Strategist & writer · brands, AI & commercial culture
Useful fictions for an age of machines.
Fifteen years at the intersection of consumers, business and creativity — and a decade spent trying to turn “cyborg creativity” from a vision into the actual work. Essays, not prophecy. Somewhere, as ever, in the middle.
Writing
WARC · 2026 · sample formatting
Scraps, Straws & Silicon
AI is not an apocalyptic threat to brands, nor a magic wand. It is the ultimate clarifying force — and it changes not just the environment of the decision, but the decision-maker itself.
For thirty years, every wave of technology has quietly rearranged the room in which a brand is chosen. Search changed where the choice happened. Social changed who was in the conversation. Programmatic changed the speed. But the chooser stayed the same: a human being, with all their glorious inconsistency, doing the deciding. That is the assumption AI breaks.
Because now there is a second audience in the room. Alongside the messy, emotional, culturally-embedded human sits an unflappably logical, infinitely knowledgeable personal shopper — the agent, the recommendation engine, the model that has already formed a statistical opinion of what your brand stands for.
In a world of infinite content, consistency is no longer repetition — it’s recognisability under variation.
The machine is only ever holding up a mirror. If it can’t describe your brand cleanly, that is usually because you haven’t. Fickle brands make patchy training data; coherent ones become almost a platonic ideal of themselves.
The five shifts
- Guidelines → guardrails. From rules that forbid to systems that steer.
- Static assets → modular ingredients. Built to recombine, not to be filed.
- Tone of voice → character sheets. Design the brand the way you’d design a mind.
- Control → stewardship. The marketer as editor and curator, not author.
- Broadcast reach → delegated relevance. Being the brand the agent reaches for.
Where this goes wrong. Chase the machine too hard and you build something optimised and hollow. Automate the craft away and you lose the distinctiveness the model was learning from. And when everyone can generate “good enough,” good enough stops being good enough at all.
AI doesn’t replace brand building. It exposes whether you’ve done it properly in the first place.
A formatting sample showing how a full essay reads on the site. Every piece in the archive is set to the same template — paste the full text and it renders like this.
About
Oliver Feldwick is a strategist and writer, currently Chief Innovation Officer and Strategy Partner at The&Partnership, and AI & WPP Open lead for JLR Global. He has spent fifteen years at the intersection of consumers, business and creativity — and since 2016, when he wrote Unleashing Cyborg Creativity, has been trying to turn that vision of human-plus-machine work into practical reality.
He is widely published and award-winning: an Admap Prize, an Atticus Award, an IPA Excellence Diploma distinction, and a place on Adweek’s 2025 Agency All-Stars for his work deploying AI. He has led strategy and new business on accounts including JLR, Coca-Cola, Amazon Audible, NatWest, EA, Mars and Google, and helped build embedded in-house agency models at scale.
A working timeline
Shelf & Marginalia
The things that actually shaped the thinking — not a reading list to look clever, but the writers, games, records and shows that keep turning up in the work. A living shelf; pull something down and it’ll usually explain an essay.
Marginalia
Lexicon
Named ideas do the heavy lifting. A running glossary of the coinages and borrowed frameworks the writing keeps returning to — each one earned, not decorative.
The Matchbox Machine
In 1961 Donald Michie built a machine from matchboxes and coloured beads that learned to play a game — no code, just consequence. This is a version of his idea. You move first; the machine moves second and remembers only its defeats, pruning each losing line until, eventually, it stops losing. The whole argument of the essays, in miniature: not human versus machine, but a small, honest demonstration of what learning actually is.