AI in the music industry

How are record labels using AI?

Record labels can use AI to organize information, prepare content, compare reports, and reduce the administrative work surrounding a release. I’d focus first on jobs that keep experienced people from spending enough time on the music and the artist.

A release involves more people than most fans ever realize. There are conversations across A&R, marketing, digital, radio, publicity, creative, distribution, legal, and management. Every department sees a different part of the picture. The challenge isn’t always that nobody has the information. It’s that somebody has to find it, make sense of it, and get it to the right person.

For A&R, I see value in background research, catalog organization, and pulling together the context around an artist. That might help somebody arrive at a meeting better prepared. It shouldn’t be mistaken for an opinion about whether the music is good.

For marketing, I’d use AI to prepare briefs, organize release schedules, compare approved campaign materials, and help turn a rough idea into the first version of something a team can review. The creative idea still has to come from somebody who knows the artist. Otherwise, a lot of campaigns start sounding like variations of the same thing.

For creative production, there’s useful work in transcripts, searching interviews, organizing footage, and finding material that has already been captured. I’ve spent enough time around concert footage to know how long it can take to locate one specific moment. Finding that moment faster doesn’t remove the need for an editor who knows which performance tells the story.

And on the operational side, basic checks matter. Missing information, inconsistent files, unclear approvals, and versions that get passed around too many times can create work that nobody planned for.

I don’t know the internal systems of every label, and I wouldn’t pretend there’s one approach that fits them all. I’d start with the work people already find frustrating. If AI can help them arrive at better-prepared decisions, that’s useful. If it’s just another system everyone has to maintain, I wouldn’t add it.

How can artist managers use AI?

Artist managers can use AI to prepare for the week, organize conversations, review information, and keep track of decisions that involve a lot of different people. I think the biggest opportunity is getting back the hours that disappear into coordination.

A manager’s week rarely follows a neat plan. An artist needs an answer. A label is waiting on an approval. A show changes. A release gets moved. A creative asset isn’t ready. Everybody’s working, but someone still has to make sure the pieces fit together.

I’d start with meeting notes and follow-ups. Those are easy to underestimate because each conversation seems manageable on its own. Across a busy week, it’s easy for a small decision to get buried. Having a clear record of what was agreed, what’s still open, and who is responsible can make the next conversation more productive.

Then I’d look at reporting. Managers receive information from multiple parts of an artist’s career, but the reports don’t all answer the same question. One might describe listening activity. Another might show what happened with a campaign. Another might involve ticket sales or a scheduled run of shows. I’d want the relevant material organized before deciding what it means.

I also see a place for help with early research. If an artist is weighing a new opportunity, the manager needs to know what questions to ask, which assumptions are being made, and what evidence would actually change the decision. AI can assist with that preparation without pretending it knows the artist better than the people in the room.

Communication is another practical place to begin. A first draft of an update, a meeting agenda, or a summary can save time. But I’d always review sensitive messages before they leave my inbox. A relationship is more than the wording of an email.

I wouldn’t automate a manager’s judgment, and I wouldn’t recommend a complicated new process just because it’s possible to build one. I’d find one recurring problem, improve it, and see whether the change holds up after a few weeks of real work.

Can AI help with music marketing and release planning?

AI can help organize a release, pressure-test a plan, and reduce the time spent creating repetitive campaign materials. It can’t give an artist a reason to matter, and I wouldn’t ask it to invent one.

I think there’s a big difference between a release calendar and a release strategy. A calendar tells everybody what happens on Tuesday. A strategy should explain why that work matters, who it’s for, and what the team wants to learn.

I’d start with the music and the artist. What’s the story behind the song? What does the artist want to say? Who’s already paying attention? What worked on previous releases, and what just looked busy? Those questions are more interesting than asking a chatbot for thirty content ideas.

Once the direction is clear, AI becomes useful for preparation. It can help turn decisions into briefs, organize deliverables, make a schedule easier to understand, and draft variations of approved messages for different places they might appear. That can remove a lot of small back-and-forth without changing the artist’s point of view.

During a campaign, I’d want regular, understandable updates. Did something change? Did we learn anything? Did a piece of content perform because people genuinely cared about it, or because it reached a large number of people once? I wouldn’t ask one number to answer a question it wasn’t designed to answer.

Afterward, I’d take the review seriously. I see too many reasons to move quickly into the next thing and too few reasons to stop and ask what the last release actually taught us. Even a small, honest review is better than repeating the same plan because it feels familiar.

Where I think AI makes marketing worse is when the tool starts making the creative decisions. Generic captions, identical graphics, and content that sounds nothing like the artist can fill a calendar while saying very little. More content isn’t automatically better communication.

My preference is to let people develop the idea and the point of view. Use AI to help them do the work around it. If the result makes the artist seem less like themselves, I’d rather leave the tool out.

How do you use streaming and radio data to make career decisions?

Streaming and radio data can help explain where an artist is getting attention and how that activity changes over time. But neither can tell the whole story of an artist’s career, and treating one number as the answer is where people get into trouble.

There are well-established sources for different kinds of music information. Luminate is known for measuring music consumption. Mediabase is widely used to examine radio airplay. Chartmetric presents a wider view of publicly available artist activity across music and social services. Each describes a different part of a career. They shouldn’t be treated as interchangeable, and access or measurement details depend on the actual service and license.

What interests me isn’t just how many times something happened. It’s what the number can reasonably tell us.

A song can receive significant radio play without proving that the same listeners will buy a ticket. A city can show strong listening activity without proving that a second show makes financial sense. A piece of social content can travel widely without necessarily building a lasting relationship with the people who saw it.

