Automation projects fail by starting with the most interesting problem instead of the most repetitive one. A simple prioritization avoids that.
- ›How to rank automation candidates
- ›The five automations most teams should start with
- ›What not to automate
- ›Keeping automations from becoming debt
Ranking candidates
Score each candidate on three axes:
Frequency. How often does it happen? Daily beats monthly, always.
Annoyance. How much do people dislike doing it? This matters more than time saved, because adoption follows relief.
Risk. What happens if the automation is wrong? Low risk means you can ship it and iterate.
Rank by frequency times annoyance, divided by risk. The top of that list is almost never the interesting problem, and it is almost always the right place to start.
Five to start with
Recurring work creation. Anything that happens on a schedule (monthly reports, quarterly reviews, weekly checks) should be created automatically rather than remembered.
Handoff notifications. When work moves to a new owner or stage, tell the person who now has it. This eliminates the most common source of stalled work: nobody knew it was theirs.
Onboarding checklists. A new hire, a new client, or a new project triggers a standard set of tasks with owners and dates. High frequency in growing companies, and consistently annoying to do by hand.
Overdue escalation. Work past its date notifies the owner, then the manager. Simple, and it removes the chasing that consumes management time.
Status-driven side effects. A closed deal creates the onboarding project; a completed onboarding assigns the first training. These joins are pure mechanical work.
None of these need AI. They are rules, and rules are cheaper and more reliable. See AI agents vs automation.
What not to automate
Anything requiring judgment about a person. Performance flags, disciplinary triggers, and anything that categorizes people should stay human.
Customer communication without review. Automated messages that miss context are worse than slower human ones.
A broken process. Automating a bad process produces bad outcomes faster and makes the process harder to change, because now it is encoded.
Things that happen twice a year. The maintenance outweighs the saving.
Preventing automation debt
Every automation is a small permanent commitment. Three habits keep the pile healthy:
Warning: The most common automation failure is not a rule that breaks loudly. It is a rule that silently stops firing, which nobody notices because the absence of a notification looks like nothing happening.
FAQ
How many automations should a small team have?
Ten to twenty covers most of the value for a team under 50. Beyond that, check whether new rules are compensating for a process problem.
When should this become an AI agent instead?
When the condition cannot be written as a rule, or when the rule has accumulated more exceptions than cases.
Who should own automations?
One named person per functional area, reviewed quarterly. Shared ownership means no ownership.
Brainis includes module automations for mechanical rules and governed AI agents for judgment, on every plan. See pricing.
Sharing insights on business operations, AI, and modern team management.
Run your company on Brainis
All 11 Operating Systems on every plan, from $29 a month. No per-seat pricing — you pay for AI capacity, not headcount.
See pricing