AI that resolves tickets,
not just answers questions
Connect your helpdesk. Kelu drafts replies inside Zendesk, Intercom, Freshdesk, Front, and Linear — grounded in your knowledge base. Deflect tickets before they reach an agent, auto-triage the ones that do, and test with simulation mode before going live.
Every layer of the support workflow
From deflection before a ticket is filed to copilot drafts inside your agents' queue
Ticket Deflection
Answer common questions as soon as they arrive. Kelu posts only when its confidence is at or above your threshold, and the ticket stays open for your team.
- Confidence threshold: only deflect when AI is sure
- Auto-reply to low-complexity tickets
- Fallback to human agent when confidence is below threshold
- Usage-based metering tracks resolved and deflected counts
- Customizable deflection message and escalation path
- Deflection rate on each helpdesk card
AI Copilot
Draft only mode puts confident answers in a private note for an agent to check and send. Zendesk agents can also ask Kelu from the sidebar.
- Draft replies in a sidebar in Zendesk (Intercom later)
- Cited from your knowledge base, not model memory
- One-click insert into the Zendesk reply box
- Agent can edit a draft, discard it, or ask for a new one
Works inside the tools your team already uses
No context-switching. Agents get AI drafts right where they work.
Zendesk
Native Zendesk app. Copilot drafts appear in Agent Workspace alongside the ticket.
Intercom
Replies or drafts on Intercom tickets and live Messenger chats.
Freshdesk
Replies, or private-note drafts, on new tickets in the groups you pick.
Front
AI drafts as internal comments inside the Front conversation thread.
Linear
AI replies on new Linear issues; the rest get your escalation label.
Test on real tickets before going live
Before enabling auto-reply in production, run simulation mode on up to 100 of your most recent resolved tickets. Kelu answers each one and shows you what it would have sent — so you can review quality, calibrate confidence thresholds, and approve the rollout with full confidence.
- Import historical tickets from any supported helpdesk
- Run AI answers against your current knowledge base
- Review each answer and its sources ticket by ticket
- Set confidence threshold before enabling live deflection
- Index solved Zendesk tickets as a knowledge source
Classify and tag tickets automatically
Kelu reads each new ticket and tags it with a category, a priority and topic tags — before a human agent opens it. Switch triage on for each helpdesk integration.
- Category and topic tags written to the ticket
- Priority classification: low, normal, high, urgent
- Runs on new tickets; live chats are not triaged
{
"ticket_id": "TKT-8821",
"subject": "API returns 429 on bulk import",
"triage": {
"category": "api",
"priority": "high",
"routing": "platform-team",
"tags": ["api", "rate-limits", "bulk"]
},
"draft_reply": {
"content": "Hi Sarah, bulk import...",
"citations": ["Rate Limiting Guide"]
}
}Rehearse first, then let it answer
Rehearse on past tickets
Run a simulation on up to 100 recent resolved tickets and see the answer Kelu would have given each one, with its confidence. Nothing is posted.
Set the confidence threshold
At or above it, Kelu replies, or leaves a private note in Draft only mode. Below it, nothing is posted.
Escalate what it can't answer
When the content is missing or confidence is low, Kelu posts nothing and adds the needs-review tag for your team.
Got questions?
Frequently Asked Questions
Common questions about Kelu Support Automation
Kelu integrates natively with Zendesk, Intercom, Freshdesk, Front, and Linear for tickets, and with Freshchat for live chat. Zendesk Messaging is coming soon. On tickets, Draft only mode leaves AI drafts as private notes inside the helpdesk, so agents don't switch context (not on Linear, where a note is a public comment). Zendesk agents also get a sidebar that drafts on request; an Intercom sidebar comes later. Jira Service Management and Salesforce can be indexed as knowledge sources.
When a user submits a ticket or contacts support via a chat widget, Kelu retrieves an answer from your knowledge base. If the confidence score exceeds your configured threshold, Kelu replies with the answer immediately and leaves the ticket open for your team. Tickets below the threshold get your escalation tag and route to a human agent as normal. You set the threshold — a higher threshold means fewer deflections but higher quality on what is deflected.
Simulation mode lets you run Kelu against up to 100 recent resolved tickets from your helpdesk before enabling live deflection. You see exactly what would have been auto-replied, review the draft answers, and calibrate the confidence threshold. Nothing is sent to customers during simulation. It is the recommended first step for any support automation deployment.
Add a Zendesk Tickets source and Kelu indexes your solved Zendesk tickets as structured knowledge documents, so an answer your team wrote once can answer the next person. Tag filters choose which tickets are added to the index. Tickets follow the knowledge base: in an internal knowledge base they stay private and answer only on the Ask page and in your bots. In a public knowledge base they are public. To keep tickets private, add them to an internal knowledge base.
Yes. The AI copilot experience is designed as draft assistance, not autonomous sending. Agents see the draft inline in their helpdesk, can edit any part of it, and choose to send or discard. Auto-reply (sending without agent review) is a separate mode you opt into explicitly, and only for tickets above your confidence threshold.
Ready to cut your ticket volume?
Connect your helpdesk and run a simulation on your most recent tickets in under an hour.
Free to start · No credit card required · Contact Sales →