Guide

Public Authority Helpdesk in OTAI Studio: Queues, Priorities, and a Narrow Public Chatbot

Set up OTAI Studio for a public authority: queue and priority Agents, run conditions, and a public chatbot limited to hours and how to file a request.

#government #otai-studio #queue-classification #chatbot
Public Authority Helpdesk in OTAI Studio: Queues, Priorities, and a Narrow Public Chatbot

This setup is for a public authority that runs its own service desk: badge requests, facility faults, workplace IT, procurement intake, and a registry of written requests. It is not a layout for a consultancy. Personal and case data stay on the authority’s infrastructure (Full On-Prem). OTAI Studio classifies tickets and answers from published FAQs. It does not decide cases or close them.

KIX is a supported ticket system; there the product is KIX Studio, and KIX has no public FAQ area, so a public chatbot there uses extra knowledge or customer-portal FAQs.

Agents to create

Open Agents. Sync values from the ticket system and describe them. Classification is zero-shot. There is no training on archived cases.

Queue classification on OTOBO or Znuny:

QueueWhen it applies
Access and badgesNew badge, replacement, visitor escort requests
FacilitiesBuilding faults, rooms, keys
Workplace ITAccounts, devices, line-of-business login faults
Procurement intakePurchase requests that only need a file number
Written requestsFormal inbound that must be registered, not answered by the chatbot
Records officeFile retrieval for staff, under the authority’s own rules

On Zammad, classify Groups. Zammad does not use queues.

Priority classification. Describe operational impact:

  • Highest: the case-file or registry system is unreachable during office hours
  • High: a badge or access fault that blocks someone who is already on site today
  • Normal: a procurement or records request inside the usual time
  • Low: a question about published opening hours

Minimum confidence: Normal (80%). Under that, flag for human review. Do not set a fallback that marks a request as granted. Studio also has a demo, Example priority classification: administration, on a sample DynamicField. Copy the description style, then use the live priority Agent. See Classification agents.

Summarization into a DynamicField for the registry handoff. See Summary agents.

Reply draft only for “your request is registered under this number.” Formal style. A clerk sends it. See Reply draft agents.

Priority classification Agent

Run conditions

On each ticket Agent, Run conditions:

  1. Ticket created
  2. State new
  3. Target field empty
  4. Activate

A second agency with its own ticket system is a second Full On-Prem installation, with separate Agents and Chatbots.

What the public chatbot may say

Chatbots → Public. Search Public FAQs only. Matching source only.

Publish a short set in the ticket-system FAQ (FAQs):

  • Office hours and the public entrance
  • How to submit a formal request (address, form, what to attach)
  • Visitor registration steps that are already public
  • Which phone or mailbox receives which topic

Keep internal procedures, file plans, and staff directories out of the public area. The public chatbot cannot create a ticket. The welcome text should say the chat does not process the request.

Public chatbot

Customer and internal

Customer, if the authority has a signed-in portal. Search Customer FAQs and allow Create ticket for a request the FAQ does not cover. See the customer chatbot.

Internal for staff (AI Chat). Search Internal FAQs: which queue registers a written request versus a facility fault. See the agent chatbot.

A Free Cloud Trial is evaluation only. Production is Full On-Prem, including OTOBO, Znuny, and the framing on the government solution page.