Prototype
Sable
AI systemAn internal knowledge assistant prototype built to answer a harder question than "does it work" — namely, how often is it wrong, and does it say so.
- AI engineering
- Retrieval
- Evaluation
- Interface design
PrototypeAI system · 2025
Atlas Voice
Where the call should end, and the human should begin.
A voice agent prototype for inbound appointment booking, built to test where automated calls should stop and a person should take over.
Overview
Status
Working prototype built to test an approach
Disciplines
- Conversation design
- AI engineering
- Telephony
- Integrations
Capabilities used
The challenge we set ourselves
Automated calling is easy to demonstrate and hard to trust. The interesting question is not whether a model can hold a conversation — it is what happens on the eleven percent of calls that do not follow the script.
We built Atlas to find the failure boundary deliberately, rather than discovering it in production.
Approach
StrategyDesignTechnology
01
Strategy
We scoped a single call type — inbound booking against real availability — and defined escalation conditions up front as rules, not as model judgement. Anything involving price disputes, accessibility needs or repeat callers hands over immediately.
02
Design
Conversation design treated as interface design: opening disclosure that it is an automated agent, explicit confirmation steps before anything is written, and a repair path when the caller is misheard twice.
03
Technology
A realtime speech model orchestrated by a Node service that owns the business logic. Calendar availability is checked against a real API, and every call ends as a structured outcome record rather than a transcript nobody reads.
Outcome
What the exploration produced
A working prototype, a documented escalation rule set, and a test-call suite covering interruption, silence, accent variation, background noise and deliberately out-of-scope requests. We use the same suite to evaluate client systems before go-live.
- Conversation design
- Escalation rule set
- Voice agent prototype
- Structured test-call suite
Why there are no numbers here. This is an internal prototype, not a client engagement. We do not publish conversion rates, revenue or satisfaction figures that no real deployment produced.
Capabilities behind this project.
- Custom AI CallingVoice agents that handle real inbound and outbound business calls — and know when to hand over.
- AI IntegrationAI applied to the specific workflows where it creates measurable leverage.
- AutomationRemoving the manual steps between your systems — reliably, and with somewhere to look when it fails.
More work
Start here
Build something like this.
If this is close to the problem you have, tell us about it. We will be specific about how we would approach yours.