June 2026 Newsletter

Dear VIVI Community, April was voice. Two weeks ago was the Operator Issue. This one is the infrastructure underneath.

We shipped a wave of things over the last two weeks that, taken together, change what VIVI is for. Agents that cite their sources and link straight to the document. Agents that return structured data your systems can act on. An API channel that lets your other agents call into VIVI as a tool. A conversation tracing layer that makes every step of an agent’s thinking visible. And SMS, finally live, with consent middleware that actually works.

The shape of the platform changed. Plus a case study from one of the more interesting things our customers have built on top of VIVI this year.

Feel free to reply directly with feedback. We read every one.

Citations have been live since March, and they remain one of the most useful things you can flip on in your agent. When a VIVI agent answers using your knowledge base, the citation now appears inline as a numbered reference. Click it and you go straight to the source document, web page, or SharePoint site. Trust without leaving the conversation.

This is the difference between an AI agent that is plausible and one that is defensible. For hospitality, a concierge agent’s recommendations are auditable. For regulated industries (legal, finance, healthcare), it is the difference between deploying AI in customer-facing roles and keeping it on the sidelines.

Your agent can now return typed JSON instead of free-form text. Define a response schema and every reply conforms to it, every time. No parsing gymnastics on the other side.

This is a feature for builders. Plug the JSON into your own UI and the cards-in-chat experiences become easy to build: menu pickers, room-upgrade comparisons, tee-time selection grids, structured order receipts. Feed the JSON into your CRM, your booking flow, or a backend workflow and the data lands clean.

Designed for the API channel’s /invoke and /batch endpoints. Not recommended for conversational channels where a natural, human-readable response is expected, but exactly right for the moments where your agent’s output needs to be machine-readable.

Define your schema three ways: visually field by field, in raw JSON Schema, or by pasting a sample JSON document. Strict mode is always on. Schema size capped at 5 KB.

When a law firm gets handed 80,000 emails from opposing counsel, every one of them has to be reviewed. Traditionally that is done by associates billing $400 an hour, page by page.

LawGPT, built on VIVI, changes the shape of that work.

A litigator uploads the document corpus. They ask a question in plain English: “Find every email where Davis discussed the Q3 supplier change.” LawGPT searches the indexed documents, returns the relevant passages, and links every claim back to the original document with citation-grade precision. The lawyer verifies in one click instead of spending six hours hunting.

Citations are what make this defensible. Structured outputs are what make it usable. Together they turn weeks of paralegal work into an afternoon, with the audit trail courts require.

Every VIVI agent can now have an API channel. Generate a channel ID and an API key, and any external system can fire messages at the agent over HTTPS.

That means your CRM. Your custom backend. Your booking flow. And the part that gets interesting: your other AI agents. Claude. ChatGPT. The custom assistant your team built last quarter. They can all now reach a VIVI agent as a tool.

VIVI agents are no longer a destination. They are a tool your other agents can reach for.

Conversation Tracing gives you a step-by-step view of how your agent processes and responds to each message. Trace logs show you the model calls, tool calls, knowledge retrievals, and intermediate reasoning that go into each response. Viewable directly in the portal. Exportable as JSON for deeper analysis.

For voice agents, each turn gets its own trace, so you can pinpoint exactly where things went right (or wrong) in a real-time interaction.

This is the observability layer the platform always needed. Especially valuable when you are debugging an edge case, reviewing a customer escalation, or proving (to yourself or to a regulator) exactly what an agent did and why.

SMS finally shipped. Provision a local or toll-free number directly through VIVI. The bigger story is what comes built in.

VIVI ships an automated consent layer on every SMS channel. Before any agent message lands, an opt-in confirmation goes out. Recipients reply YES (or Y, OK, OKAY, START, SUBSCRIBE) and the queued message is delivered. STOP, HELP, and START keywords are handled automatically per CTIA rules. Toll-free verification is integrated.

You stay compliant without writing a line of compliance code.

The shape of VIVI changed this past month. Agents that cite and link. Agents that return structured data. Agents your other agents can call. A tracing layer that makes every step visible. A new channel with the compliance baked in. And LawGPT, a customer who built something serious on top of all of it.

To see what is possible at your property, in voice, chat, SMS, or your other agents’ calls, schedule time with our team.