The Client Automation Audit

The question is not whether something will go wrong. It is how many clients find out when it does.

• A bad deploy can only affect one client at a time.
• One client's runaway loop or rate limit cannot starve another client's workflows.
• You can hand over — or delete — a single client's automations without touching anyone else's.
• You know off the top of your head which clients share infrastructure.
• No client's data sits beside another's without a scoping rule you could explain to them out loud.

Everyone knows this one is a problem. Almost nobody has time to fix it before it matters.

• No client API key or OAuth token lives in a workflow node, an env file, or a spreadsheet.
• You could revoke one person's access to one client's credentials in under five minutes.
• You know every place a given client's credentials exist.
• Credential expiry is something you get warned about, not something you discover.
• If a contractor left tomorrow you would know exactly which secrets to rotate.

Error alerts catch crashes. The expensive failures never throw anything at all.

• You would know within an hour if a scheduled workflow simply did not run.
• You are alerted when a run 'succeeds' having processed zero records.
• You have a volume baseline per workflow, not only an error alert.
• One screen answers “is everything fine for this client right now?”
• Your most recent silent failure was found by you, not by the client.

If every new client costs the same as the last one, revenue and workload rise together and the business never gets easier.

• Adding client #N takes meaningfully less time than client #3 did.
• A workflow improvement reaches every client who uses it without opening each instance by hand.
• Onboarding is a written sequence, not a memory exercise.
• You could take on three new clients next month without hiring.
• You know your actual hours-to-onboard number rather than a guess.

The cheapest problem to fix early, and the most expensive one to discover late.

• You have read the licence of every engine you run for clients — the licence, not a forum summary.
• Where the software is part of what the client buys, you hold a licence that permits that.
• Your retainer prices maintenance explicitly, rather than as 'support if you need us'.
• You can produce a per-client cost of running their automations.
• If a client asked to leave tomorrow, handover is a process rather than a project.