Skip to content
Early access: Ray will answer from the knowledge each business publishes. Request access
Ray
Sign inTalk to salesSee demoBook a demo

Ray in the consoleComing soon

Set up Ray by talking to Ray

An assistant inside the console will help each business’s team set up agents, organize knowledge and write test cases. Ray will propose each change; a person will review and approve it, and nothing will reach customers without passing the test cases.

Request accessSee a change proposal

Early access with hands-on help from the team. Ray is in development.

Illustrative example of the Ray in the console panel: an administrator asks for an agent for the Lumora Hogar brand with a Kind tone, two collections and a sensitive topic; Ray is thinking, with three steps done and one in progress.

Illustrative example with sample data. Lumora is a fictional brand.

What it will do

Requests in plain language, changes that can be reviewed

Ray in the console will work on the same things the team already sets up by hand: agents, knowledge, sensitive topics and test cases. Every request will end in a proposal, never in a published change.

  • Put together an agent

    An agent for a brand or a site, with its persona names, its greeting and its instructions, as a draft.

  • Adjust the tone

    Formal, Kind or Friendly, plus the business's own instructions. Ray will show how a reply changes before anything is applied.

  • Choose the knowledge

    Which collections each agent reads: the whole library or only the ones that belong to that brand.

  • Add sensitive topics

    The business's own topics that will always go to a person, on top of the base list, which cannot be removed.

  • Write test cases

    Cases with real questions and the expected outcome, proposed for the team to review before adding them.

  • Explain a report

    What is behind a number or a content gap, with the conversations that make it up.

Change proposal

Ray will propose; a person will decide

Every request will end in a list of changes that is reviewed line by line and applied as a draft. Whatever falls outside the role of the person asking, Ray will flag and leave out.

Illustrative example of the console: on the left, the AI Agents list with two published agents and one proposed by Ray, not applied; on the right, the administrator's request and Ray's proposal: it adds the Lumora Hogar agent, changes the tone from Formal to Kind, adds two collections, a sensitive topic and three test cases, and leaves out publishing, which a person with the Editor role does. Buttons to apply as a draft, review one by one or discard.

Illustrative example with sample data. Lumora is a fictional brand.

Before it goes out to customers

If it gets worse, it will not go out

An applied change will stay a draft. Before it is published, it will go through the business's test cases, just like a change made by hand.

  1. Ray will propose the changes
  2. A person will review them and apply them as a draft
  3. The test cases and the sensitive-topics floor will run
  4. A person with the Editor role will publish, or the change will be put on hold

Illustrative example of the test results of the Lumora Hogar agent: the business's cases pass, the sensitive-topics floor is complete, the three cases proposed by Ray were added after a review and the change is ready for a person with the Editor role to publish.

Illustrative example with sample data.

Roles

Ray respects the role of the person asking

Ray in the console will have the same permissions as the person using it, never more. Assigning the account-data lookup tools will stay with the Administrator.

  • Administrator

    Will be able to set up and archive agents, pick the default one and assign sites and tools, including the account-data lookup ones.

    With Ray in the consolePropose a new agent, its site and its tools.

  • Editor

    Will be able to edit each agent's instructions, tone and knowledge, and publish.

    With Ray in the consolePropose instructions, tone, collections, sensitive topics and test cases.

  • Agent

    Will be able to read the setup and try questions.

    With Ray in the consoleWalk through a setup and try questions, without proposing changes.
Ray in the console will be able to
  • Propose changes as a draft, with every line in view
  • Try questions against a draft before applying it
  • Walk through why a reply came out the way it did, with its source
Ray in the console will never
  • Publish a change or record a test exception
  • Do anything the role of the person asking does not allow
  • See, show or store the business's keys and secrets

Reports and gaps

From the number to the missing article

Ray will be able to walk through a report or a content gap with the conversations behind it, and draft the missing article, for a person to review and publish.

  • Explain a number

    Which conversations make up a number and what changed since the week before.

  • Draft the missing article

    A draft in Knowledge, put together from what customers asked. It is never published on its own.

  • Test it before citing it

    The published article will go through the test cases before Ray can cite it.

Illustrative example: Ray in the console walks through the “Shipping to P.O. boxes” gap and leaves a draft of the article in Knowledge, with buttons to edit or publish it.

Illustrative example with sample data. Lumora is a fictional brand.

Set up Ray with the team that builds it

A short conversation with the team that builds Ray to review the company’s case: frequent questions, existing knowledge, sensitive topics and a migration plan.

Request accessTalk to sales

Keep exploring