Are We Entering a New Age of Creativity with the Help of AI?

Harvard Business Review · article · 2026-07-07 source ↗ Reader 🖥️

Is AI Creative? (Pattern Recognition vs. Genuine Creativity)AI Industry Litigation AI Labor Disruption Authenticity Premium (Human Signal in an AI World)Credibility Gating Institutions Learning Culture & Slack Leverage & Positional Advantage Technology Diffusion & Transformation Curves

Laurene Powell Jobs's Atlantic-OpenAI content-licensing deal exemplifies the correct strategic response to generative AI for creative institutions: treat AI licensing as a short-term revenue and access play while doubling down on human-curated 'meaning' and top-tier creative talent, because genuine creativity — unlike AI's pattern recognition — remains scarce and is what audiences will still pay for.

practical · mixed — A Cold Call podcast interview (HBS Professor Caroline Elkins with host Brian Kenny) about her case study 'The Atlantic and OpenAI.' Structurally narrative/expository — the case is told as a story about Laurene Powell Jobs's decisions — but it carries a practical argument (how legacy media and creative institutions should respond to generative AI) built on a theoretical claim (what AI is and isn't capable of, creatively). The Reader content already contains the full dialogue transcript, not just show notes, so no separate transcript resolution was needed.

Structure

Part	Role
Frame: the March 2024 Atlantic-OpenAI deal and the journalists' backlash ('a devil's bargain'), set against the NYT-OpenAI lawsuit	context
Powell Jobs's dual identity (literary patron via the Emerson Collective / Ralph Waldo Emerson naming, and technologist) and the 'business of ideas' framework: idea → meaning → scale → value	conceptual scaffold
Short-term rationale: licensing revenue, privileged model access, and feeding the model to avert 'collapse' from stale training data; the S-curve strategic logic (Kodak/Blackberry as cautionary cases)	evidence
Long-term rationale: the creativity crisis (300,000-subject Torrance-test decline since the 1990s), the bet that AI won't replace highly creative long-form journalists even as it absorbs short-form/commodity news copy	evidence / argument
Cultural backlash inside the organization and the shift in creator compensation from back-end IP royalties to front-loaded salaries/advances	complication
Fair use discussion: the property-seizure analogy, the unreliable-attribution problem, and the claim that people still pay for the interpretive 'meaning' layer	argument
Meritocracy and gatekeeping: idea-to-scale requires 'meaning,' which requires curators; the funnel of who gets published narrows even as gatekeepers gain power	argument / conclusion
Big-picture reframing: the 'age of efficiency' (Taylorism/scientific management), of which AI is 'the last logical chapter,' giving way to an 'age of creativity' that requires the time and patience efficiency destroys (Amabile's kryptonite-to-creativity claim)	argument
Comparison to other billionaire media owners (Bezos/Washington Post, Sulzberger/NYT) with different motives and interventionism	generalization
Closing endorsement: Powell Jobs as visionary for combining love of literature with technological clear-eyedness	conclusion

Key terms

the business of ideas
The economic activity of creating ways of thinking, spreading them, and capturing value, intention, and influence from them — treating ideas as raw material with a whole business built around them.
meaning (as a supply-chain stage)
The interpretive, curatorial value a publisher or writer adds between a raw idea and its mass-scaled distribution — 'idea to meaning to scale to value.' The stage that can be stripped out of commodity news but not out of durable, monetizable creative work.
S-curve Technology Diffusion & Transformation Curves
A strategy-playbook metaphor for the adoption/maturity trajectory of a technology or business model; 'getting on the next S-curve' means proactively adopting the next disruptive wave rather than being disrupted by it, as Kodak and BlackBerry failed to do.
fair use
As Elkins glosses it in this interview: the current practical stance of AI companies that essentially all copyrighted creative work is usable to train models without licensing or payment — contrasted with the traditional, strictly enforced IP-protection norm (e.g., the Orwell estate).
age of efficiency / age of creativity
Two successive historical eras: a century-long 'age of efficiency' descended from Taylorist scientific management (of which AI is framed as the culminating chapter — automating humans out of efficiency work), giving way to a hoped-for 'age of creativity' that rewards highly creative people and requires patience/time that efficiency logic destroys.

