What to automate first: a one-day process audit

7 October 2026 · 5 min read · Oleksandr Shylo

"We should automate something" is the most common sentence I hear on a discovery call, and it usually means three processes are eating the week while nobody has measured which. One day of structured looking replaces the guess with a ranked list. Here is the method; you can run it yourself or have me run it.

Why "automate everything" fails

Automation has a fixed cost per workflow (building, testing, documenting) and a variable benefit (hours saved every week). A workflow that runs twice a month never pays back; one that runs forty times a day pays back in a week. Teams that start with the most annoying process instead of the most frequent one spend the budget on the wrong thing and conclude "automation doesn't work for us".

The one-day method

Interviews (morning). 30–45 minutes with each person who touches customers, orders or money. One question, asked many ways: what did you do by hand yesterday? Not what the process document says — what actually happened.
Process map (early afternoon). Each recurring workflow as a line: trigger → steps → tools → hand-offs → where it breaks. Ten to twenty lines for a typical small company.
Numbers (late afternoon). For each line: runs per week, minutes per run, who does it, what an error costs. Most people overestimate the minutes and underestimate the frequency — ask for last week's actual count.
Score and rank. See the formula below. The top three are the roadmap; everything else waits.
Write it down. One page per workflow in the top three: current steps, proposed steps, tools, what "done" looks like, fixed price. This page is the scope for the build.

The scoring formula

Monthly saving = runs per week × minutes per run ÷ 60 × loaded hourly cost × 4.3. Payback = build cost ÷ monthly saving. Rank by payback, shortest first; break ties by error cost (a lost lead is worth more than a slow report).

Example: a form-to-CRM copy done 40 times a week at 6 minutes by someone costing $30/hour saves 40 × 6 ÷ 60 × 30 × 4.3 ≈ $516 a month. At a $200 build that is a 12-day payback — build it first. A monthly report that takes 3 hours saves $90 a month; at $300 it pays back in three months — still worth doing, but third on the list.

The usual first wins

Lead intake — form, ad or calendar → CRM record → instant alert → first reply. Highest frequency, highest error cost. (CRM integration)
Booking hygiene — reminders, cancellation recovery, no-show follow-up. (case)
Recurring reports — the Monday numbers assembled from Sheets or the CRM by a model and delivered on schedule. (case)
Routing of inbound email or messages to the right person or queue, with a draft reply attached. (AI automation)

What not to automate yet

Anything the team does differently every time — automate a process, not a habit.
Anything with fewer than ten runs a month, unless an error is expensive.
Anything whose source data is not yet in a tool (a notebook, a WhatsApp group). Get it into a form or a sheet first; that step alone often removes half the pain.

Doing it with me

The audit service is exactly this day, plus the written roadmap and a fixed-price quote for each of the top three. If you already know the one workflow that hurts, skip the audit and describe it in the form.

Related work

0leads dropped since launch

CLIENT PROJECT

Lead Processing Pipeline

Website form → HubSpot CRM → Telegram alert to sales → personalised email auto-reply. Zero manual steps.

Read the case →

CLIENT PROJECT

Booking Cancellation Recovery

Cancelled Cal.com call → Pipedrive lookup → cancellation logged → "Call Discovery Cancel" tag → deal moved back a stage so the lead gets a real follow-up.

Read the case →

3hsaved every week

CLIENT PROJECT