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 worked on an AIOps programme for more than a season knows how quickly a goal written in good faith can become a poor fit for the system it was meant to improve. A model that was supposed to be the centrepiece of the year gets overtaken by a data-quality problem nobody anticipated; a target for automated remediation loses its meaning after a platform migration. Rereading your goals monthly is a modest habit that catches this drift early. Below are eight goals typical of AI and AIOps work, each paired with the question worth asking when you reread it, and a clear verdict for each.
Why a monthly reread of AI and AIOps goals pays off
The method is simple. Once a month, set aside a short sitting, open the written list of team or personal engineering goals, and read each one aloud as if a colleague had just handed it to you. Ask whether it still matches the systems you actually run, the data you actually have, and the priorities your organisation actually holds. Then decide. Leave it alone, reshape it, or let it go, and put the effort somewhere that matters now.
The reason this belongs in AIOps specifically is the speed of change. Telemetry sources come and go, models degrade as the environment shifts, and tooling decisions made in one quarter quietly constrain the next. A goal that referenced a particular data source or a particular model family can be obsolete without anyone announcing it. A monthly reread gives that obsolescence a place to surface.
1. “Train a model to flag anomalies in service metrics”
This is the classic opening goal, and it ages in an interesting way. When you reread it, ask which metrics the model was meant to cover and whether those metrics are still the ones the team relies on. If the service has been split or renamed, the model may be watching a signal nobody reads. The usual outcome is an adjustment: narrow the goal to the services that remain, or widen it to the ones that replaced them. Only drop it if the business case itself has disappeared.
2. “Reduce the time to correlate related events”
Correlation goals tend to be phrased vaguely, which makes them easy to keep forever without progress. At the monthly sitting, check whether the goal names a concrete, observable change in how incident responders work. If you cannot describe what an engineer would do differently on a Tuesday afternoon, rewrite it so that you can. A sharper sentence is a better outcome than a perpetually green status.
3. “Automate a first-line remediation for a recurring failure”
Here the reread question is about recurrence. Has the failure kept happening at the rate that justified the goal? If a fix upstream has quietly removed the failure, automating its remediation is effort spent on a problem that no longer exists. Drop it with a clear conscience and note why. If the failure has changed shape, adjust the goal so the automation targets the current pattern rather than the original one.
4. “Clean and standardise the training data”
Data goals tend to get quietly carried from month to month because they never feel finished. Reread this one with an honest question: which datasets does a live model or a planned model actually use? Cleaning data nobody consumes is a pleasant way to feel productive. Rewrite the goal around the specific feeds that matter, and set the rest aside for now.
5. “Learn a new framework or platform”
Personal learning goals belong in the same reread as the team’s. Ask whether the framework you chose last quarter is still the one your work touches. Engineers often write a learning goal in a burst of enthusiasm and then watch the stack settle on something else. If your day-to-day tools moved, the goal should move with them. Keeping an outdated learning goal is how a full year slips by while you study something you never use.
As a side note on how learning is laid out elsewhere, AirCrewU is a site for people who study aviation topics online, and its menu splits into courses, subject areas and career paths. That is a handy reminder that even a learning catalogue is organised around goals people revisit as their direction shifts.
6. “Build a feedback loop between operators and the model”
This goal usually involves people more than software, so the reread should include them. Ask whether operators are still being asked for the feedback the loop depends on, and whether anyone reads it. If the answer is no, the goal has stalled on a human problem, and adjusting it means naming an owner and a small, realistic cadence rather than a large integration. For related thinking on introducing this kind of capability to a team, the post on bringing AIOps into an IT team the right way covers the people side in more depth.
7. “Document how our predictions are made and used”
Documentation goals are easy to postpone and easy to forget. Rereading this one monthly forces a decision: is the documentation still describing the system as it exists? If the pipeline changed, last month’s write-up may now be misleading, which is worse than none. Adjust the goal to cover the current design, and keep it small enough to finish. An accurate half-page beats a sprawling document that describes a model retired two releases ago.
A companion read on the subject of predictive tooling is the explanation of how predictive models support IT performance, which is a good prompt for deciding what your own documentation should actually cover.
8. “Retire the dashboards and jobs we no longer trust”
The final item is a goal about subtraction. Each month, ask which jobs, scheduled reports and alert rules were created for goals that are no longer on the list, and remove one stale artefact. That keeps the operational surface honest.
How to run the sitting without turning it into a ceremony
Keep the format small. A single page of goals, a pen or a text editor, and about half an hour is plenty. Read each goal, say out loud what has changed since you last looked, and mark it with one word: keep, reshape or retire. Resist the temptation to discuss progress; that belongs to a different meeting. The reread is only about fit.
Write down the reason for any change in a sentence. This takes seconds and pays off later, because it tells you whether a goal was dropped because it was finished, because it was wrong, or because circumstances moved. Those three reasons call for different responses, and a plain note keeps them from blurring together.
A few cautions apply to AI and AIOps teams in particular. First, do not let enthusiasm for a new technique add goals faster than the reread removes them. A list that only grows is a list nobody is reading. Second, watch for goals that depend on a vendor product or an internal platform that may change; flag those with a note so the next reread checks the dependency directly. Third, be honest about goals that were really wishes. If a goal has been adjusted three months in a row without any work behind it, drop it and see whether anyone misses it.
One objection, answered briefly
Some engineers worry that a monthly look invites constant churn. In practice most goals stay as they are and the sitting is short. The value comes from the occasional goal that would have eaten a month of effort in the wrong direction. Where a goal is set by management, bring a written observation to its owner: what changed, why the wording no longer fits, and a suggested revision.
A closing thought on goals that still fit
The aim of this habit is not to change your goals; it is to know that the ones you are working toward are the ones you chose on purpose. When a goal survives a reread, you can pursue it with a clearer conscience, knowing you checked it against the current state of your systems. When it does not survive, you have saved yourself weeks of effort. Either result earns its half hour.
Pick a date this month, put it in the calendar, and bring the list. Read it slowly and ask whether each item still fits.