ECPA Conference Reflections
I’m writing this on the way home from Chicago, where I gave a keynote on AI and publishing at the ECPA Leadership Summit. As I said in my presentation, this is not an area of publishing I’ve worked in, and my role was to provide an outside perspective—less about the specifics of the sector, more about what AI might look like when viewed from beyond it.
More specialist publishing conferences are a very different experience to bigger, more general events like book fairs. There was a very strong sense of community, open conversation and publishers supporting each other on common issues—always in an appropriate way. I felt the same thing at the ALPSP University Press Redux conference earlier in the year, and at IPG events. That openness feels like a competitive advantage in a moment of uncertainty—though perhaps one that only works at this scale. It would certainly be hard to imagine an event with some of the larger trade publishers, gun-shy from previous competition investigations, working like this.
Being in Chicago meant that I got to catch up with my friend Andrew Savikas, whom I first met through the O’Reilly Tools of Change for Publishing conferences in New York and Frankfurt. Andrew was at the forefront of online publishing of books at Safari and subsequently Holloway, and is now documenting his experiments with AI on his website. It was great to chat to him over lunch as we’re thinking about similar things from different perspectives. Reminiscing over Tools of Change and some of the people involved in the last great wave of digital change in publishing, it made me realise how formative many of those conversations were for me, and for the industry as a whole. An open question: what is the equivalent forum for people building AI-forward publishing businesses now? It feels as though we need one.
Later in the conference, I appeared on a panel chaired by another O’Reilly alumnus, Joe Wikert. Last year, I posted a great observation from Tobias Zwingmann that there are two kinds of customer feedback: fake interest sounds like, “that’s really interesting, we should definitely explore this further”; real interest is “how do I get this?” This came to mind as I was telling Joe about my forthcoming book and he got out his phone and preordered the Kindle edition on the spot, possibly the first person to do so. Thank you Joe.
One of my fellow panelists, Bob Hutchins, had a particularly good frame for thinking about generative AI: the idea of it as a proxy failure. Writing is a proxy for something that we cannot see or experience directly: the letter of condolence or good wishes to someone we are not with that indicates we are thinking of them, or the book that serves as a proxy for the author’s thought and effort. The challenge is that AI makes it easy to simulate that proxy at near-zero cost, and therefore to undermine the trust we place in it.
Bob comes at this from a communications and culture perspective. Relating this to a domain that I’m more familiar with, it made me think of Nobel economics laureate George Akerlof’s Market for Lemons. Essentially, this was the idea that if there is a high chance that a used car is a lemon, buyers will price that probability into what they are prepared to pay for any vehicle. Vendors of good cars hold off on selling them because they can no longer achieve the price they wanted. Eventually, only lemons remain, and the market collapses because the signal (a car for sale) no longer reliably indicates quality.
Books don’t behave quite like this: content quality can be previewed, and reputation still matters—particularly in specialist areas of publishing like ECPA members or university presses. But on the other hand, AI enables a theoretically infinite supply of lemons, and even a small increase in uncertainty about what is real or earned risks lowering trust in the signal overall. The question isn’t whether parts of the market will flood with lemons—they will. It is whether publishers can credibly signal quality in ways that are costly for lemon-farmers to fake. Editorial judgement that’s visible, processes that can be explained, curation that stakes reputation on choices rather than just production efficiency.