Propositions

explicitAI, in its current form, is not creative — it is pattern recognition, not creativity in the way a human is creative.conf: high
explicitAI will absorb and replace short-form journalism and commodity news copy, but will not replace highly creative, long-form journalists for the foreseeable future.conf: high
supportsMartin Fowler & Kent Beck: Frameworks for reinventing software, again and again · p6 — Same pattern in a different domain: commodity/'middle' work is exposed to AI displacement while differentiated, high-skill creative work is comparatively insulated.
explicitAI systems require a continuous supply of new, high-quality material or model quality degrades ('collapses'), which gives producers of original high-quality content ongoing negotiating leverage with AI labs.conf: medium
explicitThe 'supply chain of ideas' runs idea to meaning to scale to value; the interpretive meaning stage cannot be stripped out of a venture without it eventually failing, even though it can be stripped out of commodity daily news.conf: high
extendsYou Probably Won’t Read This Article…and That’s OK · p3 — The Atlantic's bet on curated 'meaning' as the monetizable layer is a concrete instance of a society inventing a new gate once the prior cost-of-content-production barrier collapses.
explicitAs AI-generated and synthetic content proliferates, audiences will increasingly pay specifically for trusted human interpretation and curated meaning rather than for raw information.conf: high
supportsAdam Mosseri: AI is a tailwind for authenticity · p20 — Both argue that abundant AI/synthetic content raises, not lowers, the premium audiences place on trusted human creators and curated interpretation.
explicitdevelopedAI's current effect on creative output is to homogenize and elevate less-creative people toward a higher floor, while highly creative people continue to consistently outperform ('beat') the machine.conf: medium
extends ←How to turn your AI into a world-class designer · p1 — Supplies a specific training-mechanism explanation (RLHF-style human-rated next-token prediction converging on safe/average choices) for the homogenizing effect that the HBR/Atlantic-OpenAI doc asserts empirically ('AI's current effect on creative output is to homogenize and elevate less-creative people toward a higher floor').
explicitAs AI floods the information ecosystem with content, curatorial gatekeepers (editors, trusted publishers) gain more power, not less, even as the pool of who actually gets published narrows.conf: high
extendsYou Probably Won’t Read This Article…and That’s OK · p5 — Provides a concrete case (the Atlantic under Powell Jobs) of AI flooding an information domain and a human-judgment editorial gate consolidating power in response, rather than being displaced.
explicitGenuine creativity requires time and patience, which are structurally opposed to the century-long drive toward organizational efficiency that AI now represents the culminating chapter of.conf: high
implicitCreators' compensation is shifting from back-end IP royalties toward large upfront payments (salaries, advances) in exchange for surrendering exclusive control over how their work is used to train AI.conf: medium
explicitUnder AI companies' current practice, most creative and copyrighted work is treated as fair use — usable without licensing or payment — which functions, in effect, as uncompensated appropriation comparable to seizing someone's property without paying for it.conf: high
implicitInvesting outsized resources — well-above-industry-standard salaries — in a small number of exceptionally creative individuals is a strategic bet that positional leverage, not raw effort or content volume, is what compounds value in the AI content economy.conf: medium
supportsThe Laws of this World (Game Theory) · p2 — Powell Jobs's above-market salaries for a small number of top long-form journalists is a concrete application of investing in leverage/position rather than raw effort or volume.

Arguments

Publishers that preserve and monetize the interpretive 'meaning' layer — rather than just producing or licensing raw content at scale — will survive and even thrive in an AI-saturated information economy.

Explicit premises:

The idea-to-value supply chain requires an intermediate 'meaning' (interpretation) stage between raw idea and mass distribution.
Outlets or creators that skip straight from idea to scale without investing in meaning fail (the Buzzfeed case).
Audiences keep paying for full articles and books even when AI can regurgitate summaries, because they want the insightful interpretive layer, not just the raw information.

Implicit premises:

AI cannot currently produce the interpretive/meaning layer at the quality of highly skilled human writers, because AI is not itself creative (depends on the pattern-recognition-vs-creativity proposition).
Reader trust and attention are scarce resources that only credentialed, accountable human institutions can currently earn (a claim about source reputation as the last credible information filter).

Inferences:

If meaning becomes the primary monetizable layer, and the gatekeepers who curate 'meaning-makers' consolidate power, the AI-licensing strategy structurally advantages a small, already-successful class of publishers and star writers — a concentration effect the interview surfaces ('the funnel is getting more and more narrow') without ever naming it as a downside or interrogating who is squeezed out.

