10-lesson course · about 18 hours

Release agents that keep the record straight and ask before they act

For Artists, producers, managers and music teams responsible for creative decisions, release preparation and the details that connect them

Everything for Music teams

Build and supervise agents in OpenAI Dots or Claude Duet that keep assets, credits, listening notes and campaign drafts aligned between your reviews, and bring back what's missing.

Try lesson 1 freeSee what you’ll learn

What you’ll be able to do

By the end, you can…

Give an agent a role it can't misreadName a role such as Release Desk and write its brief: the goal, approved connections, the report you agree on, where a person reviews, and how it is paused or retired. Choose OpenAI Dots or Claude Duet, and a single run or a continuing brief, and give a reason for each choice.
Take in the overnight file drop without losing the approved masterBuild an intake agent that registers new files under one release ID. It logs a new master as a new version, leaves the approved one alone, sets the duplicate artwork aside, and asks for the missing instrumental a set number of times before it escalates.
Stop a later file from rewriting the creditsSet up a scheduled metadata agent that compares changed records against the approved credit sheet, keeping composition and recording credits apart. It drafts corrections with their sources and freezes a disputed split for a person to resolve.
Turn listening notes into the next revisionKeep a listening queue that pairs each version with the artist's written direction, and track every note to resolution. Listeners judge the sound. The agent keeps versions, notes and follow-ups, and holds any request that contradicts the direction.
Keep campaign drafts honest when the date movesKeep a dated draft queue in which every draft cites the approved facts it uses. When a date, title or artwork changes, only the affected drafts get rebuilt, superseded versions are cleared without losing the history, and nothing is published.
Test the whole workflow before a real release depends on itRun the combined workflow against duplicates, missing files, clashing credits, a moved date, a delayed approval and an interrupted run. Check that it pauses, recovers, submits nothing, and leaves a status report someone else can read.

How it works

A tutor that meets you where you are

  1. Lessons that meet you at your levelTen lessons follow one independent single from an overnight file drop to a package waiting for approval. Each lesson adapts to what you already know about releases and agents, so you skip what you know and spend more time on what's new to you.
  2. An AI tutor you can ask anythingAsk questions typed or out loud while you work. The tutor can help with a role brief, a freeze rule or an escalation message, and it walks you through the build-and-test steps a live trainer would otherwise run.
  3. Quizzes and flashcards for each lessonShort quizzes and flashcards tied to each lesson check the rules that matter, such as what a correct stop looks like and why a later file doesn't win.
  4. Progress tracking and a certificateSee which lessons and practice builds you've finished, pick up where you left off, and get a certificate of completion when you finish the course.

Curriculum

The curriculum

Five modules, two lessons each, built around one practice release. Each module adds one bounded agent to the workflow, and the last module tests the whole workflow under pressure.

Module 1Keep new assets tied to the release.2 lessons

  1. 1Register incoming assets
    • Write a role brief for a named agent such as Release Desk, covering its goal, approved connections, agreed report, human review points, and how it is paused and retired
    • Choose between OpenAI Dots and Claude Duet, and between a single run and a continuing brief, and give a reason for each choice
    • Build an intake agent that registers a batch of sample or permitted files under one release ID and reports what arrived
    • Explain which connections the agent may use and which stay off-limits for unreleased masters, private contracts and contributor details
  2. 2Handle repeats and missing files
    • Build intake rules that log a new master as a new version and leave the approved version unchanged
    • Test that running the same batch twice adds no new records and no duplicate tasks
    • Set a request limit and an escalation owner for a missing file such as the instrumental
    • Explain what a correct stop looks like and why it counts as success

Module 2Keep credits and metadata consistent.2 lessons

  1. 3Compare contributor records
    • Build a scheduled metadata agent that compares changed records against the approved credit sheet
    • Explain why composition credits and recording credits are compared separately
    • Prepare correction drafts that cite their source and wait for a person to accept them
    • Choose what contributor information appears in the agent's report and what stays private
  2. 4Escalate conflicting credits
    • Build a freeze rule that holds a disputed credit or split and keeps both versions on record
    • Write an escalation message that tells the owner what conflicts, what is frozen and what decision is needed
    • Test that a later, contradicting split sheet does not replace the agreed share
    • Explain which questions the agent must route to people and their advisers

Module 3Turn listening notes into the next revision.2 lessons

  1. 5Prepare the listening queue
    • Build a listening queue entry that pairs a version with the direction it must be heard against
    • Explain why the agent must never judge sound from file names or metadata
    • Choose which supplied comparisons and permitted audio belong in a review, and who receives access
  2. 6Track feedback to resolution
    • Build a feedback tracker that turns each note into an item with a version, owner, status and follow-up task
    • Test that a request contradicting the artist's written direction is held and routed to the artist
    • Explain the difference between the agent closing a task and a person accepting a result

