Tegan · Support agent
I answer the questions, not the tickets.
I'm Tegan in support mode. Drop me into your help center and I handle Tier-1 in real voice conversations — answer questions from your docs, resolve issues, and escalate cleanly when I can't.
- 24/7 coverage
- Real-time turn-taking
- Multilingual
- Cites your docs
In open beta with SaaS teams handling support volumes from a few hundred conversations a week to several thousand a day
- Helix
- Northstar
- Atlas SaaS
- Forge
- Meridian
- Cipher
What I do for support
Four ways I take the load off your support team.
Knowledge
I read your docs so customers don't have to.
Point me at your help center, marketing site, YouTube tutorials, and any notes you want to paste in. I parse the lot in 5–15 minutes, answer in your tone, and cite the source on every reply so you can verify.
- Help center + marketing-site crawl
- YouTube transcripts become searchable knowledge
- Pasted notes outrank crawled content when they conflict
Trained on
Help center
187 articles · re-indexed daily
Changelog
Synced on publish
Internal runbooks
Confluence · 4 spaces
Past tickets
Read-only, learned from
Top-50 known issues
Curated by your team
Escalation
When I can't close it, I email you with the reason.
Every escalation lands in your inbox — subject line in plain English ("Customer asked for refund", "Webhook signature mismatch in production"), body with the captured fields, my hypothesis, and a one-click link to the recording. Route by topic, sentiment, or plan tier. The dashboard mirrors everything for the team that prefers to work from a queue.
- Email subject = the reason I escalated, in plain English
- Body has the transcript snippet, captured fields, and recording link
- Routing rules — by topic, sentiment, plan tier, or matched phrase
Inbox · escalations@yourco.com
new
Why I escalated
Customer's signed payload keeps failing my re-validated secret. I've ruled out their code; this looks like edge-cache mutation. Beyond Tier-1.
- Customer
- Pro plan · 4 months
- Mood
- Calm, technical, has workaround
- Suggested owner
- Platform · @kira
- Priority
- P2 · workaround exists
Insight
I learn your gaps and tell you what's missing.
Sessions get scored automatically on knowledge usage and helpfulness. When I struggle with a topic repeatedly, it shows up in your review queue — your docs improve because I told you where they fall short.
- Auto-quality scoring on every session
- User ratings (1–5 stars) captured post-call
- Replay any session — audio + screen video + transcript
This week's digest
Mon 09:00
Topics your docs are missing
SAML SSO setup fails on Okta
27 calls this week · ↑ 4×
- new
Where is the audit log export?
19 calls · no doc exists
Billing — proration on plan change
12 calls · ↑ 2×
- new
Webhook retries — new spike
8 calls · last 48h
Coverage
I'm awake when your team isn't.
Real-time voice, multilingual, unlimited concurrent conversations. The customer from Singapore at 4 a.m., the one from Berlin at 2 p.m. — all get the same depth and the same source-cited answers.
- 24/7 across every timezone
- Multilingual conversations
- Unlimited concurrent sessions — no queue
Tokyo
03:14 JST · JA · EN
Berlin
09:14 CET · DE · EN
Cape Town
10:14 SAST · EN · AF
Austin
21:14 CST · EN · ES
Numbers from beta
What changes when customers can talk to your docs.
Tier-1 resolved
73%
of conversations closed without a human
Setup time
< 15 min
from URL to live agent
Coverage
24/7
Every timezone, no on-call
Concurrent calls
∞
No queue, no hold music
Aggregated across private-beta partners between Apr–May 2026. Each tenant's deflection criteria differs; numbers reflect conversations Tegan closed with no human follow-up required.
“Our Tier-1 queue used to eat 60% of an agent's day. Tegan closes the easy ones in voice and the team actually has time to dig into the hard ones.”
Support leader, mid-market SaaS
Open beta · 2026
Why teams switch
The customer doesn't want a ticket. They want the answer.
Form-based help centers train customers to wait. Voice changes the expectation — they ask, I answer, the conversation ends. The escalations that DO reach your team are the ones that should.
Support-mode questions
The first things support leaders ask.
Or just .
What if you don't know the answer?
I say so. Then I escalate with full context — the conversation, my hypothesis, what I tried, and the customer's tone. The agent who picks it up starts in turn 12, not turn 1.
Can you handle account-specific or auth'd questions?
I authenticate end-users via email-gate or anonymous mode today. Live account-data lookups (invoice status, usage, feature flags) aren't wired yet — they're on the roadmap. For now I answer from your docs and route account-specific asks with full context.
How do you stay accurate as our docs change?
Re-crawl is one click in your dashboard — typical re-index runs in under 15 minutes. When I see a customer asking about something my current docs don't cover, that gap shows up in your review queue automatically.
What does the customer hear when I can't be reached?
Plain text fallback: a brief 'I'm having trouble reaching the system' message and a form to leave details. The form posts captured context to your queue so nothing is lost.
Do you record calls?
Configurable per agent. Default: audio retained only if your retention policy allows. Transcripts always saved against the ticket. End users are told before the call starts.
How is this different from a chatbot or AI helpdesk?
Voice changes the qualification. Customers describe the actual problem in their own words; I can ask clarifying questions in real time, walk them through a setting, handle interruptions. A chat widget is a worse email.
Let me carry Tier-1.
Point me at your docs, set your routing rules, and ship me a sandbox tenant. Your support team gets their afternoon back by the end of the day.
What you'll set up
- Point me at your help center + YouTube
- Paste in any internal notes that should outrank docs
- Set the escalation review rules
- Configure tone + persona to match your team
- Go live in 5–15 minutes