Which process should you automate first?
Most automation projects pick the wrong workflow. They pick the one that is annoying, or the one someone saw a demo of, rather than the one that is both expensive and stable enough to build against. This audit separates those two things, because they are the two that decide whether a build holds.
The prize
How many hours a year the workflow eats, and what those hours cost at your staff rate. A process that runs 25 times a week for 12 minutes is 240 hours a year. That is the number worth arguing about, not whether the tool is impressive.
The readiness
Whether the process is the same every time, whether it is written down, and whether anyone would own it once it runs. A workflow can be worth a fortune and still be unbuildable this month.
The blast radius
What happens if it produces a wrong output and nobody notices. This never rules a workflow out. It decides whether a person approves the result before it leaves the building.
Three common workflows are filled in as examples. Change them, delete them, and add your own.
Wages plus on-costs for whoever does this work. Used to turn hours into dollars.
a year, worth $13,200
Big enough to matter and stable enough to build against. This is where a build pays back fastest.
a year, worth $6,600
The prize is real, but the process is not consistent or written down enough to build against yet. Run it manually the same way three times and write down what you actually did. That is a fortnight of work, not a project, and it is the difference between a build that holds and one that breaks in month two.
a year, worth $15,840
The prize is real, but the process is not consistent or written down enough to build against yet. Run it manually the same way three times and write down what you actually did. That is a fortnight of work, not a project, and it is the difference between a build that holds and one that breaks in month two.
What the audit says
hours a year across every workflow listed
of staff time, at $55 an hour
ready to build now, holding 240 hours ($13,200)
One workflow produce output that is expensive to get wrong and hard to catch. Those get built with a person approving the result before it goes out, not left unattended.
Take this with you
Copy the audit as plain text. It is yours. Paste it into an enquiry, send it to your accountant, or hand it to whichever developer you decide to talk to.
WORKFLOW AUDIT Assumed staff cost: $55/hour. Working year: 48 weeks. Total time across the workflows below: 648 hours a year ($35,640). 1. Responding to new enquiries - Build this first 25x per week, 12 min per run = 240 hours a year ($13,200) Readiness 83/100. Same steps every time. Written down, not really followed. Has an owner. If it goes wrong: obvious straight away, cheap to fix. 2. Preparing quotes - Fix the process first 8x per week, 45 min per run = 288 hours a year ($15,840) Readiness 20/100. Different every time. Lives in someone's head. Has an owner. If it goes wrong: expensive, and easy to miss for months. Needs a human approval step. 3. Chasing unpaid invoices - Fix the process first 10x per week, 15 min per run = 120 hours a year ($6,600) Readiness 45/100. Same steps every time. Lives in someone's head. No clear owner. If it goes wrong: would be noticed within a week.
The audit is the easy part
Knowing which workflow to build is worth something. Having someone in your business who can build it, change it next quarter and fix it when a supplier changes their form is worth considerably more. buildAutomation trains two or three of your own staff to do exactly that, and builds the hard flows alongside them.
Bring your audit. It is the fastest possible start to a first conversation.
See buildAutomationQuestions
Do I need to give you my email to use this?
No. Nothing is sent anywhere, nothing is stored, and there is no download gate. The audit runs entirely in your browser and the copy button puts the result on your clipboard. If you want to talk to us afterwards you can, but the tool is not a form in disguise.
What counts as one workflow?
A sequence of steps with a beginning and an end that produces one outcome, like turning an enquiry into a booked job, or turning a completed job into a paid invoice. If you cannot say what finishing it looks like, it is probably two workflows.
Why does a high-value workflow sometimes say to fix the process first?
Because a build can only be as consistent as the process it copies. If the same job is done three different ways by three people, an automation will encode one of those ways and quietly break the other two. Running it manually the same way three times and writing down what you did is usually a fortnight of work, and it is what makes the build hold.
Does a high blast radius mean I should not automate it?
No. It means the output needs a human approving it before it goes to a customer. Plenty of the highest-value automations in a business are draft-and-review rather than fully unattended, and that is a design decision rather than a failure.
What do I do with the result?
Copy it out and use it. It is a scoped starting point you can hand to any developer, take to your accountant, or use to argue the case internally. It is deliberately portable, because a scope that only makes sense to the agency that wrote it is a sales document rather than a scope.
Read next
The AI Readiness Audit
The people side of this: the twelve questions that surface which workflows are bleeding time in the first place.
Break a workflow into tasks
Automation decisions are made at the task level, not the workflow level. How to split one and mark the human steps.
A prompt pack is not automation
Why a documented prompt library is still a person doing the work by hand, and what has to change for it to count.
Run it manually three times first
What to do with a workflow the audit says is valuable but not ready to build against yet.