Generative AI has dramatically reduced the cost of producing words, images, ideas and variations. But abundance creates its own economic problem: when almost anyone can generate more content, producing something is no longer necessarily the difficult part.
Deciding what deserves to exist may be.
That tension sits at the centre of Dr. Morissa Schwartz’s work. A writer, editor, publisher, marketing strategist and AI evaluator, Schwartz operates across two worlds that generative AI is increasingly forcing together: the traditional disciplines of authorship and editorial judgment, and the emerging systems through which machines generate, evaluate and refine language.
Schwartz is the founder of GenZ Publishing and Dr. Rissy’s Writing & Marketing. A Forbes 30 Under 30 honoree with a Doctorate in Literature and a Master’s in Communications, her career spans publishing, ghostwriting, public relations, marketing and digital media alongside work involving LLM evaluation, prompt and rubric design, reasoning, tone and language quality.
Her writing and commentary have appeared across major media, while her work and social presence have built an audience of more than half a million followers. But the particularly interesting intersection today is between her experience deciding what humans should publish and her work evaluating what machines can produce.
Because generative AI introduces a strange inversion into creative economics.
For centuries, publishing was constrained at almost every stage. Writing took time. Editing required people. Production had costs. Distribution was controlled. Professional presentation demanded infrastructure. Even producing a credible first draft required significant human effort.
AI weakens several of those constraints simultaneously.
A single person can now generate drafts, alternatives, summaries, marketing copy and editorial variations at extraordinary speed. Small publishers and independent creators can access capabilities once associated with much larger organisations. Production moves towards abundance.
But attention does not.
Neither does credibility. Nor taste. Nor accountability.
Schwartz argues that this changes where creative value accumulates. If production becomes cheap, judgment becomes expensive. If plausible language becomes abundant, provenance matters more. If machines can reproduce the statistical characteristics of a writer’s style, then having an actual point of view becomes more—not less—important.
And simply putting a human somewhere in an AI workflow does not necessarily solve the problem.
A person clicking “approve” after an AI system produces something is technically human-in-the-loop. Schwartz argues that meaningful human oversight demands much more: expertise, authority, evidence, intervention and ultimately responsibility for the decisions being made.
The emerging question is therefore not simply whether creative work was made with AI.
It is whether a human made consequential decisions about what the AI produced.
We spoke with Dr. Morissa Schwartz about infinite content, human judgment, authorship, publishing scarcity, the difference between voice and point of view, and what happens when “human-in-the-loop” risks becoming an accountability label rather than an accountability system.
If content becomes effectively infinite, what actually becomes scarce?
Trusted attention. We are not short of things to read, watch, or hear. We are short of reasons to believe a particular thing deserves our time.
Judgment becomes scarce because judgment requires exclusion. It means deciding what not to publish, which claim is too weak, which sentence is merely competent, and which idea adds something useful. Provenance, credibility, editorial coherence, and the courage to omit become more valuable too. In an age of infinite output, the rarest creative act may be saying, “This does not need to exist yet.”
Production can be automated. Earned trust cannot be mass-produced on command.
What tells you that human judgment genuinely shaped AI-assisted work rather than merely approved it?
I look for decisions that cannot be explained by polishing alone: a clear thesis, surprising but defensible exclusions, a hierarchy of evidence, deliberate structure, and moments when the writer resists the easiest or most statistically familiar answer.
Human-shaped work carries consequences. The author can explain why a claim is present, why another was removed, which sources were trusted, and what changed during review. I care less about whether a human clicked “approve” and more about whether the human challenged the premise, corrected the framing, checked the facts, and materially changed the result.
Source notes, version history, fact checks, documented reversals, and a named decision-maker reveal real intervention. If the human only removes typos and obvious hallucinations at the end, that is not meaningful oversight. It is signing the delivery receipt.
Where is the boundary between reproducing a writer’s voice and possessing a point of view?
Style is a pattern. Point of view is a position.
A model can reproduce sentence length, vocabulary, cadence, and favourite metaphors. A point of view comes from experience, values, memory, commitments, blind spots, and stakes. It determines what a person notices, what they refuse, and what they are willing to defend or revise publicly.
AI can imitate the sound of conviction, but it does not bear the consequences of being wrong. It can resemble a writer’s voice without inheriting the life that formed it. I see AI as useful for testing, translating, organising, and refining ideas. The line is crossed when assistance becomes counterfeiting, attribution disappears, or a recognisable identity is simulated without meaningful consent.
Which forms of publishing scarcity does AI destroy, and which does it make more valuable?
