Looking back, however, I have to admit that my opinion was based largely on instinct rather than experience. I had formed a view about a technology that I had never actually used. Whether that instinct eventually proved to be right or wrong seemed less important than recognising that it wasn’t a particularly good foundation on which to build a considered opinion.

For most of my life I have tried to avoid dismissing ideas before understanding them, so in March 2026 I decided to spend some time exploring Suno. There was no grand plan and certainly no intention of becoming an AI musician. I wasn’t trying to reinvent my practice or discover a faster way of making music. Curiosity was the only real motivation. I simply wanted to understand what these systems were capable of and whether they had anything meaningful to contribute to the way I already worked.

I had no idea where that decision would lead.

Over the following three months I probably produced more music than at any other comparable period in my life. At first glance that sounds rather extraordinary, but it would be misleading to imagine that AI had somehow replaced thirty years of musicianship, production experience and collaboration. The opposite is probably closer to the truth. The work only existed because of everything that had come before it.

Every project began with something that already had a human history.

Light On Water emerged from a collection of unfinished piano pieces and instrumental sketches that had accumulated over many years. The Songs of Matthew Dench began, quite naturally, with Matthew’s songwriting. Ghosting revisited recordings that Dan Pierce and I had made more than twenty years ago, while Say Something Dumb returned to songs that Dan had written and that I had originally produced a quarter of a century earlier. Even the drum and bass tracks I made with my fourteen-year-old son, Euan, began with two-bar loops that he had created himself using a simple app on his iPad.

The more I reflected on these projects, the more I realised that none of them had actually started with AI. They began with songs, sketches, rehearsal recordings, friendships and unfinished ideas. AI entered much later in the process. It wasn’t the source of inspiration. Instead, it became a remarkably powerful means of reinterpretation, allowing me to ask questions of existing material that I would probably never have thought to ask otherwise.

That distinction has become increasingly important in my own thinking because it describes much more accurately what I was exploring. I wasn’t asking a machine to become a songwriter. I was inviting it into music that already carried its own history, relationships and creative intent.

Some of the results genuinely surprised me.

There were moments when I heard possibilities that I don’t believe I would ever have discovered on my own. Some arrangements revealed aspects of songs that had somehow remained hidden for years. Say Something Dumb, in particular, became one of the releases I have returned to more than any other this year. I cannot honestly dismiss that simply because it complicates my conclusions. If I am going to write honestly about this period, then I also have to acknowledge that I genuinely enjoyed much of the music that emerged from it.

Had the technology simply produced poor results, the story would have ended there. I would almost certainly have closed the laptop, satisfied my curiosity and carried on exactly as before. Instead, I found myself facing a much more interesting problem. The technology was capable of producing work that I found engaging and, in some cases, genuinely moving. Rather than settling the debate, it forced me to think much more carefully about the questions I was asking.

As the months passed, I found myself reading more widely than I ever expected. I listened carefully to musicians who believed AI represented the most significant democratisation of music making since affordable home recording. Their argument deserves to be taken seriously. Throughout history, technological developments have repeatedly lowered barriers to creativity. Recording equipment became smaller and cheaper. MIDI made complex arrangements accessible to individuals. Digital audio workstations brought professional production into spare bedrooms. AI appears, at least on the surface, to continue that progression by allowing people with ideas but limited resources to hear their music realised in ways that were previously beyond their reach.

There is something genuinely attractive about that vision. As someone who has spent much of his career trying to encourage other people to make music, I find it difficult to argue against greater accessibility.

At the same time, I found myself increasingly persuaded by another line of argument. The more remarkable these systems appeared to become, the more important it seemed to ask how they had acquired those abilities in the first place. Current generative models appear to have been trained on enormous collections of human creative work, much of it gathered without the explicit permission of the musicians, composers and artists who created it. Whether that ultimately proves to be lawful is a matter for legislators and the courts. My own concern gradually shifted towards something slightly different. If the extraordinary capabilities of these systems have been built upon generations of creative work contributed without consent, can we really separate the quality of the outcome from the ethics of the process that produced it?

That question stayed with me throughout the remainder of the experiment, and it remains unresolved.

I often find myself imagining how differently I might feel if companies such as Suno chose to invest in future systems built upon recordings, performances and compositions that artists had knowingly contributed and for which they had been fairly compensated. Imagine thousands of musicians choosing to participate because they believed in the project and because they were treated as collaborators rather than simply sources of data. The technology itself would remain just as remarkable, but its foundations would feel very different. I suspect the debate would become considerably more interesting because it could begin to focus less on how the systems had been built and more on how they ought to be used.

As the months passed, I realised that my concerns were no longer limited to copyright or consent. Those questions remained important, but they gradually became part of something much larger. The more I read about the extraordinary investment flowing into artificial intelligence, the construction of vast data centres, the increasing demand for energy and water, and the concentration of power within a relatively small number of technology companies, the more I found myself wondering what kind of future this technology was really encouraging.

It would be easy to dismiss those concerns as being outside the scope of music, but I don’t believe they are. Music has always existed within a wider cultural, economic and political landscape. The tools we choose are never entirely separate from the systems that produce them. Throughout my career I have tried, however imperfectly, to make creative decisions that reflect the values I hope my work embodies. It became increasingly difficult to separate my excitement about the possibilities of generative AI from the wider questions surrounding its development.

