AI Request Classification

AI request classification automatically determines the Category, Impact, Service Instance, and Team for new requests that arrive without a clear classification. This reduces misrouted tickets, accelerates response times, and ensures requests reach the right specialists without manual triage.

How it works

Classification runs on a new request, shortly after it is created, when:

• its category is Other, or it is a Case without a service instance, or the Virtual Agent created it without a request template;
• its status is Assigned;
• it has a Requested for person; and
• it is not a request group.

Requests that arrive with any other category are not classified. Sera AI first looks for a matching request template, then for the most similar past request, and then for a service. It posts a note on the request saying which template or request it used. It determines values for:

• Category: The type of request (Incident, RFC, RFI, Complaint, etc.)
• Impact: The urgency and scope of the request
• Service Instance: The specific service affected
• Team: The team responsible for handling the request

Classification happens within seconds of request submission. When a request is created through the Virtual Agent, the assigned category and service instance typically appear on the request card almost immediately.

Suggest vs. Auto-apply

In the Account Settings, administrators can choose how classification behaves. The setting is not offered on directory accounts.

• Suggest: Sera AI proposes values for Category, Impact, Service Instance, and Team. A specialist reviews and confirms before the values are applied.
• Auto-apply: Sera AI sets these field values directly. Specialists can still override them.

Both options apply to the same requests, described above.

Organizations can start with Suggest mode to build confidence in classification accuracy, then switch to Auto-apply when ready.

Triage and prioritization

The combination of fields that AI classification sets on arrival effectively automates triage. When Sera AI classifies a request and assigns Category, Impact, Service Instance, and Team within seconds, the request is triaged without a specialist touching it.

Impact drives priority. The Impact value assigned by Sera AI directly influences the request's priority and, by extension, the SLA targets that apply. A request classified as high impact is immediately subject to shorter response and resolution targets than one classified as low impact. This means the urgency of a request is established at the moment it enters the system, not when a specialist first opens it.

Team assignment eliminates routing delays. By identifying the responsible team automatically, classification removes one of the most common causes of slow triage: requests sitting in a general queue waiting for someone to read them and decide where they should go. The request lands directly in the right team's inbox.

Sentiment adds a prioritization signal. Within the team's inbox, Sera AI's sentiment analysis provides an additional layer. When a specialist has multiple requests with similar resolution targets, a negative sentiment indicator can help them decide which to address first. See AI Sentiment Analysis for details.

Automation rules extend triage logic. For organizations that want to go further, automation rules can act on classification results to implement custom triage workflows. For example, a rule could automatically escalate requests where Impact is high and the affected service is business-critical.

When to use automation rules vs. AI classification

Automation rules and AI classification serve different needs:

• Automation rules are best for standard, repeatable classification scenarios where you can define clear patterns (for example, "all requests containing 'password reset' should be categorized as Incident").
• AI classification adds the most value for less structured scenarios where the correct routing depends on understanding the meaning and context of the request content, not just matching keywords.

Both can be used together. For example, automation rules can handle well-defined patterns while AI classification catches everything else.

Using ai_similar_request in automation rules

The ai_similar_request expression is available for use in automation rules. It gives a rule the most similar past request, so the rule can copy values such as its service instance or team. See AI Similar Requests and Ticket Clustering for details.

Classification scope

Classification only runs on the requests described under How it works. The requester must be covered by a service instance; when no service covers them, the request stays unclassified. Requests that already have a specific category are left as they are.

‍