Automated triage & ticket generation for Topdesk support emails

The Challenge: Handling high-volume & inconsistent support emails

Handling a high volume of support emails with inconsistent structure and mixed content (bugs, queries, updates, etc.). Manual triage was time-consuming, error-prone, and lacked consistency. Ticket creation delays caused slower response times and poor tracking.

Objective:

Topdesk aimed to achieve several key objectives using this Agentic automation implementation:

Automate email classification (bug vs. non-bug)
Reduce manual triage and decision-making burden
Improve consistency in issue identification and response
Create a feedback-driven system that learns over time
Seamlessly integrate with Lienion for ticket creation
Provide auto-generated, reviewable email responses

TalosNation's solution: Agentic automation for email classification

An Agentic Automation crew processes incoming support emails. Agents fetch, classify, and route emails, create tickets in Ilenion, and generate email replies. The crew is coordinated by a Knowledge Hub and refined through human-in-the-loop feedback.

Technologies:

TalosNation Agentic Framework (The Fraim)
Large Language Models (LLMs) for classification & response generation
Prompt library for structured reasoning and task routing
Knowledge Hub integrated with support SOPs and prior ticket data
Custom dashboard tabs for finetuning, validation, and human-in-the-loop feedback
Integration with Lienion (for automated ticket logging)
Email Monitoring APIs (IMAP)

Technical Details:

Agents Roles

Email Fetcher Agent: Pulled new emails periodically
Bug Verification Agent: Analyzed content for issue classification
Ticket Creation Agent: Created actionable tickets in Lienion
Auto Reply Agent: Generated templated email responses using context

Feedback/FineTune

Two-level human review system (Tab 1 & Tab 2)
All feedback logged and used to finetune prompt logic and model accuracy

Knowledge Sources

Ticket format rules, bug classification criteria, prior examples

Task Routing

Controlled via prompts and validation phases
Crew coordination managed through finetuning dashboard

Process: TalosNation's Agentic crew workflow in action

Inbox Monitoring:

Performed by: Email Fetcher Agent
Monitors the support inbox and retrieves all new incoming emails in real time for further processing.
Content Classification:

Performed by: Bug Verification Agent
Analyzes the email content to determine whether it’s a valid bug report or a non-bug message. Classification is based on prompt-driven logic and knowledge hub rules.
Finetuning Tabs

Coordinated by: Crew Orchestrator Logic
Routes emails to the appropriate validation queue:

Tab 1: Bug Emails, Tab 2: Non-Bug Emails
Allows for efficient human-in-the-loop validation and performance feedback tracking.
Human Review:

Supported by: Validation Interface
Human reviewers verify the agent’s classification decision, make corrections if necessary, and provide structured feedback used to improve agent performance over time.
Ticket Creation:

Performed by: Ticket Creation Agent
Once a bug email is approved by a human reviewer, this agent creates a task in the Lienion system, assigns it to the correct team, and tags it for tracking.
Auto Reply Generation

Performed by: Auto Reply Agent
Generates a contextual email response using structured prompt templates, referencing the ticket details. The response is queued for human review.
Final Review & Dispatch

Reviewed via: Response Validation Interface
A human reviewer checks the auto-generated email for clarity and accuracy. If approved, the response is sent; otherwise, it's manually edited and then dispatched

Results: TalosNation's Agentic automation success metrics

As a result of implementing the Agentic Automation crew, the following measurable outcomes were achieved