I use AI in music production almost every day. I also have a hard line I won’t cross. People assume those two facts contradict each other; they don’t. Here’s the actual workflow inside my studio in 2026 and the rules I run on.

What AI does in my sessions

Five places it’s currently in the chain:

  • Stem separation — pulling reference vocals or drums out of finished tracks for study. LALAL.AI and similar tools made this trivial.
  • De-noise and de-hum — iZotope RX and the free Adobe Enhance equivalent. Saved more vocal takes than I can count.
  • Sketching arrangement ideas — Suno or Udio as a brainstorming partner, never as a final source.
  • Mastering pre-pass — services like LANDR or my own Ozone-driven chain. Pre-pass only; final pass goes to a human.
  • Lyric ideation — Claude as a thesaurus and rhyme-spar partner. Never delivering a finished verse.

What AI doesn’t do

  • Write lyrics I publish under my name
  • Generate vocals that ship as my voice
  • Compose top-line melodies I claim authorship over
  • Master a final commercial release

Those are the four lines. Crossing them isn’t a creative tradeoff — it’s an identity problem.

The principle behind the rules

The test I run is: could I, in good faith, sit across from a fan, point at this section, and say “I wrote that”? If yes, AI was a tool. If no, AI replaced me. Tools are fine; replacement is the whole problem.

Where most artists go wrong

Two failure modes I see:

  1. Total avoidance — refusing AI on principle while shipping at half the pace of peers who use it. The market doesn’t care about your principles; it cares about your output.
  2. Total surrender — outsourcing songwriting to Suno and lying about it. Audiences eventually find out, and “eventually” is faster than ever.

The middle path — AI as instrument, human as author — is the only sustainable one.

The 2027 question

The question I keep asking myself: when AI music gets so good that listeners can’t tell, do my rules still hold? My honest answer: yes, but for a different reason. The contract between artist and listener isn’t about audibility — it’s about the fact that someone meant something. AI can generate; only humans can mean. That’s the line, and it doesn’t move.

Producer wrestling with the same questions? Let’s talk.