Preprint conference paper describing the use of email as a standard AI endpoint
Read the full paper on GitHub email_as_ai_endpoint.pdf
- We present a method to use email to chat and interact with AI, broadly
- We outline a three-layer architecture: email interface, webhook service, and AI orchestration
- The implementation can run fully local; enabling businesses to deploy in-house AI interactions through existing email systems
- We describe the benefits of email’s inherent characteristics of broad accessibility and asynchronous processing
- We provide a public demonstration; accessible by sending email to interface@integralbi.ai
- Users do not need to adapt to new interfaces or frequently switch between their inbox and AI chat
- Works across all email platforms and all AI model providers; local and API
- Asynchronous nature alleviates the burden of minimizing latency/time-to-first-token
- Asynchronous nature allows certain processing workloads (EG: detailed multi-agent report analysis, etc.) to be assigned a deadline to facilitate management of inference workloads
- And many others
We provide a public demonstration - accessible by sending email to interface@integralbi.ai