Service

Support triage and reply drafting

Tickets sorted and drafted before your team opens them. A person sends the reply.

Most support time goes on reading: working out what a message is about, finding the account, and deciding what needs doing today. Typing the reply is the quick part.

We automate the reading and preparation, and your team still sends. Each ticket arrives classified, prioritised and linked to the account, with a draft based on your help centre and past resolutions. Once the numbers show that routine cases need no edits, you can let those send automatically.

Scope this Growth

Channels: Email · Helpdesk · Web form · Chat transcript
Classification: Intent, product area, sentiment and urgency
Enrichment: Account, plan, orders and recent history
Drafting: Based on your help centre and past resolutions
Citations: Every draft names the sources it used
Default mode: Draft for approval. Auto-send is opt-in per intent
Escalation: Confidence and risk thresholds route to a person
Evaluation: Set of past tickets re-run on every change
Reporting: Volume by intent, deflection and edit rate

Specifications

How it runs

The stages of a typical support triage setup. Yours may skip some or add others.

Ingest
Tickets come in from your helpdesk or mailbox and are put into one format.
Classify
Intent, product area, urgency and sentiment are set against your own categories.
Enrich
The account, plan, recent orders and earlier tickets are attached before anything is drafted.
Retrieve
Matching help-centre articles and past resolutions are pulled in for the draft to work from.
Draft
A reply is written in your tone, citing its sources, with a confidence score.
Gate
Billing, legal, security and cancellation intents always go to a person, whatever the confidence.

Example output

An example of the format. We agree the exact fields with you before building anything.

Triage output, abbreviatedjson

{
  "ticket_id": "T-99214",
  "intent": "refund_request",
  "intent_confidence": 0.93,
  "product_area": "billing",
  "urgency": "high",
  "sentiment": "frustrated",
  "account": { "id": "acct_5512", "plan": "growth", "mrr": 490,
               "tenure_months": 26, "open_tickets": 2 },
  "risk_flags": ["churn_signal", "second_contact_same_issue"],
  "grounding": [
    { "source": "help/refund-policy", "relevance": 0.91 },
    { "source": "ticket/T-71880", "relevance": 0.84 }
  ],
  "draft": "…",
  "routing": { "decision": "human_required",
               "reason": "billing_intent_always_escalates" }
}

In detail

Always routed to a person

Refunds, chargebacks and billing disputes
Cancellation and downgrade requests
Legal threats or regulator mentions
Security and data-protection reports
Anything touching an account above your value threshold
Second contact on an unresolved issue
Sentiment below the agreed floor
Any intent the classifier is not confident about

What you get

This list goes into the scope document, and we check the finished work against it.

1Ingestion from your current helpdesk or mailbox
2Classification against your own list of intents
3Account and order history attached to every ticket
4Draft replies that cite your documentation
5Escalation rules for billing, legal, security and churn risk
6Test set built from your past tickets

Questions

Can it reply to customers on its own?

Only for intents where you switch that on, after reviewing the edit rate. Draft and approve is the default, and most teams keep it for anything involving money.

How do you stop it inventing policy?

Drafts are built from your documented answers and cite the ones they used. If nothing relevant is found, the ticket goes to a person with no draft.

Does this replace our helpdesk?

No. It works inside the helpdesk you already use.

How do we know it is working?

Watch the edit rate: how much an agent changes a draft before sending it. We report it per intent, so you can see which kinds of ticket the system handles well.

Also in Growth

Service