Блог EN
2026-11-14 15:00 Telegram and Automation

Bot accepts requests but nobody replies: assign and monitor a responsible manager

A Telegram bot accepts a form, creates a request number and tells the customer “We received your enquiry.” But that does not establish that an employee has read it, accepted responsibility or responded. Lost enquiries usually occur after the conversational interface: in task queues, ownership rules, unavailable staff and unmonitored exceptions. Design the complete handover, not merely a menu of buttons.

Define the request states after submission

A practical workflow distinguishes received, waiting for assignment, assigned, in progress, waiting for customer, resolved and, if needed, escalated or reopened. Teams can choose their own names, but each state needs an exact meaning, responsible role and permitted transitions. An acknowledgement to the customer must say only that the request was registered if nobody has accepted it.

Give the queue an owner

Document working hours, competence groups, routing priorities and a backup person. Assignment may be automatic by topic, handled by a dispatcher or based on an agreed rota. An enquiry without an available owner belongs in a visible exception queue. Telegram bot and Mini App development can include a handover workflow, but the business must supply realistic employee availability and substitution rules. A bot cannot force an unavailable manager to respond.

Distinguish bot notification and human acceptance

A delivered push notification, an assigned CRM card and an actual first human response are three separate events. The manager-handover process should record when the human acknowledges a case and whether they can see its conversation history. Do not mark a ticket “in progress” just because a message was delivered to an employee's chat.

Technical events and audit trail

Telegram's Bot API can receive updates through webhooks or getUpdates; it does not include a ready-made business CRM queue. Plan idempotent processing so repeated update delivery cannot create a duplicate enquiry. Store the original update identifier, request ID, assignment/reassignment, acknowledgement, first substantive answer, escalation and closure, with appropriate protection of personal data. Operational monitoring must check whether requests actually reach the CRM and remain visible, not only whether the bot responds.

Filled hypothetical TASK-098: two incoming requests

Synthetic tickets T-101 and T-102 arrive at 10:00 and 10:03. A fictional internal rule requires assignment within 15 minutes; this is an example policy, not an agency SLA. T-101 is assigned to manager M1 at 10:04, while T-102 cannot find an owner.

  1. Create the two unique records with receipt timestamps and acknowledge only registration to customers.
  2. Save M1 as the owner of T-101; leave T-102 explicitly “waiting for assignment”.
  3. At 10:18, escalate T-102 to a dispatcher without closing or removing the record from the queue.
  4. After assignment, record the first meaningful human reply separately from the bot acknowledgement.
  5. Replay the same Telegram update; verify that no third ticket is created.

Positive acceptance: every case has a visible current state, ownership history and escalation when the internal rule is exceeded. Negative: the bot claims “a manager is working on it” even though T-102 is unassigned, or a repeated webhook creates an extra ticket.

Scope CRM integration and access rules

Specify categories, assignment and substitution rules, notification destinations, role permissions, queue views, event logs and failure tests. The exact CRM API and authentication methods must be checked separately; integration is not a built-in Telegram queue feature. A bot specification with scenario-based acceptance should state who may reassign, who may close and what users see during system errors.

Commercial boundaries for the automation

The published Telegram development offer covers business workflow analysis and possible integrations but does not automatically include specific duty rosters, deadline tracking, escalation rules or guaranteed human response. Agree on those as part of an individual technical specification and quote. For an estimate through agency contacts, provide the existing path of an enquiry, CRM, duty schedule, roles, replacement procedure, alert channels and anonymised failure examples.

Second filled scenario: H-98 and an absent manager

In a fictional consultation bot H-98, ticket H98-001 arrives at 09:00. The customer receives a registration acknowledgement, not a promise that a human is already replying. The internal model uses received, assigned, acknowledged, in_progress, resolved and reopened; these are design states, not fields guaranteed in a particular CRM.

  1. At 09:00 create H98-001 with an ID and submission time, then select the on-duty manager A according to the agreed rota.
  2. Store the assignment time and alert channel. Do not treat delivery of the alert as acceptance.
  3. If A has not accepted within a hypothetical 10-minute internal threshold, check staffing and reassign to standby manager B. This is not an agency response-time guarantee.
  4. B acknowledges the task, sends a substantive answer and records the outcome; keep the reassignment history.

Failure, access and reassignment acceptance

Positive: H98-001 has one active owner at a time, a traceable transfer and an understandable customer status. Negative: two people independently promise a response, or none accepts while the bot displays “specialist already working”. Test leave, weekends, unavailable CRM, repeated webhooks, erroneous closure and reopening. Give the customer a safe way to follow up when automation fails. Maintain separate permissions for private enquiry data and operational queue supervision.

The deliverable is a tested routing-and-handover system with real staffing responsibilities, not proof that every possible employee will always respond.