AI weakens the scarcity of production, first drafts, routine copyediting, formatting, and basic packaging. It lowers the operational barrier to making something that looks publishable. That can be liberating for independent authors and small presses like GenZ Publishing because good ideas no longer need enterprise-sized machinery to reach a professional baseline.
It does not eliminate the scarcity of attention, trust, distribution into meaningful communities, rights clarity, or excellent developmental judgment. It makes those things more valuable. Editorial work shifts away from mechanical correction and toward selection, context, ethics, originality, and long-range development.
Publishers increasingly earn relevance by choosing well, improving deeply, building trust, and standing behind the work. I grew up in my family’s small business, and the lesson applies here too: the storefront is not the business. The relationship and reputation are.
Who reviews the reviewer, and how do we stop “human-in-the-loop” from becoming an accountability label?
The human needs more than a seat in the diagram. They need relevant expertise, sufficient time, clear standards, access to evidence, and the authority to stop or escalate the work. If the reviewer cannot disagree with the system or delay release, the human is decorative.
Reviewers should be reviewed through sampling, second-reader checks for high-risk material, disagreement tracking, incident analysis, blind-spot rotation, and periodic external or cross-functional audits. Organisations should measure interventions, not merely approvals. What did the reviewer change, reject, escalate, or send back? Which failures slipped through, and what changed afterwards?
For smaller organisations, responsible governance does not require a bureaucracy worthy of a Russian novel. A practical system can include risk tiers, named ownership, source requirements, a short checklist, and a second set of eyes for consequential claims. “Human-in-the-loop” should describe evidence of accountable decisions, not the presence of a pulse.
The Economics of Judgment
There is a useful distinction running through Schwartz’s answers: production and creation are not necessarily the same thing.
Generative systems are extraordinarily effective production machines. They can create grammatically competent text, reproduce styles, generate alternatives and reduce the operational cost of reaching something that looks finished.
As that capability becomes ubiquitous, however, technical competence becomes less differentiating.
A reasonably polished paragraph was once evidence that someone possessed a certain degree of writing ability. A professional-looking publication suggested access to editors, designers, production systems and distribution. AI progressively separates the appearance of professional production from the infrastructure and expertise traditionally required to produce it.
That does not make expertise irrelevant.
It changes where expertise becomes visible.
The editor’s value moves from correcting sentences towards deciding which sentences matter. The publisher becomes less differentiated by the ability to manufacture a book and more differentiated by what they select, improve and ultimately put their reputation behind. The writer’s value shifts away from simply generating language towards developing a perspective worth expressing.
This is why Schwartz’s distinction between style and point of view matters.
Style can increasingly be modelled as a pattern. Sentence length, vocabulary, rhythm, recurring expressions and structural tendencies provide signals that generative systems can reproduce with growing sophistication.
Point of view is harder to reduce to those characteristics because it involves stakes.
A person has experiences behind an argument, values informing it, and consequences attached to defending it. The machine can generate the linguistic appearance of conviction without experiencing the consequences of that conviction.
That difference becomes especially important as synthetic content becomes harder to distinguish from human work.
The relevant question may eventually cease to be “Was AI used?”
AI will be embedded in too many creative tools and workflows for that question alone to tell us very much.
More useful questions emerge instead:
Who made the consequential decisions?
Who selected the evidence?
Who rejected the plausible but weak argument?
Who decided something should not be published?
And who is prepared to stand behind the result?
The same logic applies to human oversight of AI systems.
“Human-in-the-loop” has become one of the reassuring phrases of the generative AI era. But the mere presence of a person somewhere in a workflow says remarkably little about the quality of oversight.
Schwartz proposes a more demanding standard: measure interventions, not approvals.
That is a useful principle well beyond publishing.
A meaningful human-in-the-loop system should leave evidence of human agency—changes, disagreements, escalations, rejected outputs and decisions to stop. Without the authority and expertise to intervene, the human can easily become what Schwartz describes as “decorative.”
And that brings the argument back to scarcity.
Generative AI gives us more.
More drafts. More images. More alternatives. More variations. More content.
The economic and cultural response to that abundance may therefore be greater demand for something AI cannot generate simply by increasing its output: reasons to care about one thing rather than another.
In that environment, perhaps the defining creative capability is no longer production.
It is judgment.
Or, as Schwartz puts it more sharply:
“In an age of infinite output, the rarest creative act may be saying, ‘This does not need to exist yet.’”
More Interviews
Building Creative Machines has published dozens of exclusive conversations with founders, researchers, executives, artists and other leading voices shaping artificial intelligence and its impact on society.
Editorial Disclosure
This interview was conducted independently by Building Creative Machines. No payment, sponsorship, or other editorial consideration was received in connection with its publication. The views expressed are those of the interviewee.