Engineering managers are pulled in too many directions. The automation question isn’t “what could I automate?” — it’s “what should I automate first, given that my time is the constraint?”
Here’s my current answer, based on what I’ve actually seen pay off.
Start with things that are high-frequency, low-judgment, and currently manual
The ROI formula for automation is straightforward: frequency × time saved, discounted by implementation cost and maintenance burden. But judgment work doesn’t belong in that formula — automating it creates the illusion of efficiency while removing the thing that makes the work valuable.
The sweet spot is work that’s:
- Done repeatedly (daily or weekly, not occasionally)
- Largely procedural — the same process produces the right outcome most of the time
- Currently consuming time that could go to higher-value work
What this looks like in practice
Sprint and delivery reporting
Pulling metrics from your project tracking system, formatting them, and sharing them — this is a solved problem. If you’re doing it manually, you’re wasting the time that should go to interpreting the data, not producing it.
Most team dashboards can be automated in a day. Status report templates connected to live data take a bit longer but are worth the investment.
The goal isn’t to eliminate the manager’s judgment — it’s to eliminate the prep work so the judgment gets to happen on time.
On-call rotation scheduling
Rotation scheduling has well-defined constraints: coverage requirements, equitable distribution, individual unavailability. These are exactly the kinds of problems that scheduling tools and scripts solve well.
The manual version of this takes 20-30 minutes every few weeks and generates more edge-case complaints than almost any other operational task. Automate it.
New engineer setup and onboarding checklists
The first week of a new engineer’s tenure should not depend on a manager remembering to do things in the right order. Automated onboarding workflows — provisioning access, sending welcome information, scheduling the right conversations — make the process consistent and reduce the cognitive load on both the manager and the new hire.
This is also where most onboarding failures happen. Inconsistency in the process leads to inconsistent results.
Recurring meeting agendas
Standard 1:1 prompts, team retrospective formats, sprint review templates — these should exist in a tool that surfaces them automatically, not in someone’s memory or a document they have to find.
This sounds trivial. In practice, the overhead of maintaining meeting structure adds up, and automation makes it more likely the structure actually gets used.
Alert routing and incident assignment
Alerts that land in a shared channel and require someone to manually assign them create inconsistency and delayed response. Automated routing to the right person on-call, with clear escalation rules, is both faster and more reliable.
If your current process is “whoever sees it first deals with it,” you have a people-and-luck system masquerading as a process.
What not to automate
Performance management. Tools that generate automated performance summaries or recommend action based on activity metrics are appealing and risky. They surface numbers but not context, and they can produce confident-looking outputs that are systematically wrong about the things that matter.
Hiring decisions. Screening automation that filters candidates without human review creates legal exposure and misses the people who don’t fit the filter but would be excellent hires.
Judgment calls disguised as process. Any decision where the right answer depends significantly on context — team dynamics, organizational politics, technical risk — resists automation. If you find yourself automating something and the results feel off, that’s usually why.
The implementation order
- Metrics and reporting — immediate ROI, low maintenance cost, visible impact.
- Onboarding workflows — benefits every new hire from the moment it’s deployed.
- Meeting structure — low effort, consistent value.
- On-call scheduling — medium effort, eliminates a recurring source of friction.
- Alert routing — higher complexity, but high value if your incident volume justifies it.
The principle behind the list
Good automation extends a manager’s capacity for judgment work by eliminating the procedural work around it. Bad automation substitutes for judgment — and then you have all the risk of the decision without any of the human understanding that makes the judgment worth having.
The test I use: “If this automation made a mistake, would I notice before it mattered?” If yes, automate. If no, keep the human in the loop.