"What should we automate first?" is the most common question I get, and most answers to it are hedged non-answers. So here is a ranked list with an opinion attached, for a company around $10M revenue with 20 to 150 staff and no dedicated ops engineer.
First, the ranking criteria, because they matter more than the list. A task is worth automating when it scores high on three things: frequency (daily beats monthly), pain (someone's week is worse because of it), and rule-ness (the decision can be written down: if the invoice total is over X, route to Y). Frequency times pain times rule-ness. Anything low on rule-ness is off the table no matter how annoying it is, and that is not a euphemism. Fuzzy rules produce brittle automations that get quietly abandoned.
The ranking
1. Invoice and document data capture. Invoices arrive as PDFs and get typed into the accounting system. This is the highest-frequency, highest-rule-ness task in almost every company, and modern document AI handles it well with human review on exceptions. It is also the easiest win to measure: count the invoices, count the minutes, multiply. If you do one thing this year, do this one.
2. Report distribution that is currently a person exporting and emailing. Someone exports a sales report every Monday and pastes it into an email. This is pure robot work wearing a human costume. The automation is unglamorous: a scheduled job that pulls the numbers, formats them, and sends them. The compounding benefit is that the report now always exists, always on time, and the person who used to build it gets their Monday mornings back.
3. CRM-to-accounting handoffs. Won deal in the CRM becomes a customer record and an invoice draft in accounting, without re-typing. This one ranks third not because it is less valuable but because it usually requires the CRM hygiene work to be done first. When the data is clean, the handoff is straightforward; when it is not, you are automating chaos.
4. Approval routing. Quotes, discounts, refunds, time-off: anything that currently waits in an inbox. The automation is a routing rule plus a deadline plus an escalation. Small project, direct effect on the quote-to-cash cycle, and it makes your approvers better at approving because nothing sits invisible.
5. Customer status notifications. "Your order shipped," "your request is received," "your renewal is in two weeks." Customers rate you partly on communication, and status emails are the cheapest way to look organized. They also deflect the "any update?" calls that eat your team's afternoons.
What not to automate first
Anything you do personally that involves judgment. Your read on a customer, your call on a hire, your pricing instinct on a weird deal: these are not rule-based, and an automation that mimics them will be wrong in ways you will not notice. Automate the paperwork around the judgment, never the judgment.
Anything with fuzzy rules, no matter how painful. The exception-handling process that changes every quarter. The pricing decision that depends on the relationship. The complaint that needs a human. These are the automation graveyard, and every company I know has a project there they do not talk about.
The honest caveat on all five
Every item on the list assumes the underlying data is trustworthy. Automations amplify what they touch: a clean CRM with a handoff automation is a money machine, a dirty CRM with a handoff automation is a duplicate-email machine. If your data is a mess, the real first project is the two-week cleanup, and the automation ranking starts after that.
One more thing: ship them one at a time. The companies that automate successfully do one, watch it for a month, fix what drifted, then start the next. The companies that automate catastrophically launch four at once and spend the next quarter firefighting automations instead of doing the work they were supposed to replace.
Jai Indiongco
FOUNDER & PRINCIPAL CONSULTANT