Questions in one channel, answered or routed
Colleagues' questions used to arrive by email, Slack and Teams, and each one needed back-and-forth before anyone could help. Now one Slack channel reads every question, answers the common ones, looks things up when needed, and opens a ticket for a person when it has to.
Three inboxes, one small team
Every question landed wherever the asker happened to be.
Day-to-day operational questions reached the team as an email, a Slack message or a Teams chat. Whoever saw it first had to ask for the missing details, wait, ask again, and only then start investigating.
The same kind of question came back again and again, and each time somebody typed the answer again. When a case needed a ticket, the context was copied over by hand.
The team wanted one place to ask, where a question gets dealt with.
One channel, three helpers, three outcomes
Post the question once. The channel sorts it.
A reader bot reads every message in the channel, no tagging needed. If it is a known question, it answers straight away from an prepared answers the team keeps for recurring issues.
If the question needs looking up, it calls a second bot that searches internal documents and writes a clear reply. If a person has to step in, a third bot opens a Jira ticket with the original message, what was found and a suggested next step, and tags the right teammate.
The bots fetch and type. People judge.
No bot closes a case on its own. When a question needs real work — a fix, a decision, a call to someone — a ticket lands with a teammate who investigates and resolves it. I prepared the first set of answers; the team can add an agreed answer once, and the first bot reuses it instead of a person retyping it.
Bookings for one of my hotels dropped since last week — did anything change on our side?
Known issue — the same drop was raised last week and a fix is in progress. Nothing to change on your side; I'll post here when it's done.
answered · from the prepared answersA booking shows the wrong cancellation policy. Can someone fix it before the guest calls?
This needs a person. @Ticket bot, open a ticket with the details.
Ticket created and @Priya tagged — she handles booking issues.
- Context: Ana's message and the booking details
- Found so far: policy text differs from the hotel's current one
- Suggested next step: confirm with the hotel, then correct the booking
Gone: the grunt work
Repetition
Typing the same answer for every new asker. Recurring issues get one agreed answer, reused automatically.
prepared answersWaiting
Three inboxes and a round of "which hotel?" before anything started. One channel, read the moment a message lands.
reader botCopy-paste
Writing up a ticket by hand. The ticket arrives with the message, the findings and a suggested next step.
Jira handoffListen, map, try, wire, fill
Listened first
I interviewed every team member: what do you handle, which cases should reach you? Their answers became the routing rules.
Mapped the flow
With my manager I turned the team's needs into a workflow map: when to answer, when to look something up, when to ask the asker for more, when to open a ticket.
Tried four versions
Each version was shaped by what the one before could not do. The fourth — three existing bots working together — is the one in use.
Wired the three
I wrote the behaviour rules: the reader bot reads without being tagged and decides which helper to call; the research bot answers in a clean format; the ticket bot connects to the team's Jira.
Filled the prepared answers
Prepared answers for recurring issues — what is known, whether a fix exists, what to do next — so the first bot can reuse them instead of a person retyping.
Honest credit
My part
- Interviewed each team member and turned that into routing rules
- Designed the whole flow: intake, asking back, looking up, escalating
- Chose and combined the three existing bots; wrote their behaviour logic
- Designed the Jira handoff: what a ticket must contain before a person sees it
- Prepared the answer-bank entries for recurring issues
Not my part
- The AI bots themselves — existing AI tools supplied by Agoda
- The idea of one team channel — my manager's requirement
- Platform choices — Agoda's guidance was to reuse existing bots
- Training or building any AI model — none was built