fix this week | Assign an owner and a date |
| Low impact | Batch with routine work | Disable or document |
High impact means a client would notice, a deadline would move or money would be wrong. Low impact means a small internal annoyance. The matrix is a sorting tool, not a score to defend.
Keep a recovery record
When a workflow fails, the next person needs to know what was done, not just that it was fixed.
| Field | What to record |
|---|---|
| Workflow and run | Name and execution reference |
| Symptom | What the team saw and when |
| Affected records | Clients, projects or invoices touched |
| Immediate action | Re-run, manual fix or rollback |
| Root cause | Trigger, mapping, permission or platform change |
| Prevention | Check, alert or design change added |
| Owner and date | Who closed it and when |
The record turns one incident into a check the next audit can verify. Without it, the same failure returns with a new name.
Review usage and cost
Compare recent billing periods with actual usage, not a forecast alone. Look for schedules that run more often than the task requires, scenarios that process the same records repeatedly and tiers whose limits are far above real volume.
Vendor pricing changes, so check the current plan and unit on the vendor's own page rather than a screenshot in a proposal. The Make pricing guide for agency workflows explains how operations and credit limits are counted, and the n8n vs Make comparison covers the trade-off for small teams.
Schedule the next review
Record a review date and make ownership part of staff onboarding and departure. After changing a workflow, update its inventory row and retest its failure path.
A quarterly review is usually enough for stable workflows; review sooner after a platform migration, a staff change or a client incident. Put the date in the same place as the workflow list so it cannot be forgotten.
Frequently asked questions
What is an automation audit?
A scheduled review of the workflows an agency relies on, covering owners, triggers, data access, error handling, cost and recovery. It produces a list of verified, broken and unnecessary automations.
How often should we audit automations?
Quarterly for stable workflows, and immediately after a platform change, a staff departure or a client-facing failure. The right cadence is the one that finds problems before a client does.
Who should own the audit?
The person who operates the workflows, backed by whoever administers the accounts. Ownership of each workflow matters more than ownership of the audit itself.
What should we test first?
Anything that can lose a client request, send the wrong information or change billing. Silent failures come next, then unused workflows and oversized schedules.
Can we audit without a test environment?
Partly. You can review configuration, permissions and logs, but you cannot confirm failure behaviour. Mark untested workflows clearly and schedule a safe test with fictional data.
Should we delete unused workflows immediately?
Disable them first and keep the configuration for a review period. Deleting removes the record of what was tried and may break an integration another workflow depends on.
Related reading
- Automate client onboarding with Make for a documented workflow with failure tests.
- Make pricing for agency workflows for how operations and credit limits are counted.
- n8n vs Make for a small agency for the platform trade-off.
- Client onboarding checklist for the manual steps automation should support.
- Client onboarding hub for the surrounding process.
- How we test for the evidence standard behind this guide.
Next step: export or list every live workflow, add an owner and a last-checked date to each row, and test the one that would hurt most if it failed silently.