Module 4Keep campaign drafts aligned with the release.2 lessons

  1. 7Prepare a dated content queue
    • Build a dated draft queue in which every draft cites the approved release facts it uses
    • Explain why the campaign agent has no publishing connections and produces drafts only
    • Write honest AI disclosure language that describes what AI did and what people did
  2. 8Rebuild only affected drafts
    • Build a rebuild rule that finds and redrafts only the drafts citing a changed fact
    • Test the queue against a moved release date and a replaced artwork file
    • Explain how superseded dates and artwork are cleared from the review queue without losing the history

Module 5Assemble a release package for approval.2 lessons

  1. 9Check the delivery manifest
    • Build a packaging agent that checks the delivery manifest against the distributor's spec and the release record
    • Explain why a complete package still waits for artistic, permissions and release approval
    • Write a release status report that another person can read without asking questions
  2. 10Test a delayed approval
    • Test the workflow on duplicates, missing files, clashing credits, a moved date, a delayed approval and an interrupted run
    • Choose reminder limits and escalation owners for an approval that does not arrive
    • Explain what a correct stop and a clean recovery look like in the agents' reports
    • Build the pause and retire steps for each agent and carry them out on the practice release

Who it’s for

Made for you if…

Made for: Artists, producers, managers and music teams responsible for creative decisions, release preparation and the details that connect them

Independent artists who run their own releases

You want the admin to keep moving while you're in the studio, and you want the creative calls and your consent to stay with you.

Producers who handle the files between studio and distributor

You've seen the wrong bounce go up or a session player left off the credits, and you want version control and credit checks to run without you watching.

Managers keeping several releases on track

You need disputed splits, missing assets and late approvals to reach the right person early, with a status report you can pass on.

Small label and music teams

You want a repeatable, supervised release-prep workflow that pauses in the right places and never publishes or submits on its own.

Questions

Good to know

What do I need to know before starting?

You should already be able to give AI a single task with a written brief that names the song's fixed element and marks assumptions. You should keep a track sheet that separates composition credits from recording credits, with an owner for every unsupported field, and write release copy only from approved facts. You should also know the basic parts of a release: masters and versions, artwork, credits and splits, release date and the distributor's delivery spec.

Which agent tool do I need?

You'll need access to OpenAI Dots or Claude Duet, or a way to get it. The course treats them as equal options and helps you choose one. You'll also need sample files, or release files you have permission to practise on.

Will the agents publish posts or submit my release to a distributor?

No. Every agent you build prepares, drafts and reports, and the campaign agent has no publishing connections. Even a complete package waits for artistic, permissions and release approval from people.

Will the agents judge my mixes or masters?

No. Listeners judge the sound. The agent keeps versions, notes and follow-up tasks, uses only supplied LANDR comparisons and permitted audio, and never judges sound from file names or metadata. Mixing and mastering technique aren't covered.

Does the course give legal advice on splits, ownership or samples?

No. You'll set up agents to send consent, ownership, sample clearance, voice imitation and disputed credits to the right people and their advisers. The course doesn't settle those questions itself.

Does it cover song generation or agent teams?

No. Generating songs with tools such as Suno is covered in the Intro. Agent teams sharing one release record are covered in the Capstone. This course covers single, bounded agents and how to combine them into one supervised workflow.

How long does it take?

About 18 hours across ten self-paced lessons.

Whose ideas does the course draw on?

The lessons draw on perspectives from will.i.am (own your agent), Imogen Heap (delegate the admin, keep the creative calls), Brian Eno (seeds that carry their own instructions), Wyclef Jean (a placeholder is not a performance) and Tom Gray (credits as attribution). They are independent and don't endorse the course.

Music AI Intensive: Proactive Agents for Release Preparation

$397

One payment · lifetime access

  • 10 self-paced lessons in 5 modules, built around one practice release
  • Hands-on builds in OpenAI Dots or Claude Duet: intake, credits, listening, campaign and packaging agents
  • Tests for duplicates, missing files, clashing credits, a moved date, a delayed approval and an interrupted run
  • An AI tutor you can ask anything, typed or out loud
  • Quizzes and flashcards for each lesson, plus progress tracking
  • Certificate of completion

Secure checkout by Stripe. Have a promotion code? Enter it at checkout.

Refunds: Full refund if you ask before you open your first lesson. Once you’ve started, purchases aren’t refundable, except where the law gives you that right.

Already bought it? Sign in