Kieran Flanagan
SVP of Marketing, HubSpot
On giving up AI tools for writing and building an audience
Our takeOpinion
The SVP of Marketing at HubSpot raises what happens when a time-poor creator builds an ever more elaborate AI system for producing content and finds the work drifting away from the enjoyment and originality that made the output worth publishing. A reader can take from it a useful split between AI as a helper around the edges (research lookups, structural and pacing feedback on a finished draft) and AI as the author of ideas and sentences, which the piece says left a channel flat until the hosts went back to topics they actually wanted to discuss. The split applies most to executives, founders and small content teams whose audience follows a recognisable human voice, and to internal documents where fluent but empty memos can pass for thinking. Our view is that this rests on one experienced creator's own channels and a single show, with no controlled comparison, so it may not transfer to high-volume programmes, product documentation, or brands where no individual voice carries the audience.
Written from the text of the post as published, which we do not store. How we write these
Why this is here
A Voices entry passes four tests: the post makes a specific claim, a reader at another company could use it, it is the author's own work, and it does not sell the author's employer. Two separate model reviews have to agree that it passes before it is listed.
The label and the take are ours. The take is the editors' opinion, written only from what we read and named in the line under it; it says what a reader can use and where the post stops short. The role shown comes from our own list of people, checked against a page that states it.
More Voices
- Kieran Flanagan: On how AI search summaries are reshaping content marketing strategy for companies (21 October 2026)
- Kieran Flanagan: On hiring marketers with entertainment and creator backgrounds as AI commoditizes content (16 October 2026)
- Kieran Flanagan: On the problem of AI-generated content sounding generic and interchangeable (2 October 2026)