Skip to content
ONE3.ai

How it works · Flow F4

Tickets triaged, answered from Knowledge, and closed without reopen

Email, WhatsApp, chat, call and portal requests become tickets. A triage agent classifies them. Answers come from Knowledge with citations, and gaps become new articles.

What the Issue to resolution flow does

Issue to Resolution is the support flow. It is built for the agency and SMB beachhead first, where most requests arrive on WhatsApp and email.

A request arrives on email, WhatsApp, chat, a call or the portal. Service Desk creates the ticket and identifies the customer through CRM. The triage agent sets category, priority, sentiment, language and SLA. If Knowledge can answer with high confidence and the skill has earned A3, the agent replies with citations and marks the ticket pending-customer. If confidence is lower, it drafts a reply for a human to approve (A2). If it cannot answer, the ticket routes to a skill-based queue.

Defects become linked issues in Work, and the customer is told when the fix ships. Every resolved ticket gets a CSAT survey. The Knowledge Curator then looks for gaps — questions Knowledge could not answer — and drafts new articles for review before they publish.

Helpdesk and client portal lite are planned for W1–W2. Issues & Sprints is a W3 module, so defect linking is lighter in Year 1.

Every step on this page is an Action: a typed verb with a risk tier (R0 read-only to R4 irreversible or financial), a permission check and a reversibility class. Agents and humans call the same Actions, so the same approval rules apply whether a person clicks a button or a Digital Employee acts.

Issue to resolution, step by step

Twelve steps. The key measure is resolution *without reopen*, not raw deflection.

  1. 1

    1. Request arrives

    **Service Desk · WhatsApp channel · email · chat · voice · portal.** Any channel starts the same flow.

  2. 2

    2. Create the ticket and identify the customer

    **Service Desk · Helpdesk.** Action `desk.ticket.create` opens the ticket and links it to the customer in CRM.

  3. 3

    3. Triage

    **AI Workforce · Support Triage & Reply Assistant.** Category, priority, sentiment, language and SLA are set.

  4. 4

    4. Can Knowledge answer with high confidence?

    **Intelligence · Knowledge.** Permission-aware retrieval looks for a grounded answer with citations.

  5. 5

    5a. Auto-reply with citations (A3)

    **Support agent.** Where the skill has earned A3, it replies with citations and marks the ticket pending-customer. Replies are external, risk tier R3.

  6. 6

    5b. Suggested reply for approval (A2)

    **Human gate.** Where confidence is lower or autonomy is not earned, a human approves or edits the draft.

  7. 7

    5c. Route to a skill-based queue

    **Service Desk.** Tickets Knowledge cannot answer route to the right human queue with a structured handoff note.

  8. 8

    6. Bug or defect?

    **Service Desk.** The agent or human flags defects.

  9. 9

    7. Create a linked issue

    **Work · Issues & Sprints.** A linked issue is created and the customer is notified when the fix ships.

  10. 10

    8. Human resolves

    **Service Desk.** Non-defect tickets that need judgement are resolved by a person.

  11. 11

    9. CSAT survey

    **Service Desk.** Every resolved ticket gets a CSAT survey.

  12. 12

    10. Knowledge gap detection and new article

    **Knowledge Curator agent.** Unanswered questions become article drafts. A human reviews them before they publish.

KPIs

ItemDetail
KPIsFirst response time, resolution time, auto-resolution rate *without reopen*, CSAT, cost per ticket

Questions about this flow

Will the AI answer customers without a human?

Only for questions Knowledge can answer with high confidence, and only after that skill has earned A3. Otherwise it drafts and a human approves.

What stops the AI from making things up?

Answers must come from Knowledge with citations and pass a groundedness check. One3 prefers "I don't know" over guessing.

Can I see why a ticket went to a person?

Yes. Handoffs carry a structured note: what was done, what is needed, evidence links, sentiment and SLA.

Which languages are planned?

English plus Hindi and Tamil in Year 1, each released only after its evals pass.

When is this flow planned?

Helpdesk is planned for W1, with client portal lite in W2. Linked issues in Work arrive with W3.

Bring your support queue to the pilot

Teams that handle support on WhatsApp and email help us measure resolution without reopen.

Become a design partner