What can an AI-generated outlaw country band teach us about collaboration?

← Back to Blog

I built an AI-generated outlaw country band called Southern Oracle with my childhood friend. Three women: Syd, Blair, and Maria. They play a sound I describe as outlaw country meets tarot and alchemy. They have lyrics, voices, visual identities, a debut album called Turn the Cards (on your streaming platform now). None of them are real. This is the part of my AI work that makes people in boardrooms look at me sideways. So let me explain what it has to do with everything else.

Some context: I am an artist. My first degree is art history. I paint, I write poetry, I play the piano badly, and I have been writing songs for as long as I can remember. What I am not is a music producer, arranger, or vocalist. When Suno made it possible to take lyrics, specify a genre direction and vocal character, and generate a fully arranged, mixed, and mastered track, it closed a gap that had kept these songs in notebooks. DistroKid put them on every streaming platform. But I want to be precise about what the AI is actually doing here, because the discourse around generative music tends to land on one of two poles ("the AI made it" or "the AI is just a tool") and neither is accurate.

Suno is not a recording studio that follows instructions. It interprets a text prompt and makes compositional decisions: melody, harmony, arrangement, instrumentation, vocal timbre, dynamics, phrasing. I provide lyrics, a genre tag, a mood direction, sometimes a structural note. The system generates a complete musical performance. Some of what it produces I would never have conceived. Some of it I reject and regenerate. The result is a negotiation, not a dictation. That negotiation is the interesting part, and it is the part that connects to my research on human-AI collaboration.

This is also a project rooted in friendship. My childhood friend and I have been writing stories together since we were kids, building worlds. Southern Oracle gives life to those stories with tools that did not exist two years ago. The catharsis of hearing a song you wrote performed by a voice you shaped is real. But the creative line between human contribution and AI contribution is not where I expected to find it.

The hypothesis: creative projects with generative AI produce insights about human-AI collaboration that are inaccessible through analytical work alone, because they engage the builder's identity, aesthetic judgment, and emotional investment in ways that enterprise projects do not.

Three takeaways

First, building Southern Oracle taught me something about co-evolution that my research frameworks had not captured. In my collaborative intelligence work, co-evolution is the mechanism where humans and AI systems learn from each other over time. In enterprise settings, I struggled to find documented evidence of it actually occurring. Working on a band with AI, I experienced it directly. Suno generated musical ideas that reshaped what I wrote next. My revisions (different lyrics, different genre tags, different structural cues) changed what the system could produce. The outputs were genuinely collaborative in the sense that neither party alone would have arrived at the result. Co-evolution, it turns out, requires creative stakes, not just task completion.

Second, the parasocial relationship with fictional performers is not new, but AI makes it available to creators, not just audiences. Horton and Wohl (1956) first described parasocial interaction as the illusion of face-to-face relationship with media performers. KPop Demon Hunters, named TIME's 2025 Breakthrough of the Year with over 500 million Netflix views, proved that fictional characters can generate the same intensity of parasocial fandom as human performers. Their songs charted higher on Spotify US than BTS or Blackpink ever had. But there is a difference between consuming a parasocial relationship as a fan and constructing one as a creator. Designing Syd, Blair, and Maria meant building what Winnicott (1971) would recognize as a transitional space: characters that are authored and generated simultaneously, existing between self and other. The question is what this creator-side parasocial experience teaches about human-AI trust more broadly.

Third, the line between human creativity and AI creativity may not exist as a line at all. Benjamin (1935/1969) argued that mechanical reproduction strips art of its aura. Generative AI complicates this: the outputs are novel productions emerging from human intent and machine pattern. When I write lyrics and Suno generates music, the contributions are not sequential. They are entangled. I chose the words, the emotional arc, the genre direction, the vocal character. The system composed the melody, the arrangement, the sonic texture, and sometimes those compositions pushed the lyrics in a direction I had not imagined, prompting rewrites that generated new outputs in turn. Csikszentmihalyi's (1996) systems model describes creativity as the interaction among individual, domain, and field. Generative AI introduces a fourth element: a non-human collaborator operating within the domain's rules but sharing neither the individual's motivation nor the field's evaluative standards. The boundary between contributions is not clean. I am increasingly unsure it needs to be.

The longer view

Winnicott (1971) described transitional objects as items existing between self and other, between inner reality and external world. Southern Oracle occupies that space: authored and discovered simultaneously. Learning to hold creative intent while remaining open to what the system offers translates directly to how I think about leading AI transformation. The best outcomes come from holding your vision loosely enough that the system can teach you something.

Fernando Pessoa created over 70 literary alter egos, each with distinct styles and philosophies, describing them as personalities he channeled rather than invented (Zenith, 2021). Building a band with generative AI resonates with Pessoa's practice: the characters develop a coherence that feels directed and emergent at once. What generative tools change is the medium of emergence (from text to sound, image, and persona) and the speed at which it happens.

My two cents

Southern Oracle is the most fun and, paradoxically, the most instructive part of my AI work. The boardroom work is important. The research is rigorous. But the creative work is where I learn things I cannot learn any other way, because it is the only context where I am willing to be genuinely surprised by what the system produces, and where surprise is a feature, not a failure mode.

I came into this expecting the creative boundary to be clear: my words, the AI's music, distinct contributions, clean line. Instead I found entanglement. The AI's output changed what I wrote next. My revisions changed what the AI produced. The negotiation between human intent and machine interpretation is where the interesting work lives, in music and I suspect in every other domain where generative AI is being used seriously.

A portion of proceeds from Southern Oracle will go to Girlstart, a national nonprofit that empowers girls through STEM education. Somewhere out there is the girl who is going to build the first AI popstar that changes everything, and I cannot wait to see what she makes. The tools are already here. Making sure girls know they are allowed to use them is the work.

Read to learn more

Academic: Csikszentmihalyi, M. (1996). Creativity: Flow and the psychology of discovery and invention. HarperCollins.

Industry: TIME. (2025). KPop Demon Hunters: Breakthrough of the Year 2025.

References