Skip to main content

Orchestra Overview

Learn what Orchestra is, how it fits with Cue and Fountain's other AI tools, and how to build, run and share custom agents.

Orchestra is where you build your own AI agents in Fountain. Describe a task or a process in plain words, and Orchestra turns it into an agent that owns that work: it knows what to do, when to run, what it may use, and when to ask a person first. Once it is on, the agent runs on demand, on a schedule, or when something happens in your account.

Limited Feature

This is currently a limited feature. Please contact your Fountain representative for more information.


How Orchestra Fits With Fountain AI

Each of Fountain's AI tools does a different job.

Tool

What it does

Use it when

Cue

Your AI assistant. You ask, it answers or carries out the task with you, one conversation at a time.

You want an answer, an analysis or a change made now.

Cue scheduled tasks

Runs one prompt on a daily, weekly or monthly cadence and saves the result as a conversation.

You want the same report or check repeated.

Orchestra

Builds agents that own a piece of work. Every run sets a goal, follows a plan and reports whether it reached the goal.

You want work handled end to end, without starting it yourself each time.

Talent Agents

Fountain's AI voice agents, such as Sam, that call workers and applicants.

You want structured conversations by phone at scale.

Automations

A fixed rule: when one event happens, take one action.

The same step always follows the same event.

Orchestra is built on Cue, so you reach it from the Cue menu and its agents work with the same Fountain data and the same permissions. The Talent Agents in your account appear in the Orchestra library next to the agents your team builds.


Understand a Custom Agent

Every agent you build in Orchestra is made of the same parts:

  • Purpose: the work it owns, in your own words.

  • When it runs: when you start it, on a schedule, or when an event happens in Fountain.

  • What it uses: Fountain skills, and the apps your account connects to it.

  • Asks first: the actions that wait for a person to approve them before the agent goes ahead.

Each run starts by setting one goal and a plan, then works through the plan. When it finishes, the run shows how it ended, for example Achieved, Partly done, Needs your approval or Failed, so you always know what the agent did.


Open Orchestra

Open Cue in the left navigation, then select Orchestra in the menu under it.

Orchestra home has two parts: a box at the top where you describe a new agent, and the agent library below it. The library lists every agent you can use, with filters by state and owner and a Search an agent field.

Before You Start

Your role needs the Can access Orchestra permission to build agents. An admin turns it on in Settings > Roles & Permissions. A person without it still sees Orchestra once an agent is shared with them, and can use that agent.


Build an Agent

Building an agent is a conversation. Orchestra shows you the whole agent before it goes live.

  1. In the box at the top of Orchestra home, describe the work you want the agent to own, then send it. For example: Every Monday morning, send me a summary of my openings that received fewer than five applicants last week.

  2. Answer the questions Orchestra asks to fill in the details.

  3. Review the agent in the Agent draft panel: its purpose, when it runs, what it uses and what it asks you about first. Ask for any change in the conversation.

  4. Select Turn on when the agent looks right.

The agent is now live in your library. Select Open agent to go to its page.

Start From the Outcome

Describe the result you want, not the clicks. Re-engage idle applicants who are good matches for my open roles gives Orchestra more to work with than a list of steps.


Run and Manage an Agent

Each agent has its own page with how it works, what it uses, the actions it asks about first, and its recent activity. Open a run in Recent activity to read what the agent did and why.

The ⋮ menu on the agent page holds the everyday controls:

  • Run now: start a run of a scheduled agent without waiting for its next time.

  • Pause and Resume: stop the agent's schedule and event triggers without deleting it, then start them again.

  • Pin: keep the agent in your navigation.

  • Copy link: copy the agent page's address to send to a teammate.

  • Tool permissions: choose which tools the agent's scheduled runs may use without asking.

An admin with the Can manage the company's custom agents permission can also Disable any agent in the account, which turns it off for everyone.


Share an Agent

Select Share on the agent page, then add people, roles or, with the right permission, everyone in the account.

The people you share with can use the agent. Only you edit it. Two permissions control sharing: Can share custom agents with people and roles, and Can share custom agents with everyone in the account.

Shared Agents Use Each Person's Access

A shared agent works with the permissions of the person who runs it. Two teammates can get different results from the same agent. If an agent needs an app a teammate has not connected, they connect it the first time they use the agent.


Find Ideas for Agents

A few agents to try:

  • Talent rediscovery: when an opening goes live, find people already in your pools who match it and invite them to apply.

  • Idle applicant follow-up: each week, find good candidates who have stopped moving and nudge them to the next step.

  • Pipeline watch: every morning, flag openings that are falling behind their hiring goal and explain why.

  • Weekly hiring report: send your team a summary of applicants, hires and drop-off by location.

Agents draw on your account's agentic credits as they run. See How Agentic Credits & Billing Works.

Did this answer your question?