Upgrade & Secure Your Future with DevOps, SRE, DevSecOps, MLOps!
We spend hours scrolling social media and waste money on things we forget, but won’t spend 30 minutes a day earning certifications that can change our lives.
Master in DevOps, SRE, DevSecOps & MLOps by DevOps School!
Learn from Guru Rajesh Kumar and double your salary in just one year.
Anyone who has carried a pager for a few years knows the strange kind of tiredness that arrives at three in the afternoon on a day when nothing actually broke. The cause is rarely the incidents. It is the hundred small choices that pile up around them: which dashboard to open first, which query language to reach for, which of five similar remediation scripts to trust. For people building a career in AIOps, learning to limit those choices is a skill as real as learning Python, and it tends to get noticed by the people who decide promotions.
This piece follows one imaginary engineer, an operations analyst moving into an AIOps role at a mid-sized company, and the way her working week changed once she stopped treating every decision as a fresh one. None of this really happened to one person, yet the pattern repeats across teams, and the habits translate to almost any team that mixes machine-learning tooling with ordinary infrastructure work.
The week that taught her about decision fatigue
In her first quarter she was handed a broad brief: use anomaly detection, event correlation and automation to make the platform calmer. The company already had four monitoring tools, two notebook environments, a half-built feature store and a shared drive full of experimental models. Nobody told her which to use, and she took that as freedom. By Wednesday of her third week she had opened eleven tabs, started three prototypes and finished none of them.
What she noticed was not a lack of ability. It was that every morning began with a menu. Which model family should she try for seasonality? Which data window? Which alert channel for the results? Each question was small, and each was answerable, but answering them one at a time left her with less attention for the question that mattered: does this change help the on-call engineers sleep?
Psychologists have written about this for decades under names like decision fatigue and choice overload. The details of the research are debated, and it would be dishonest to promise a neat formula. The practical observation survives anyway: a person who has to re-decide the same things every day spends part of each day on bookkeeping rather than on judgment.
Deciding once: the short menu
Her fix was unglamorous. On a Friday afternoon she wrote down, on a single page, the choices she kept re-making, and for each one she shortened the menu. For time-series work she would start with exactly two approaches, a simple seasonal baseline and one gradient-boosted option, and only leave that pair when both had clearly failed. For notebooks she picked one environment and archived the rest from her bookmarks. For output she chose one destination for every experiment summary.
The point was not that these were the optimal picks. Several of them were arguably second best. The point was that the choice was made once, calmly, with a clear head, instead of forty times under pressure. When she later wanted to deviate, deviating became a deliberate act with a written reason, which is a very different thing from drifting.
- A short approved list of modeling approaches for each common problem shape.
- One notebook and one repository layout for every experiment.
- A fixed template for the end-of-experiment summary, so no one debates the format.
- A named place where results go, so reviewers always know where to look.
What a career in AIOps rewards
It would be easy to read this as advice for the lazy. It is closer to the opposite. The engineers who advance in this field seldom turn out to be the people with the longest list of algorithms. They are the ones whose work is easy to review, easy to repeat and easy to hand to someone else. A short menu produces exactly that. When your teammates can predict how you will approach a problem, they trust the output faster, and trust is the scarce currency on any team experimenting with automation.
If you are planning your own growth, this changes what you practice. Instead of collecting tools, practice narrowing. Take a messy problem, such as noisy events from a deployment pipeline, and ask which two options deserve a trial and which eight can be dismissed without a trial. Being able to explain why you dismissed them is a stronger interview answer than being able to name them all.
Some of this connects to the way we write about event handling on this site. A walkthrough of cutting through event noise with smarter tooling shows what happens downstream when the upstream choices are tidy, and a piece on how decision support works inside IT operations makes a similar point about restricting the options an engineer sees at the moment of action.
Building the habit without a manifesto
The analyst in our story did not announce a new method to her team. She simply began keeping her one-page list pinned in the corner of her screen. Three habits made it stick.
Review the list on a schedule, not on a whim
Once a quarter she read the list and asked whether any entry had stopped earning its place. That kept the menu short without freezing it. A list that never changes becomes dogma, and a list that changes daily is no list at all.
Keep an exceptions note
Whenever she stepped outside the approved list, she wrote two lines: what she used and why. Over time this note became a rich source of promotion-worthy evidence, because it showed a pattern of reasoned departures rather than random ones. It also revealed which menu items needed replacing.
Protect the hard decisions
With the routine choices settled, the remaining decisions were the interesting ones: whether a detected anomaly was worth waking someone for, whether a proposed automation was safe to run unattended, whether a model was drifting. She had more attention left for these, and her reviewers noticed that her judgment calls were calmer and better explained.
A note from outside the operations world
Decision fatigue is not a tech-only problem, and the same logic appears in every function a team depends on. Inoroff, a site presenting time-off and check-in software for smaller companies, offers a handy illustration from a different corner of the workplace: it describes moving scheduling questions into a fixed flow of requests and approvals, so that people are not renegotiating the same arrangements by message each time. You can see how the Inoroff leave and attendance site frames that idea for itself; whether it suits any given team is a separate question.
Common objections, answered honestly
When she described her approach to a colleague, she heard three objections, and they are worth answering because you will hear them too.
“Won’t I miss better options?” Sometimes, yes. The cost of a narrow menu is that occasionally a superior method sits outside it. The counterweight is the exceptions note and the quarterly review: new candidates enter the list deliberately. In practice most of the gains come from finishing things, and finishing is easier when you are not choosing from scratch.
“Isn’t flexibility the whole point of experimentation?” Experimentation needs a controlled variable. If the notebook, data window, summary format and channel all change between runs, you cannot tell what caused a difference. A fixed frame is what makes the experiment readable.
“Does a junior engineer even have the standing to set defaults?” You do not need to set them for the team. Set them for yourself, document them, and let others adopt what helps. Several teams have drifted toward shared lists that way, starting from one person’s tidy page.
Turning it into portfolio material
If you are building a career story, convert the practice into something you can show. A short document titled with the problem you solved, the options you ruled out, the two you tried and the reason one won is a compact piece of evidence about how you think. Hiring managers in operations-heavy roles tend to value this over a long list of certificates, because it demonstrates the thing they cannot easily test: how you behave when the problem has no instructions.
Another useful habit is to practice decision logs on small personal projects. Pick a toy dataset of server metrics, restrict yourself to two methods, and jot down half a page about what you would change with a larger menu. The exercise costs an evening and builds the reflex.
For teams that are earlier in their journey, the article on bringing AIOps into an IT team step by step makes a good companion read, since adoption goes more smoothly when the tooling choices are limited at the start.
What changed for her
By the end of that year the analyst had shipped four small models, retired two noisy scripts and written the team’s first one-page guide on how experiments are run. None of it required a heroic effort. It required deciding, early and once, what she would not spend attention on.
The lesson for your own path is modest. Attention is a limited resource, and every option you keep on the table draws from it. Shorten the menu, write down the rule, review it now and then, and save your best thinking for the calls that actually carry risk. In a field where machines handle more and more of the repetitive work, the people who protect their judgment will be the ones others come to for it.