← operator playbook
Apr 28, 20264 min read

Telegram-first ops: why we push every event to chat before the dashboard

Every payment, screenshot, and balance change surfaces in Telegram in real time. Here's why we designed the ops layer around chat instead of a dashboard, and what that unlocks.

operationsproductengineering

A dashboard is a pull interface. You open it, you scan for problems, you drill in. It is only useful when the operator remembers to look.

Chat is a push interface. Events surface where the team is already paying attention. For operations at high-risk operators — where a 15-minute delay on a flagged transaction can compound into a support incident — push wins.

What lands in the channel

Every operator gets three default channels:

  • ops-events: every state change on every intent. Verbose. Read by systems, not humans.
  • ops-review: transactions that failed at least one automation rule and need agent eyes. Read by agents actively on shift.
  • ops-alerts: incidents, rule-threshold breaches, upstream rail issues. Read by leadership.

Each event is a structured message with an intent ID, a compact summary, and a link that opens the transaction in the admin console. Agents can respond to the message with an action — approve, decline, escalate — and the response updates the transaction state without opening the console.

Why chat, not email or SMS

Email is asynchronous and unstructured. SMS is one-way. Chat is the only channel that supports low-latency, two-way, structured interaction natively. It also happens to be where operator teams already live, especially in international operations where the team spans time zones.

Telegram specifically because it has bot-level interactions that don't require enterprise integration, message threading that scales past a few thousand events per day, and end-to-end encryption for the operator-to-agent private channels.

What the dashboard is still for

The dashboard exists for aggregate views: weekly settlement statements, customer lifetime totals, historical dispute analysis. Anything the operator plans around, rather than reacts to.

The pattern is: chat for events, dashboard for state. Operators who invert this — dashboard for events, chat for occasional summaries — end up either missing the flag that mattered, or spending agent hours refreshing the dashboard. Neither is what the ops team should be doing.

// related reading