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.
A few winters ago a small platform team decided to try anomaly detection on their payment service logs. The idea was sound, the budget was approved, and the engineer leading it was keen. Three months later the project had produced one half-finished notebook and a lot of apologetic stand-ups. Nobody had done anything wrong. The work had simply been sitting in other people’s inboxes, and nobody could say whose.
That story is why I am a fan of the humble waiting-for list. It is not a framework and it needs no tooling. It is a running note of each favour or deliverable you have requested from another person, plus the name of the owner, the day the request went out, and the date on which a friendly reminder goes out. For anyone running AIOps work, where almost every step depends on a person outside your team, it is the cheapest insurance I know.
Why AIOps work waits more than most
Machine-assisted operations looks like a technical subject, and the technical parts are real. But look at what a typical week of an AIOps effort contains. You need read access to log archives owned by another group. You need a security reviewer to approve a data export. You need the on-call team to agree what counts as a useful alert. You may need a vendor to answer a question about how their product stores history. Every one of those is a request, and every request has a quiet failure mode: it gets acknowledged, then forgotten.
Explained plainly, an AIOps project is mostly a chain of small handoffs around a small amount of modelling. If you only track your own tasks, you track the part that moves fast and ignore the part that decides your calendar. The waiting-for list tracks the other part.
What a good waiting-for list actually contains
Aim for a page you can scan in sixty seconds. Every entry needs five ingredients, and I have found that dropping any one of them makes the list useless within a fortnight.
- What you are waiting for, written as a deliverable and not a topic. “Read-only access to the archived application logs” works; “logging stuff” does not.
- The owner, named by role and team. If you cannot name an owner, that is itself the first finding, and you should go and get one.
- When you asked, because the age of a request is the single best predictor of whether it is stuck.
- What you will do if it arrives, so the item is attached to a next step on your own plan.
- When you will nudge, as a real date, not a vague “soon”.
Notice what is missing: priority scores, colour codes, status workflows. Those belong to the project plan. This list answers exactly one question, which is “what is out of my hands right now, and who has it?”
A concrete example from an incident-prediction pilot
Return to that payment-logs team. Had they kept a waiting-for list from day one, it might have looked like this after the first week.
- Archived service logs, read-only access. Owed by the data platform group. Asked on Monday. Nudge on the following Monday.
- Approval to copy a sample of logs into the experiment environment. Owed by the security reviewer. Asked on Tuesday. Nudge on Thursday, because reviews tend to sit unless someone prompts.
- A definition of “incident” that the on-call rota accepts. Owed by the on-call lead. Asked on Wednesday. Nudge after the next rota meeting.
- Answer on how long the monitoring vendor keeps raw metrics. Owed by the vendor’s support desk. Asked on Wednesday. Nudge in two working days; escalate through the account contact if silent.
Four lines, and suddenly the project has a shape. You can see that three of the four blockers are internal and cheap to chase, and that the fourth is the one with a genuine unknown. A manager glancing at it learns more than a status report would tell them.
The nudge is the point
Here is the human part that most guides skip. Chasing feels rude. Engineers in particular hate to be the person who writes “just checking in” for the third time. The waiting-for list quietly removes the awkwardness, because the nudge is no longer a mood; it is a scheduled, neutral task with a date on it. You are not nagging, you are doing what the list says.
Good nudges are short and make replying easy. Restate the request in one sentence, say what it unblocks, and offer the smallest possible action. “Could you confirm yes or no on the log copy? It unblocks the first model run” gets answered faster than a paragraph of context. If the second nudge is also met with silence, the list tells you it is time to change channel: a quick call, a word with the owner’s manager, or a note in the shared planning thread.
Where the list sits next to your existing tooling
No new product is required here. A pinned note, a section at the bottom of your team’s planning page, or a plain text file will do. What counts is opening it every day, not the file format. I like to review mine at the start of each day with a hot drink, scanning only for lines whose nudge date has arrived. Ten seconds per line, tops.
If your team already works with AIOps platforms, the list complements them rather than competing. Tools tell you what the systems are doing; the list tells you what the humans owe each other. For readers curious about the decision side of operations, this site has a piece on how machine-assisted analysis supports operational decision-making, and the waiting-for list is what keeps the people-side requirements of such a setup from being forgotten. Likewise, anyone starting from a messy alert stream may find the write-up on taming event floods with smarter tooling a useful companion, since the cleanup it describes also relies on approvals from other owners.
Common ways the list goes wrong
It grows into a graveyard. Lines pile up because nobody removes resolved ones. Fix: delete or archive an item the moment it arrives, and glance at anything older than three weeks with suspicion. An item that old is either dead, or it is a risk you should have raised.
It lists vague owners. “The security team” is not an owner; a role within it is. Vague owners are why requests evaporate, since everyone assumes someone else has it.
It confuses waiting with blocked. Some items you wait on do not stop you, and some do. Mark the ones on your critical path with a star so that the nudge schedule favours them. A project that is waiting on a non-critical approval can keep moving; one waiting on data access cannot.
It is private. A waiting-for list that only you can see helps only you. Sharing it at the weekly sync turns it into a lightweight way to surface cross-team dependencies without anyone being singled out.
A small aside on research you do outside work
The same habit travels well beyond engineering. Someone I know used a waiting-for list while researching their first home purchase, noting which questions they had put to relatives and advisers, and when they planned to follow up. For general reading on that topic, this property research site for home buyers in India says it offers general information only, so it makes sense as one input among several rather than a final word.
Starting tomorrow morning
Do this much and no more. Open a blank note. Write down every request you currently have in someone else’s hands, even the ones from last month. For each, add the owner, the date you asked, and a nudge date. You will probably find two or three items you had honestly forgotten, and at least one that has been stuck for weeks. That discovery alone repays the effort.
Then keep the list for a full sprint. Add lines when you ask, strike them when you receive, and send nudges on the dates written down. When you finish, count how many items were caught by a nudge rather than by a missed deadline. That number is the argument for continuing.
The team in my story did eventually revive their anomaly-detection project. They started with a list of four lines taped above a monitor, and the first model ran two weeks later. Nothing about the technology had changed. Only the waiting had become visible, and visible waiting is waiting that someone can end. If you also want a look at how forward-looking maintenance ideas are discussed in IT settings, the article about predictive maintenance in IT systems is worth a read alongside this one.