Most teams automate the wrong thing first. They pick whatever annoys them most that morning instead of whatever quietly eats the most hours every week. Here's the ranking we use in every Greyflows audit — and the seductive first project that usually wastes the budget.
Score every candidate before you build anything
For each repetitive process, estimate three numbers:
- Hours per week the process consumes across everyone who touches it (be honest — count the re-checking and the fixing, not just the doing).
- Volatility: how often the inputs change shape. A report that always has the same 6 columns is a 1. An inbox where every email is different is a 4.
- Blast radius: what breaks in the business when it's done wrong. A mis-filed internal note is a 1; a mis-shipped customer order is a 5.
Your first automation should score high on hours, low on volatility, and low-to-medium on blast radius. That combination is where automation is boring, reliable and immediately worth the money.
The shortlist that usually wins
- Report assembly. Someone pulls numbers from 3–5 tools into a spreadsheet every Monday. Deterministic inputs, scheduled trigger, zero judgment. Two days of build, years of payback.
- Lead routing. New enquiry lands → enriched → assigned → Slack ping → CRM record. The logic is boring and the volume is what makes it expensive manually.
- Document intake. Invoices, PODs, applications arrive as attachments. OCR + rules + exception queue. Medium effort, huge hour savings.
- Onboarding sequences. New customer or new hire: the same 12 steps every time. This is what workflow software was invented for.
The one to skip
Almost every team wants to start with the AI chatbot that answers everything. It's the most seductive project on the list and the worst first build: high volatility inputs, high blast radius (it talks to your customers), constant prompt babysitting, and it automates a task nobody can fully specify yet. If a human can't write the rulebook, the machine can't follow it.
Rule of thumb: automate the process that already has a checklist. If it doesn't have a checklist, write the checklist first — that's a Tuesday, not a project.
What good looks like
A first automation is done when: it runs on a schedule or trigger without anyone remembering it exists, failures page a human with context (not silence), and there's a one-page runbook someone else could follow. Anything less is a demo, not an automation.
If you want an outside pair of eyes on your list, the free audit does exactly this: 45 minutes, a ranked candidate list, and an honest go/no-go.
Want this in your stack instead of your reading list? Start a project →