A three-person team notices they spend forty minutes every morning copying data between two tools. Someone writes a script. The script runs for a week, breaks on an edge case, and now one person spends Friday afternoons debugging it instead of shipping. Net time saved: negative.
This pattern repeats everywhere small teams operate. The instinct to automate is correct — repetition is expensive and dull. But execution goes wrong when teams automate too early, automate the wrong thing, or automate a process they haven't actually figured out yet.
Large organizations can absorb a bad automation. They have dedicated platform teams, on-call rotations, monitoring dashboards. A broken workflow gets triaged and fixed during business hours.
Small teams don't have that buffer. When an automation breaks at 2 a.m. on a Saturday, the person who fixes it is also the person who sells, ships, and supports. Every hour of maintenance comes directly out of something else that matters.
There's also a psychological trap. Small teams feel repetition more acutely — fewer hands, more tasks per person. That urgency pushes them toward quick fixes before they understand the problem well enough to automate it correctly.
1. Auto-routing that hides broken inputs. A team builds a rule to sort incoming requests into buckets. The sorting works, but it quietly absorbs malformed requests that a human used to flag. Bad data accumulates downstream for weeks before anyone notices.
2. Scheduled reports nobody reads. Easy to set up, easy to forget. The report runs every Monday morning. After the first month, nobody opens it. But it still consumes compute, generates notifications, and occasionally breaks in ways that trigger false alarms.
3. Auto-retry loops without limits. A job fails, so the automation retries it. And retries. And retries. Without a circuit breaker, a small failure cascades — a queue fills up, a rate limit gets hit, a downstream system gets hammered.
4. Approval workflows that remove the conversation. A manual approval process is slow, but the conversation around it surfaces context: "This request looks unusual — did you mean to change that field?" Automating the approval removes the friction and the judgment at the same time.
5. Data syncs with no conflict resolution. Two systems hold the same record. An automation keeps them in sync — until both sides change at the same time. Now you have a merge conflict with no human in the loop and no clear rule for which version wins.
You can spot a backfiring automation before it causes real damage.
Rising exception volume. The automation works most of the time, but "most of the time" means someone handles exceptions manually. If the exception rate climbs, the automation is masking a deeper problem, not solving it.
Unclear ownership. Ask the team: who's responsible when this breaks? If the answer involves shrugs or pointing at someone who left three months ago, you have a liability, not a tool.
The underlying process keeps changing. Automating a stable, well-understood process is smart. Automating something you're still iterating on means you'll rewrite the automation every time the process evolves. That rewrite cost adds up fast.
Nobody can explain what it does without reading the code. If the logic lives only in the implementation and not in anyone's head, the automation is fragile. The next person who touches it will either break it or be too afraid to change it.
When you suspect an automation costs more than it saves, run it through three questions:
What does this save per week, in minutes? Be honest. Measure it. Don't count the theoretical time saved — count the actual time reclaimed by the people who used to do the task.
What does this cost per week, in maintenance and attention? Include debugging, monitoring, the mental overhead of worrying about it, and the time spent handling exceptions.
If this broke permanently tomorrow, what would we do? If the answer is "go back to doing it by hand and barely notice," the automation isn't critical. Kill it or simplify it.
Any automation where the cost column exceeds the savings column — even slightly — deserves scrutiny. Small negative margins compound over months.
Automate the boring parts. Tasks that are stable, well-understood, high-frequency, and low-judgment. Data formatting. Notifications. Scheduled cleanup of things that follow clear rules.
Never automate what you don't yet understand. If you're still learning why a process works the way it does, keep doing it by hand. The manual version teaches you things the automated version hides.
The best automation a small team can build is one they forget about — not because they lost track of it, but because it runs quietly, handles its own edge cases, and never needs attention. That kind of automation starts with patience, not speed.
Be the first to comment.
0 comments
Loading comments...