How tech companies are using poets, novelists, playwrights, and TV writers to engineer the personalities of AI virtual assistants
Elizabeth Dwoskin / Washington Post :
Context & Ripple Effects
Elizabeth Dwoskin's Washington Post piece captures the moment Silicon Valley decided an assistant's voice was a product feature: tech companies began putting poets, novelists, playwrights, and TV writers on staff to engineer how their AI virtual assistants talk. What reads as a curiosity in 2016 turns out to be the seed of a durable labor market.
Seven years on, the same hiring pattern has industrialized. Data-labeling firms like Scale AI and Appen now recruit writers with humanities backgrounds to lift the literary quality of AI output ([[a:844402]]), and by 2024 a staff writer at a tech company was describing his day job inventing pretend responses to hypothetical chatbot questions to train models ([[a:874743]]). The throughline: humanistic writing skill migrated from marketing copy into model development itself.
First-order effects
- Writers become a recognized engineering-adjacent job category at tech companies, with assistant personalities — tone, humor, restraint — treated as a designed deliverable rather than an emergent afterthought.
Second-order effects
- Data-labeling suppliers like Scale AI and Appen follow the labs' lead and build writer benches of their own, turning literary craft into a priced input in the AI supply chain.
Third-order effects
- As authored personalities meet chatbots that can imitate any author's style — the hollow approximations one writer found in Writer's LLM being the cautionary case — authorship itself splits between those who script machine voices and those whose voices get scripted, pushing toward formal governance of how companion AI personas are built and disclosed.
The trend: Humanities talent is being absorbed into AI development as a core input, evolving from scripting assistant personalities in 2016 to training model voice at scale today.