Guava's latest speech model uses a Minskian model of agents to perform language understanding, language generation, supervision, inter-system communication, planning, and question answering.
Published
Signal category
Research & Knowledge
Quote
“The latest speech model at Guava uses a Minskian model of agents, with different systems performing language understanding, language generation, supervision, inter-system communication, planning, question answering, and so on.”
— Anthony Scodary|Guava team
Company
Guava
The voice platform for regulated industries.
- Industry
- Software Development
- Location
- Los Angeles, US
- Company size
- 30 employees
Guava is the voice platform for regulated industries — where compliance and call resolution aren't optional. When the call is a prior-auth, a KYC check, or an FNOL, being wrong isn't an awkward moment — it's a compliance event. Guava is built for exactly those calls. Trained and tested on 10B+ minutes of regulated voice production over 13 years, it plugs into your CRM, EHR, and workflow engine through code, not prompts — so your systems stay the source of truth, and a supervisor agent catches hallucinations on the hard calls. SOC 2 Type II, HITRUST i1, and PCI DSS L1 — designed into the runtime, not bolted on. 100+ regulated deployments across healthcare, banking, insurance, BPO, and government. From kickoff to a live call in 14 days. Built for calls that have to be right.