None of this led me to conclude that AI itself is inherently bad. Technology rarely fits comfortably into such simple categories. Rather, I became uneasy about the direction in which so much of its development appears to be heading. When extraordinary technological capability becomes closely tied to the interests of a relatively small number of powerful corporations, whose commercial incentives do not always align with those of artists, communities or the environment, it seems entirely reasonable to ask difficult questions before embracing the technology uncritically.

I found myself wondering whether my growing discomfort was really about AI at all. Perhaps it was about power. Who develops these systems? Who benefits from them? Who carries the environmental cost? Who is asked for permission? Who is compensated? Those questions seemed to extend well beyond music, yet they increasingly influenced how I felt about using AI within my own practice.

Perhaps the most revealing part of the whole experience wasn’t the music I chose to explore but the music I deliberately left untouched.

During those same three months it never once occurred to me to involve AI in The Powdered Earth. Shane and I have spent years developing not only a body of work but also a creative relationship that has become inseparable from the music itself. Our conversations, disagreements, experiments and shared discoveries are as important to me as the finished recordings. Introducing AI into that relationship simply never felt appropriate.

The same instinct applied to Big Blue Sun and Mindful Systems, the album Dan Pierce and I recorded together in the Highlands of Scotland in 2022. Those projects represent living collaborations. They continue to belong entirely within the creative relationships from which they emerged.

Interestingly, I also felt no temptation to revisit SPACES or TRACKS. Both projects already felt complete. They had become the works they were intended to become and I saw no reason to reinterpret them simply because a new technology had appeared.

The projects that did invite exploration all had something else in common. They contained ideas that had never quite reached their intended destination. Light On Water remained a collection of sketches. Matthew Dench’s songs deserved productions that circumstances had never allowed us to create. Ghosting belonged to another period of my life and naturally lent itself to reinterpretation rather than preservation. Even there, however, AI has not become the destination. The Ghosting reinterpretations are already beginning to serve as production templates for new recordings featuring real singers and musicians. Similarly, I have another collection of nine Dan Pierce songs that never progressed beyond rehearsal recordings. Those, too, seemed to invite exploration because they represented unrealised possibilities rather than completed artistic statements.

Looking back now, I realise that I wasn’t making decisions about technology at all. I was making decisions about stewardship.

Some projects asked to be preserved exactly as they were. Some represented living creative relationships that I wasn’t prepared to alter. Others remained unfinished, waiting patiently for an opportunity to become something more than they had previously been. AI became one means of exploring those possibilities, but it never replaced the relationships, histories and creative decisions that lay behind them.

As my understanding deepened, another change quietly took place. I didn’t decide at the outset that the exploration would be limited. That decision emerged gradually. The more I learned about the technology, the more I found myself reflecting not on what it could do, but on where I wanted to spend my own creative life. I had learned what I had come to learn. Continuing indefinitely no longer felt necessary.

I have no intention of hiding the work that emerged from those months. On the contrary, I think it represents an honest record of an important stage in my development as a producer. The recordings will be available on SoundCloud, where they can be heard as the explorations they genuinely were. I have deliberately chosen not to treat them in the same way as the rest of my catalogue. Bandcamp, whose approach to AI I respect, does not permit AI-generated music, and I have recently made a conscious decision to use streaming much more sparingly as a means of distribution in any case. Those decisions feel consistent with where I have arrived.

When I began this journey, I imagined I was setting out to understand artificial intelligence. Looking back, I think what I actually discovered was something much closer to home. The experiment reminded me that the part of music I value most has never been efficiency, productivity or even the finished recording itself. It lies in the relationships that shape the work, the slow development of ideas, the conversations that happen along the way and the shared experiences that eventually find their way into songs.

Looking back now, I don’t think the most important decision I made was to spend three months exploring AI. The more important decision was recognising when I had learned enough. Curiosity took me into the technology. Discernment told me where it belonged.

I remain genuinely fond of much of the music that emerged from that period and I have no intention of pretending it never happened. Those recordings represent an honest piece of my creative life and I will make them available as such. They are not hidden away because they embarrass me, nor are they presented as the future of my work. They simply document a period when I believed it was important to understand something before deciding what I thought about it.

If there is a thread running through everything I have made over the last thirty years, from Big Blue Sun to SPACES, from TRACKS to The Powdered Earth, it is not really about technology at all. It is about people. Songs. Places. Friendships. Stories. Ideas that deserve to be realised and collaborations that deserve to be protected. AI briefly became part of that story, but it never became the story itself.

I don’t imagine these conclusions will be the last thing I ever think about artificial intelligence. If the past thirty years have taught me anything, it is that creative practice is a continuous process of questioning rather than arriving. New technologies will emerge, ideas will change and my own understanding will continue to evolve. This essay isn’t intended to be the final word on AI, nor even my final word. It is simply the point at which one particular investigation reached its natural conclusion.

The AI chapter forms an important part of my story, but it remains only that: a chapter. It clarified my thinking, challenged some of my assumptions and produced music that I remain genuinely pleased to have made. More importantly, it reminded me why, after thirty years of making records, I still believe that the most valuable part of music has always been, and perhaps always will be, the people who make it together.

Between March and June 2026 I set out to understand a technology that I initially viewed with suspicion. I found remarkable possibilities, encountered difficult ethical questions and, perhaps most importantly, learned something about my own practice. That feels like enough for now. The recordings remain as evidence of that exploration, but my attention has already begun to return to the things that first drew me to music: people, places, collaboration and the slow unfolding of ideas over time.

Curiosity took me into the AI chapter. It is curiosity, once again, that now leads me elsewhere.