Solutions · Support
Tier-1 deflection that doesn't tank CSAT.
I answer the easy questions in voice with citations from your docs. Your agents see the ticket only when the customer actually needs a human — and they walk in with the conversation already loaded.
- 24/7 first response
- Voice + screen-share
- Multilingual
- Cites docs, doesn't hallucinate
Trusted by support teams from SMB-volume help desks to enterprise white-glove operations
- Helix
- Northstar
- Atlas SaaS
- Forge
- Meridian
- Cipher
Where the queue gets stuck
Three moments your support team loses ground.
The problem isn't your agents. It's the work below their pay grade soaking up their day. I take that work.
Drag 01
The same five questions, all day, every day.
Reset password. Where's the invoice. How do I export. The repetition that wears down your best agents and drives the attrition number up. I take those questions — politely, with the right citation, in the customer's language.
~60% of Tier-1 tickets are repeat questions
Drag 02
Weekend backlog landing Monday at 9am.
Customers in Singapore at 4pm, Berlin at 2pm, São Paulo at 8pm — your agents are off. By Monday the queue is 200 tickets deep and CSAT slips because everyone's now waiting four hours for first response. I take that shift.
First-response time is the #1 CSAT correlate
Drag 03
A knowledge base nobody reads.
Your team has written 800 docs and yet the inbox is full of questions that are answered on page 1. Customers don't find them because search is bad, because they'd rather ask, because they don't know what to look for. I read every doc and find the answer for them.
Self-serve rate < 25% on most help centers
Where I sit in your support workflow
First touch to weekly digest — one voice, your knowledge.
First touch · pick up first
I answer before the queue forms.
I live on your help center, your in-product widget, the support phone line. The customer talks, I respond — in voice, in their language, in under a second. No ticket created if I can resolve it.
- Web widget or in-product surface
- Real-time pickup
- Multilingual conversation
Help-center widget · 0:08 elapsed
Anonymous · Berlin · 03:14 local
Customer
Hi — how do I export the audit log to CSV?
Tegan
Settings → Security → Audit log → Export. You'll see a CSV / NDJSON toggle on the top right. Want me to walk you through it?
Tegan · internal
Cited: docs/security/audit-log.md · § Export
Answer · cite, don't hallucinate
I quote your docs. With the link.
Every answer comes with the doc section I read it from. Customers see the source; your team sees what I'm grounded in. When I genuinely don't know, I say so — and route the question with the conversation context.
- Every answer cites a doc section
- No fabricated procedures
- Says "I don't know" cleanly when needed
Answer with citations
"Yes — paid invoices can be reissued from Billing → History → ⋯ → Reissue. The reissue keeps the original number with an -R suffix."
Grounded in
docs/billing/invoices.md
§ Reissue a paid invoice
changelog/2026-04-12
Invoicing v3 release notes
kb/faq/tax
VAT handling on EU invoices
Escalate · context attached
I hand your agent a record they can act on.
When the conversation needs a human, the session lands in your dashboard with audio, full transcript, the customer's actual words, captured fields, and a one-line summary of what I already tried. Your agent picks up where I left off — not from scratch.
- Session record in your dashboard, ready for review
- Audio + transcript + captured fields
- Per-session quality score on persona, KB usage, accuracy
Created from voice session
● routed to Tier-2
- Ticket
- Session #42188 · Priority high
- Customer
- Meridian Labs · Pro plan
- Issue
- OAuth callback failing intermittently
- I tried
- Walked through token refresh; symptom persists
- Sentiment
- Frustrated · 2nd contact this week
- Audio
- 06:12 · attached
Report · what your docs are missing
Every Monday, I tell you where your help center has gaps.
The top ten questions I answered well, the top ten I couldn't, the doc sections customers cited as confusing. Your docs team gets a roadmap; your CX leadership gets ammo for the next quarterly review.
- Top deflected vs. escalated questions
- Confusing-doc heatmap
- Suggested edits ranked by ticket impact
Weekly digest · Doc gaps
● 159 escalations
- Webhook retry policy47
- SAML group mapping39
- Data export format31
- Rate-limit thresholds24
- Tenant migration steps18
What changes for the support org
The numbers your VP Support actually cares about.
Tier-1 deflection
72%
Of inbound resolved without a human
First response
Live
Real-time voice pickup, day or night
CSAT delta
+4
Net points vs. pre-Tegan baseline
Agent attrition
-50%
Quarterly turnover after Tier-1 lift
Aggregated across private-beta partners between Apr–May 2026. Deflection and CSAT vary by product complexity and policy on hallucination guardrails.
“Agent attrition halved when Tegan took Tier-1. The team now does the work they trained for — the gnarly stuff with the customer who's been around three years. They actually like coming to work.”
Head of CX, late-stage SaaS
Open beta · 2026
What changes for your team
Agents on the deep stuff. I run Tier-1.
The work that was killing morale — password resets, where's-my-invoice, how-do-I-export — is the work I'm built for. Your agents do the deeper work that retained customers actually remember.
Support leadership questions
The questions before the next ops review.
Different question? .
Looking for the technical spec? Read the support-agent product page.
Will Tegan hallucinate?
I'm grounded in your docs and I cite every answer. When the docs don't cover the question, I say so plainly and route the ticket — I don't make up procedures. You can audit every answer against its cited source.
Does this replace agents?
No. Tier-1 deflection means your agents stop doing repetitive work — they don't stop existing. Teams I've worked with redeploy agents to Tier-2, deep customer-success work, or proactive outbound. Net staffing is usually flat with a much better mix of work.
How does my support team get notified?
Every escalation triggers an email to the address (or Google Group) you set, with the reason I escalated in the subject line — "Customer asked for refund", "Webhook signature mismatch", "Lost data after upgrade" — and the captured fields + a one-click link to the recording in the body. Everything also lands in your Tegan dashboard. Native write-back to Zendesk / Intercom / Front is on the roadmap.
What languages do you handle?
Multilingual voice — natural conversation across the major languages your customers speak. I follow the conversation in whichever language the customer's most comfortable in.
How is my customer's data handled?
Audio and transcripts are kept per session in your Tegan dashboard so you can review and replay. Data is wiped on request. Sign-in for end-users is via email-gate or anonymous mode. Security review for SOC 2 is on the roadmap as we move toward enterprise readiness.
What if a customer explicitly asks for a human?
I capture the request and the full conversation context in your dashboard with email + in-app alert to your support owner. They follow up with the context already loaded so the customer doesn't restart. (Live in-session takeover isn't shipped yet.)
Tier-1 in voice. Agents on the deep stuff.
Point me at your docs, your help desk, and your top deflection patterns. By Friday I'm answering Tier-1 in voice — and your agents are working the tickets they actually trained for.
Rollout — one week
- Day 1 · I crawl your help center + changelog
- Day 2 · I learn from your top deflected tickets
- Day 3 · Embed + escalation rules wired
- Day 4 · Guardrails + PII policy locked in
- Day 5 · Live on a slice of inbound chat
- Week 2 · A/B against your current first-response motion
- Week 4 · Reforecast deflection + CSAT + agent retention