When a six-person travel desk in Patna decided to prepare for AI-assisted operations, nobody on the team could write a monitoring rule, and only one of them had ever read a server log. This is the story of how they got there anyway, told plainly, because the lesson matters for every small outfit promoting Bihar to the world: the technology was never the hard part. The people part was.
Call the team Kiran’s crew. They run a portal that lists village stays, festival dates, food trails and seasonal itineraries across Bihar. Their income depends on enquiry forms working, pages loading fast during festival rushes, and someone noticing quickly when either stops being true. For years, “noticing quickly” meant whoever happened to check the inbox first.
The chores that were quietly eating the team
Before any talk of smarter tooling, Kiran made the crew write down every repetitive task they did in a fortnight. The list was embarrassing and useful in equal measure:
- Checking each morning whether the enquiry form had actually delivered overnight messages, because it failed silently about once a month.
- Watching page speed by hand during the Chhath and Sonepur Mela season, when traffic triples and the server occasionally buckles.
- Spotting duplicate bookings for the same homestay dates, usually after an awkward phone call from a host.
- Restarting a flaky image service whenever gallery photos stopped loading, a fix everyone knew but only two people remembered how to run.
- Scanning payment-gateway emails for failed transactions and matching them to bookings, line by line.
Every item on that list follows a pattern: a signal appears, a human notices it late, a known response gets applied by hand. That pattern is exactly what AI-assisted operations exists to absorb. Software that learns what normal looks like can flag the abnormal form queue or the odd traffic dip before a guest complains, and routine fixes like the image-service restart can be wired to trigger themselves. The team had read about this style of working in a piece on running dependable digital services with continuous automation and assumed it was for big companies. It is not. It is for anyone whose festival-season revenue dies when a form breaks on a Sunday.
Why the first attempt at AI-assisted operations stalled
Kiran’s crew did what most small teams do first: they installed tools. A free uptime checker, an alerting plugin, a half-configured log viewer. Within six weeks the alerts were being ignored because nobody could tell a real problem from noise, and the log viewer had never been opened after the first demo.
The diagnosis, in hindsight, was simple. Tools answer questions; the team didn’t yet know which questions to ask. Nobody understood how a detection model decides something is unusual, so nobody trusted it. Nobody could safely script an automatic fix, so every alert still ended in a manual scramble. The gap wasn’t software. It was skills.
Upskilling one person at a time, on a schedule
The fix the team settled on was unglamorous: structured study, one person at a time, while everyone else covered the desk. They picked a structured AIOps training path at AIOps School because it broke the subject into levels rather than dumping everything at once — a ladder that starts with fundamentals and climbs through implementation toward strategy and architecture, so a beginner from a travel portal doesn’t have to pretend to be an engineer on day one.
Rohit, the de facto tech person, went first. His foundation-level course ran a month, with a weekly commitment in the ten-to-twelve-hour band that he covered using evenings plus a single weekend block, manageable alongside the desk. He came back able to explain, in Hindi and plain English both, what the uptime checker was actually measuring and why half their alert rules were useless. The crew rewrote them together in an afternoon.
Three months later, Rohit took the engineer-level step: a forty-five-day stretch at a heavier weekly load, built around teaching models to recognise unusual behaviour, assembling a proper monitoring stack, and scripting responses that run without a human. The practical labs mattered most — he rehearsed on realistic systems before touching the live portal. His first production win was small and perfect: the image service now restarts itself and posts a one-line note to the team chat, instead of waiting for someone to notice grey boxes where Bihar’s village homestay photos should be.
Readiness is personal before it is technical
The crew’s second insight deserves its own heading: a team is not “ready” as a block. Each person needed a different on-ramp.
- Rohit needed the full technical track, and got it.
- Kiran, who runs commercial decisions, did only the first-level fundamentals — enough to budget for the work and challenge vendors without bluffing.
- The two content writers needed just two internal sessions: what an anomaly flag means, and when to escalate instead of ignoring it.
- The host-relations pair needed the duplicate-booking detector explained once, with examples, so they’d trust its warnings during Mela season.
Mapping people to levels like this cost nothing and prevented the classic failure where one trained person becomes a bottleneck and everyone else keeps old habits. When the crew later evaluated new software, they followed the same discipline they’d recommend for any purchase — the approach laid out in their favourite method for vetting business software through digital catalogues — but this time with someone in the room who understood what the vendors were actually selling.
What changed, in numbers the crew actually tracks
No miracle claims here, just desk-level facts. Overnight form failures now page a phone instead of waiting for the morning inbox ritual, so the response gap shrank from hours to minutes. The festival-season speed watch stopped being a person’s job; a model watches the curve and speaks up only when the pattern genuinely breaks. Duplicate bookings get caught before the host does. The crew estimates they reclaimed close to a full working day per week across the team — time that now goes into new food-trail content, which is the work that actually grows a Bihar travel business.
If your own desk — travel or otherwise — recognises itself in that fortnight list of chores, the practical takeaway is this: write your own list first, pick the one person with the most patience for systems, and put them on a time-boxed, level-by-level course rather than a pile of tutorials. The tools will still be there when your people are ready for them. Kiran’s crew would tell you the order matters, and they’d be right.
