Pay for resolutions. Nothing else.
A conversation counts only when the customer confirms or goes quiet after an action. Handoffs to your team are free. Seats are unlimited.
Starter
For teams under 3,000 conversations a month. Chat and email, one help desk, the core 40 actions.
- Chat + email channels
- 40 pre-built actions
- Policy engine, 10 rules
- Shadow mode pilot
- Email support, 1 business day
Growth
Most teamsFor 3,000–50,000 conversations a month. Every channel, every action, custom actions via SDK.
- Chat, email, voice, in-product
- All 140 actions + SDK
- Unlimited policies, staged rollout
- Analytics + warehouse export
- Slack support, 2-hour response
Enterprise
For 50,000+ conversations, regulated industries, data residency and dedicated models.
- Everything in Growth
- EU / UK / US data residency
- Dedicated fine-tuned agent
- SSO, SCIM, audit log export
- Named engineer, 99.99% SLA
| Feature | Starter | Growth | Enterprise |
|---|---|---|---|
| Channels | |||
| Chat & email | Included | Included | Included |
| Voice | — | Included | Included + custom voices |
| In-product SDK | — | Included | Included |
| Actions & policy | |||
| Pre-built actions | 40 | 140 | 140 + bespoke |
| Custom actions (SDK) | — | Unlimited | Unlimited |
| Policy rules | 10 | Unlimited | Unlimited + red-team suite |
| Approval queues | 1 queue | Unlimited | Unlimited + routing |
| Trust | |||
| Data retention control | 30 days | Custom | Custom + residency |
| SSO / SCIM | — | SSO | SSO + SCIM |
| Uptime SLA | 99.9% | 99.95% | 99.99% |
Honest answers to the awkward ones.
A conversation where the agent took an action or gave an answer, and the customer either confirmed or did not reply within 72 hours. Reopened conversations are refunded automatically. Every week a 2% sample is re-checked by humans and the invoice is corrected downwards if we were wrong.
It hands off. You set the confidence threshold per intent. Handoffs include a summary, a suggested reply and the evidence, and they are free — we don’t get paid for work your team does.
No. Your transcripts and knowledge train your agent only, inside your tenant. Our model providers run under zero-retention agreements. You can export or delete everything at any time.
Two weeks. Week one: connect your help desk and knowledge, write policies, run shadow mode on last month’s tickets. Week two: live on one intent, then widen. Most teams are at 50%+ autonomous resolution by day 30.
Yes. Set a monthly cap; past it, Oriel keeps triaging and summarising for your team but stops autonomous resolution until the next cycle, or until you raise the cap.
Thirty minutes, your tickets.
Bring an export of last month’s conversations. We will run Oriel against them live and show you what it would have resolved, what it would have handed off, and why.
You see the product on your data, not a deck.
Pilots are month-to-month with a spend cap.