"Innovate When You Can, Not When You Must"

Darryll Colthrust had barely started his new job — he was due to officially join Macmillan Publishers as Chief Technology Officer on April 7 — when he took the keynote stage of the dpr and Publishing Perspective's digital conference ai@media to deliver what he called "a reminder." His message was simple, and he repeated it throughout his talk: innovate when you can, not when you must.

Colthrust opened by pushing back on the notion that AI in publishing is a recent phenomenon. Holtzbrinck, the privately held media and science group spanning fifteen thousand employees across one hundred and twenty countries, has been running AI and machine learning projects for years. When the group formally established Chaptr — an independent AI specialist firm operating within the Holtzbrinck ecosystem — in October 2022, it was codifying something that had already been quietly building: through global hackathons run as virtual events with physical hubs around the world, yearly visits to the West Coast to stay close to companies like Google, OpenAI, and Anthropic, and a newly established AI lab in San Francisco.

The point was not to boast but to argue that sustained, incremental investment creates something that cannot be acquired in a hurry: an organisational capacity for change.

Colthrust was candid about what has not worked. Chaptr launched an illustrated storytelling product at Google I/O in 2023, only to see a new model arrive shortly after and commoditise its core differentiator. A second early project, White Mirror, pushed into AI-generated short film at a moment when the technology was far from ready. Both came too early — but Colthrust framed them as essential: they revealed the limitations of the models and ultimately accelerated the team's judgment about where and when to deploy AI.

Chaptr's current approach applies a strict two-month scoping rule to all internal projects: anything that cannot be meaningfully delivered within that window is deprioritised. No project begins without a formal commitment of time and resource from the relevant business unit. The pace is deliberate — it keeps stakeholders engaged and prevents the common failure mode where a technology project loses its internal champion halfway through.

Much of the success, Colthrust argued, comes down to a principle he distinguished carefully from the fashionable rhetoric around "humans in the loop": experts in the loop. Getting a manuscript summary that meets editorial standards — "the one percent of the one percent" — requires people with deep domain knowledge to be actively involved throughout, not consulted after the fact. When they are, outputs improve and trust in the system grows.

Colthrust's sharpest remarks were reserved for the traditional publishing posture of watchful patience. "If you take the posture of waiting and seeing, you are absolutely moving back as fast as the market is moving forward." The reference point, he stressed, is not simply what other publishers are doing — it is the democratisation of the underlying technology itself, which is lowering barriers for entirely new entrants. Publishers need to be watching adjacent industries, not just their traditional peers.

Looking ahead, Colthrust sees the greatest opportunity in the convergence of publishing's content assets with an expanding range of modalities. As models become capable of assessing and transforming audio, video, and other formats, the definition of what a publishing company can offer its authors, readers, and marketers broadens considerably — and opens the door to entirely new kinds of creators beyond the traditional author.

Organisations that innovate from a position of strength have options; those that wait until they are forced to move do not. "A new experience has come into the market, and you have no choice about what you want to be giving to your customers — because that's already been defined for you by somebody else."

He will have the chance to test that thesis at Macmillan from April.