In the Loop with Anoop is a monthly series where Anoop Mohan, Chief Product & Technology Officer at Augury, shares his perspective on Industrial AI, the future of manufacturing work, and what it actually takes to build technology that serves the people on the plant floor. Subscribe on LinkedIn to get each new post delivered directly to you.
The alert came in overnight. By the time the reliability engineer got to their desk, it was already waiting: a critical fan, flagged red, status unknown.
What happened over the next three days is something a colleague on our field team described to me recently, and it has stayed with me. Not because it was unusual, but because it was completely ordinary.
Day one: the alert. The engineer opened the APM dashboard and started working through it, clicking across screens to understand the scope. One asset or several? One plant or more? The root cause analysis didn’t give a clear answer, so they reached out for support and waited. Day two: confirmation of the findings, passed back through the team. Day three: a planner dispatched to the floor to scope the repair before anyone could write the work order.
Three days. A fault getting worse, while everyone did exactly their job.
Nobody failed here. The diagnosis simply lived in multiple places at once, and no single person had the full picture until all the pieces had traveled through enough hands. That is the real shape of the problem. The time doesn’t disappear in one visible block. It bleeds out in the handoffs.
The swivel chair at work
If you’ve read my earlier thinking on the Agency Gap and swivel chair operations, you know I use that term to describe any moment where a person has to jump between systems to complete one task. The more systems, the wider the gap between knowing and doing.
For a reliability engineer managing a fault, the swivel chair count is significant. At a well-run site, they’re moving across five or six touchpoints before a work order can be written: the APM dashboard, the RCA tool, the CMMS, the production system, and whatever communication tools the team uses to coordinate. Five or six touchpoints is the floor, not the ceiling.
And there’s a second compounding pressure layered on top of all of it: the work order timing decision. Open it too early and you’re sending a technician to a problem that hasn’t been fully diagnosed. Open it too late and a developing fault becomes an unplanned failure. The window to get it right is narrow. The information required to hit it is scattered.
Now consider what happens at the majority of sites, where there is no dedicated reliability engineer running this process at all. The reliability function is often a half-job, shared with other responsibilities. In smaller plants, it belongs to no one in particular. The faults still come in. The work still needs to happen. It lands on whoever is closest: a maintenance lead, a supervisor, a planner who already has a full day ahead of them.
When that’s the reality, there simply isn’t enough bandwidth to act on every alert. Teams have to prioritize, and prioritization under pressure means the most urgent issues get attention and everything else waits. That’s where the expensive failures come from. Not the fault someone watched and misjudged, but the one nobody had time to get to.
What the same morning looks like with a reliability agent
A reliability agent runs the same workflow a reliability engineer would, starting the moment the alert fires. How critical is this? How severe? How widespread? Unlike a person managing competing priorities across a shift, the agent works through red alerts and orange ones. Everything that warrants attention gets attention, on every shift, without the constraint of available hours.
It performs the root cause analysis. It checks maintenance logs for prior occurrences and historical fixes. It produces a recommendation on the work order: not a decision, but a documented recommendation with its reasoning visible, for the reliability engineer to act on.
The human remains in the loop. What changes is that they’re no longer spending their morning reconstructing a picture across five systems that the agent has already assembled.
When we show this to our design partners, the reaction moves through two stages. First: disbelief. “Is this actually possible?” Then, quickly, practicality. Once people see it working, they stop asking whether it works. They ask whether it works with their setup, their CMMS, their alert structure, their team.
That’s the right set of questions. The agent does the diagnostic work, surfaces the recommendation, and documents its reasoning. The reliability engineer reviews it and decides whether to move forward. Everything is done for them up to that point, which is itself a massive step forward in productivity. The expertise and the final call still belong to the person who’s spent years earning them.
The productivity math
Our internal goal at Augury is to bring 30% productivity improvement to every persona in a manufacturing plant. That figure isn’t arbitrary. It reflects a conservative estimate of how much of a specialist’s day is currently consumed by digital coordination work: manually syncing data between systems, chasing status updates, and navigating software that was never designed to talk to each other. The agent reclaims that time. And 30% at the individual level compounds quickly: across a team, across multiple plants, the business-level impact is well beyond any single number.
For a reliability engineer moving across five or six systems to diagnose a single fault, 30% is not a stretch target. It may be conservative.
The goal of the reliability agent is not to replace the reliability engineer. It’s to return their time, so they can apply their expertise, their judgment, and the knowledge they’ve built over years on the plant floor to the work that actually requires it. The agent handles the swivel chair. The engineer handles everything the agent cannot.
That three-day fault my colleague described? In an agent-led workflow, the diagnosis, the recommendation, and the documented reasoning are ready before the engineer has finished their first cup of coffee. The decision still belongs to them. It just doesn’t cost them the better part of a week to get there.
Ready to learn more about the Industrial AI Workforce? See how it works.
