How I Used Salesforce to Try to Make It in Music (And Why I Ended Up Building Something Way More Interesting)
There’s a saying in investing: it takes money to make money. Music has its own version of that problem – it takes listeners to get listeners.
If you’re an independent artist on Spotify, you already know this. The algorithm rewards momentum. Playlists amplify it. But breaking into playlists requires proof of an audience you don’t have yet. It’s a catch-22 that keeps most independent artists perpetually invisible, regardless of how good their music is.
I’m a software developer and a musician. When I ran into this wall, I did what any developer would do: I tried to build my way out of it.
The Problem With Being Indie
Streaming platforms like Spotify are simultaneously the best and worst thing to happen to independent music. The best, because distribution is now essentially free and global. The worst, because visibility is entirely algorithmic – and algorithms, by design, amplify what’s already working.
Playlist placement is one of the few meaningful levers an independent artist can pull. Get your track on the right playlist and your streams spike, your follower count grows, and suddenly the algorithm starts noticing you. Listeners lead to more listeners. Streams generate (modest) revenue. Fans accumulate. The flywheel starts to spin.
But getting there is manual, opaque, and exhausting. Playlist curators receive hundreds of submissions. There’s no standard process, no guaranteed feedback, and no systematic way to know which playlists are worth pursuing for a specific track.
That’s the gap I decided to close.
Start Simple: A Playlist Submission Tool
The first version of this project was straightforward. I built a custom Salesforce application to track my music catalogue – releases, tracks, and the Spotify metadata attached to each one. Then I added a playlist submission workflow: find a playlist, submit a track, log the outcome.
Nothing revolutionary. But it gave me something I didn’t have before: a structured dataset of what was working and what wasn’t.
Over time, patterns emerged. Certain types of tracks landed better on certain types of playlists. Acceptance rates varied by genre, tempo, energy. Submission timing seemed to matter. The data was telling a story, but manually extracting the signal from the noise wasn’t scalable.
I needed something smarter.
Enter Agentforce
This is where the project started to get interesting.
The plan is to layer Agentforce into the workflow as an intelligent submission advisor. Instead of manually scouring the internet for playlists that might fit a track, the agent will analyze performance and acceptance history across my catalogue and suggest where to submit next.
Think of it like a recommendation engine that knows your catalogue intimately. It looks at which tracks performed well, which playlists accepted them, what those playlists had in common – and uses that context to prioritize where a new track should go. Rather than spraying submissions at random, the process becomes strategic, guided by data.
That’s the shift I’m building toward. The submission process as a system, not a lottery.
The talk I’m presenting at True North Dreamin‘ – “How Agentforce Helped Me Go Punk (And Viral)” – is the story of building that system. What it took, what surprised me, and what it actually accomplished.
The Pivot That Made Everything More Interesting
Here’s where things took an unexpected turn.
For Agentforce to make genuinely intelligent playlist suggestions, it needs to understand the musical characteristics of a track. Not just the genre or tempo, but deeper qualities – energy, danceability, mood, how acoustic or electronic it sounds, how it would feel next to the other tracks on a playlist.
Spotify used to offer an API endpoint called Audio Features that provided exactly this. It was the foundation I planned to build on. Then Spotify deprecated it.
No replacement. No timeline. Just gone.
At first, that felt like a dead end. But the more I sat with it, the more I realized the deprecation was actually an invitation. If Spotify’s proprietary system for measuring musical characteristics was off the table, I’d have to define and compute those characteristics myself.
That means in addition to my planned Scoring Engine, I’d have to build a Music Intelligence Engine from the ground up – a microservice that takes raw audio, runs it through a purpose-built analysis pipeline, and produces a rich set of audio feature scores: energy, danceability, valence, acousticness, tempo, key, and more. Not borrowed from someone else’s black box, but computed from first principles using signal processing and music theory research.
What started as a workaround is turning into something I’m genuinely proud of.
What’s Coming
This blog series is a companion to the talk and to the project itself. Over the next few posts, I’ll go deep on the two technical pillars that make this thing work:
The Scoring Engine – how raw audio features are combined into composite scores that place a track on a map of mood, energy, and rhythmic drive, and how those scores are used to evaluate playlist fit.
The Music Intelligence Engine – the microservice and data pipeline that replaces Spotify’s deprecated endpoint with proprietary analysis computed from audio waveforms.
Both of these were designed to fit inside a Salesforce data model and integrate with Agentforce. One just went live. One I’m racing to finish before I board a plane and fly to Toronto.
I built the majority of this in the margins of a full-time job and life as a father of two. Whether I make the deadline or not, I’ll tell you exactly how it went.
Matt McGuire is an independent punk artist and Salesforce architect. He’s presenting “How Agentforce Helped Me Go Punk (And Viral)” at True North Dreamin’ in May 2026.