Key takeaways
- The best automation candidates are consistent: the same trigger, the same steps, every time.
- Start with one data source and a schedule. Every extra platform is another thing that can break.
- Exceptions and corner cases are where automations get expensive. Handle them by hand.
- If an automation needs constant fixing, the work has just moved to someone else.
FireOak is a small team, and we run dozens of automations every day. Maybe hundreds, depending on how you count. They do research, back things up, move information between systems, and remind us about things we’d otherwise forget.
People sometimes assume that means we automate everything. We don’t. With today’s tools, almost anything can be automated, so the more useful question is whether something is worth automating. Here’s how we think about it, starting with the automation that taught us the most.
The email that came from a missed deadline
Several years ago, around 2020, we missed the deadline to cancel an expensive platform. It renewed automatically for another year, and that was that. Vendors are notorious for this: many send nothing before an annual renewal, so staying on top of it is entirely your responsibility. Miss a few, and you’re wasting thousands of dollars a year.
I didn’t want that to happen again, so we built a daily “important upcoming dates” email.
Version one ran on Microsoft Power Automate. After some trial and error with different Microsoft 365 tools (Tasks, Planner, and others), we tied it to a group calendar. Anything I added to that calendar showed up in an email every morning listing the next 14 days: deliverable deadlines, monthly and annual subscription renewals, tax and other administrative deadlines, payroll runs, whatever I wanted.
The 14-day window turned out to be the key design choice. If I was traveling or buried in a project and ignored the email for a few days, the item was still there the next time I looked, well before the deadline passed. Nothing depended on me reading one specific email on one specific day.
Version two, rebuilt a few years ago, pulls from a database instead of a calendar. Each subscription or recurring item has a status (active or not) and a frequency (monthly or annual), and the automation works out the next upcoming date by itself. The daily email is formatted HTML now, much easier to scan than the original.
We keep the data up to date, but the automation itself hasn’t needed a single change in several years. It just runs.
At one point we had a truly ridiculous number of cloud subscriptions. (We’ve since cut many of them, partly because the daily email made the total so visible.) Getting notice before each automatic renewal, especially the annual ones, has paid for the time spent building this many times over.
It’s also a good example of what makes an automation worth building:
- One data source. A calendar, then a database. Not five places to look.
- One consistent trigger. Every morning, same time.
- The same steps every time. Look at the next 14 days, list what’s there, send the email.
- A real cost when it’s forgotten. Missed renewals cost real money.
What else we automate
Without giving away every detail, here’s the kind of work our automations handle:
- Research and monitoring. Technology trends, new cybersecurity threats, news in the fields our clients work in, and competitive intelligence for us and for clients. The automations gather and sort; people decide what matters.
- Backups. Our databases get backed up on a schedule. So do the automations themselves, because rebuilding one from memory after it breaks is no fun.
- Moving information between systems. When the same information needs to live in two places, an automation copies it so nobody has to retype it (and nobody introduces typos doing so).
- Small daily helpers. Reminders, renewal notices, and other nudges that keep things from slipping through the cracks.
None of these is glamorous. That’s the point.
Where automations go wrong
In our experience, automations almost always fail because the requirements get too complicated.
It usually starts with a simple, sensible workflow: when this happens, do that, then do something else. Then the exceptions arrive. This one client needs it done differently. Skip it for this person. Do it on the first Tuesday of the month, unless it’s a holiday. Each exception might be possible to automate. But the effort, and the technical debt you take on, is out of proportion to what you’re automating. A straightforward workflow becomes unwieldy, fragile, and constantly breaking.
It helps to separate what you really need from what would be nice to have. Corner cases and exceptions are usually in the second group, and they’re often cheaper to handle by hand.
The other warning sign is sprawl. Once a workflow touches seven different platforms, every one of them is something that can change, break, or expire. At that point, maintaining the automation can take more time than the original task did. And if keeping it running falls to your IT team, a technical consultant, or an outside vendor, the work has simply moved to someone else.
How to spot a good candidate
With the tools available now, the best candidates are “set it and forget it” workflows. Here’s a quick checklist we use:
- Is it consistent? Same trigger, same steps, same result. Do this, then that, then something else is a prime candidate.
- Does it come from one place? One system as the source of truth is ideal. Each additional platform adds risk.
- Does it run on a schedule or a clear trigger? Every morning; whenever a form is submitted; when a record changes status.
- Does it happen often enough to matter? A five-minute task done daily adds up to more than 20 hours a year.
- Is forgetting it expensive? Missed renewals, missed deadlines, lost data. Some automations are worth it even if they run rarely.
- Can you live without the exceptions? If the answer is “only if it handles every special case,” handle the special cases by hand and automate the rest.
If something checks most of these boxes, it’s probably worth automating. If it fails the first two, think hard before you start.
Start with something boring
If you’re looking for a first automation, start small and boring: a daily reminder email, a scheduled backup, copying form submissions into a spreadsheet or database. You’ll learn how the tools work, and you’ll build something that keeps quietly paying off.
For the bigger picture, see Why Automate?, and for help brainstorming, try the “Identify Automation Opportunities” prompt in How We Use AI (and How You Can Too).
If you’d like a second opinion on what’s worth automating in your organization (and what isn’t), that’s part of our process optimization work. You can also read more about how we test tools ourselves before recommending them, or get in touch.
Enjoyed this? Get the next one.
Signal, the FireOak newsletter: practical tips and thoughtful takes on technology, knowledge, and AI. No spam. Unsubscribe anytime.




