# Decagon agents by Decagon

Canonical page: https://technooptimist.io/ai/decagon-agents
Frontier: [AI](https://technooptimist.io/ai.md) · Field: Agents · Company: [Decagon](https://technooptimist.io/companies/decagon.md)
Stage: Scaling (5 of 5: Research → Proto → Pilot → Deploy → Scaling)
Updated: last update 2026-10-01, facts checked 2026-10-11
Goal: Replace hold queues with an AI concierge that solves most customer problems instantly in any channel and language, while the company keeps control of policy and tone.
Status: Voice 3 and Chord speech model launched Oct 2026; self-improving Duet Autopilot since June.

## Milestones

| Status | Milestone | Date | Slip |
| --- | --- | --- | --- |
| Done | Company founded | Aug 2023 |   |
| Done | $250M Series D at $4.5B | 27 Jan 2026 |   |
| Done | Duet Autopilot: agents that propose their own fixes | 9 Jun 2026 |   |
| Done | Voice 3 and Chord, its own speech model | 1 Oct 2026 |   |

## Most important updates

- **1 Oct 2026**: Decagon launches Voice 3 and Chord, its own speech model for customer calls. Voice 3 runs two models side by side: a fast one that listens and talks, and a stronger one that reasons, calls tools and enforces rules. ([our story](https://technooptimist.io/stories/decagon-launches-voice-3-and-chord-its-own-speech-model-for-customer-5c9d8de0.md) · [original source](https://decagon.ai/blog/voice-3))
- **9 Jun 2026**: Decagon launches Duet Autopilot, agents that find and propose fixes to themselves. Duet Autopilot reads production conversations, finds the most valuable problems, such as needless hand-offs to humans, and drafts targeted fixes. ([our story](https://technooptimist.io/stories/decagon-launches-duet-autopilot-agents-that-find-and-propose-fixes-to-ec9023e5.md) · [original source](https://decagon.ai/blog/autopilot))
- **27 Jan 2026**: Decagon raises $250M Series D at a $4.5B valuation. Decagon raised $250 million led by Coatue and Index Ventures, tripling its valuation to $4.5 billion in under six months. ([our story](https://technooptimist.io/stories/decagon-raises-250m-series-d-at-a-4-5b-valuation-4236bc3f.md) · [original source](https://decagon.ai/blog/series-d-announcement))

## Current obstacles

- **Head-to-head with Sierra**: Sierra and large software vendors chase the same enterprise contracts, so features and prices are under constant pressure.
- **Edge cases in regulated work**: Banks and telecoms need agents that never break policy; rare cases still go to humans.

## Physics limits

- **Language models are not deterministic**: The same question can get different answers, so exact rules must live in code or checks, not only in the model's instructions.
- **Voice latency floor**: Hearing, understanding, deciding and speaking each add delay; parallel models hide some of it but cannot remove it.

## How it works

### Procedures: Write the playbook in words

Support teams describe each process, such as a refund, in plain language plus code for exact rules; the agent follows it.

### Act: Connected to company systems

Integrations let the agent look up orders, change bookings or issue refunds instead of only answering questions.

### Speak: Voice that keeps talking

A quick conversational model listens and replies while a stronger model works on the request in parallel.

### Improve: Learn from live conversations

Autopilot finds failures, drafts fixes, tests them on a set of past chats and waits for human approval.

## Spec sheet

| Spec | Decagon agents | Provenance |
| --- | --- | --- |
| Channels | Chat, email, voice, SMS | reported |
| Voice languages | 70+ | reported |
| Listeners who spotted the human | About 10% (blind test) | reported |
| New enterprise customers, past year | 100+ (to Jan 2026) | reported |

Provenance: reported = stated by the company; estimated = our estimate; sample = a sample figure.

## About Decagon

Decagon builds AI agents that answer customers by chat, email and phone and take actions like refunds, for companies such as Chime, Avis and Deutsche Telekom. Sierra's main rival.

- Valued at $4.5B in Jan 2026, triple its value six months earlier; added 100+ enterprise customers in a year.
- Agents follow Agent Operating Procedures: plain-language instructions with code where precision matters.
- Voice 3 (Oct 2026) splits listening/speaking and reasoning into two parallel models to cut lag.

- Founded: 2023
- Headquarters: US
- Status: private
- Valuation: $4.5B (private valuation, 2026-01)
- Raised: $250M
- Last round: Series D $250M, 27 Jan 2026, led by Coatue Management, Index Ventures
- Website: https://decagon.ai
- People: Jesse Zhang (Co-founder & CEO); Ashwin Sreenivas (Co-founder & CTO)
- Partners, customers, investors: Chime (customer); Deutsche Telekom (customer); Avis Budget Group (customer); Andreessen Horowitz (investor)

Company page: https://technooptimist.io/companies/decagon.md

## Related programs

Everything in AI: https://technooptimist.io/ai.md

---

The Techno Optimist tracks every frontier of technology, program by program: https://technooptimist.io/ (a Markdown version of any page: add .md to its URL; site map for agents: https://technooptimist.io/llms.txt).
Data API (JSON, realtime stream): https://technooptimist.io/developers. People read 5 pages free, then Reader $10/month; these Markdown pages are open to agents.