None of that makes those measurements useless. It means the questions have to get better.

If I were advising an artist considering a career decision, I’d want to understand what outcome they’re actually trying to create. Is it growth in a particular market? A more sustainable touring business? A better understanding of who keeps coming back? The answer affects which information matters and what additional context is missing.

I’d also want to be honest about uncertainty. Different sources can have different definitions, coverage, time periods, and limitations. A clean-looking chart doesn’t make those differences disappear. And a confident answer isn’t the same thing as a defensible one.

This is the kind of thinking that led me to build Sorvo. I want decisions to be made with a clearer picture of the evidence, while keeping the experience of artists, managers, programmers, and other professionals at the center. The data should improve the conversation. It shouldn’t end it.

Can AI fix royalties and metadata?

AI can help people notice inconsistent information, missing fields, and records that deserve another look. It can’t decide who owns a song, fix a disputed agreement, or guarantee that the right person gets paid.

Royalties are complicated partly because a recording and the composition behind it can involve different rights and different people. Information can be entered by different parties at different times. If those records don’t line up, the problem may sit unnoticed until somebody starts trying to understand a payment.

Metadata sounds boring until the wrong recording is associated with the wrong information. Artist names, writers, identifiers, titles, alternate versions, and ownership claims all matter. Even a small inconsistency can create questions that take real work to resolve.

I see a useful role for AI in helping people review the material they already have. It may help prepare a list of inconsistencies, organize correspondence, summarize what the documents say, or show where the available records don’t appear to agree. That could make the next conversation with the right administrator more productive.

But I draw a hard line between flagging a possible issue and declaring the answer. A tool cannot read a conflicting document and simply decide which party should own the rights. That can have financial and legal consequences, and the people making those decisions need accurate records and the proper authority.

There are also different organizations involved in different rights and payment types. Public resources from The Mechanical Licensing Collective and SoundExchange can help explain some of those distinctions in the United States, but neither replaces the need to understand a specific agreement or catalog.

I’d be skeptical of any product promising to “fix your royalties” in one click. A real correction may require registrations, documentation, communication with rights holders, and sometimes legal advice.

The opportunity I see is making complicated information easier for qualified people to examine. If that helps someone catch a mistake earlier, that’s worthwhile. If it encourages people to make confident ownership claims without evidence, it can make the situation worse.

What AI tools are useful for radio programmers?

I’d use AI in radio to help with preparation, research, transcription, and the repetitive work surrounding a good show. After fourteen years at iHeartRadio, I wouldn’t start by replacing the things that make a station feel local and human.

There’s a lot of work in radio that listeners never hear. Somebody needs to prepare an interview, check the details of a promotion, find a quote from a previous conversation, or make sure the information on the station’s website is current. Those jobs matter, but they can eat into the time programmers and talent have to work on the actual show.

For editorial preparation, I’d consider tools that help organize source material and make long documents easier to review. I’d want links back to original reporting and clear dates. An incorrect story delivered with a confident tone isn’t helpful. In radio, it can become an on-air mistake very quickly.

For production, transcription and searchable audio can be useful. Finding a specific moment in an interview is easier when there’s a reliable way to search what was said. I’d still want a producer or editor to make the creative call about what belongs in the final cut.

For promotions, I’d use AI to check copy against approved details and help prepare different versions for different channels. But I wouldn’t trust it to invent contest rules or decide whether a promotion satisfies legal requirements.

I also wouldn’t confuse an AI-generated music recommendation with knowing a market. Radio programmers have to think about the audience, the station’s identity, the music, and everything else happening in that city. Information can help support those decisions. It doesn’t absolve anybody of making them.

The category I’d avoid is a tool that makes every station sound like it came from the same template. Radio already has plenty of ways to become less local. I’d rather use technology to give programmers and talent more time to be interesting.

That’s the test I’d apply to any product: Does it help the people who know their audience do better work, or does it just make something easier to automate?

What can’t AI do in A&R?

AI can help organize artist research and surface things a person might have missed. I don’t think it can reliably explain why one performer connects with a room, why a song becomes part of somebody’s life, or which unusual artist deserves more time.

A&R isn’t just about finding activity. It’s about listening, recognizing a point of view, understanding an artist’s potential, and making a decision before the outcome is obvious. Some of that judgment comes from experience. Some comes from taste. And some comes from being willing to believe in something that doesn’t look like an easy bet on paper.

I understand the appeal of using numbers to make that process feel more objective. There are useful questions to ask about an audience, a release history, and how people’s attention changes over time. I would absolutely want that context before committing to a major decision.

But the strongest number in a report isn’t necessarily the most interesting artist in the room.

Sometimes an artist hasn’t figured out how to present their music yet. Sometimes the live show communicates something the recorded material doesn’t. Sometimes the thing people initially find strange turns out to be the reason an artist is worth paying attention to. If everybody uses the same rules to decide what’s promising, I worry we start looking for copies of things that have already worked.

I’d rather use AI to challenge an opinion than have it manufacture one. If I love an artist, I want to understand the evidence that supports or contradicts that feeling. If an artist has impressive activity but the music isn’t connecting with me, I’d want to know what other listeners are responding to. Neither side wins automatically.

What would change my mind? I’d want repeated, well-tested evidence that an AI-based approach identifies genuinely promising artists before their success is already apparent, across very different styles and circumstances. And I’d want to compare that performance fairly against people making the same decisions with the same information.

Until then, I think A&R still needs people willing to hear something unexpected and trust their judgment enough to investigate it.