Narratives

The hypothetical of scaling a brilliant writer's ideas (Jill Lepore) directly to mass, low-friction distribution ('the Buzzfeed of Jill Lepore') without an interpretive layer — and Buzzfeed's actual failure as the real-world referent.
→ Implicitly supports the claim that the 'meaning' stage of the idea-to-value chain cannot be skipped (p4) and grounds the interview's central business thesis.
Elkins's own 'crossing the river' anecdote: sitting in Harvard's arts-and-sciences faculty meetings debating whether to ban AI outright, then walking across the river to the business school, where colleagues talk about AI solving world peace.
→ Illustrates and legitimizes her positioning as a synthesizer between two poles ('somewhere down the middle'), and frames the interview's overall 'it's complicated, not utopian or dystopian' stance.
The hypothetical Elkins poses to a room of real-estate executives: would you accept the government seizing your property without payment?
→ Makes vivid and emotionally persuasive the claim that AI training on copyrighted work without payment functions as uncompensated appropriation (p10).
The George Orwell estate's aggressive IP enforcement (any quotation triggers a licensing-fee demand from his lawyers) as the pre-AI norm, set against AI companies' current blanket fair-use posture.
→ Sharpens the sense that AI represents a break from an established IP-protection norm, by contrast.
Contrast among billionaire media owners: Powell Jobs (Atlantic, non-interventionist, self-described commitment to literature and democratic process) versus Bezos (Washington Post, more interventionist, shifting posture across the two Trump administrations) versus Sulzberger (NYT, litigating rather than licensing, leaning on complementary assets like Wordle).
→ Implicitly argues that an owner's underlying motive and values — not the technology itself — determine whether an AI partnership sustains or undermines an outlet's editorial independence and journalism quality.

Questions addressed

✅ Is AI creative?
No — not in the way a human is; it is pattern recognition, though it can elevate less-creative output toward a higher floor.
⭕ Should legacy media partner with AI companies or sue them?
Elkins clearly favors partnering-and-adapting (holding up Powell Jobs as the model), but explicitly frames it as a live, contested question — she notes her own MBA and exec-ed students often argue the opposite ('I would sue them to protect my IP... there are different schools of thought').
⭕ What happens to the writers and journalists who are not part of the small, highly-paid, top-creative tier that outlets like the Atlantic are now investing in?
⭕ Does AI's homogenizing effect on less-creative people's output change what a healthy information ecosystem or a healthy definition of 'creativity' looks like at scale, beyond the individual writer/outlet level?

Gaps

incompleteThe interview never follows up on what happens to the writers and journalists who are not top-tier creative talent and are being 'weeded out' as legacy outlets like the Atlantic double down on a smaller roster of highly paid stars. Elkins states plainly that Powell Jobs 'is weeding folks out' and that the publishing 'funnel... is getting more and more narrow,' but the whole argument stays optimistic about the shrinking population of winners without examining the losers' fate, alternative markets, or what this means for journalism's overall labor market and coverage breadth (e.g., local news, beat reporting). This is the sharpest gap: an argument about creative-industry winners built on an unexamined and possibly large population of creative-industry losers.
Read next: Labor-economics and media-industry literature on winner-take-most/superstar dynamics in creative markets, and reporting on the collapse of local and freelance journalism under both the advertising-model crisis and AI displacement.
uninformedThe claim that AI 'elevates' lower-creativity people and homogenizes ideas is asserted via unnamed 'MIT' and 'other' studies, with no citation, sample, methodology, or effect size given. Given the persona's stated bar for AI-and-creativity claims (practitioner/benchmark data over hype), this is exactly the kind of unsourced empirical claim that should be treated as a lead to verify, not a settled finding.
Read next: The actual MIT (and related) research on generative AI and creativity/ideation variance — e.g., Doshi & Hauser's work on AI and human creativity — to check whether the 'homogenizes below-average, doesn't help top performers' finding replicates and under what task conditions.
incompleteNo engagement with attribution and compensation mechanics at scale beyond a single anecdotal figure (a ~$3,000 Anthropic authors' settlement payout, split with the publisher). There is no discussion of how AI-licensing revenue is actually distributed to individual journalists versus captured by the outlet or owner, nor any comparison of deal terms across the '70+ similar deals' the piece mentions were struck between publishers and AI companies by March 2024.
Read next: Reporting comparing the actual terms of publisher-AI licensing deals (News Corp, Axel Springer, Condé Nast, the Atlantic, etc.) and how much of that revenue reaches working journalists versus owners/shareholders.
illogicalThe central 'meaning cannot be commodified out' argument rests on one negative case (Buzzfeed's failure) and one still-unresolved positive case (the Atlantic, whose own outcome Elkins repeatedly disclaims — 'we'll see where it goes... I have no crystal ball'). No comparative base rate of publisher-AI licensing deals succeeding versus failing is offered, so the argument's generalizability is asserted rather than demonstrated.
Read next: Longitudinal tracking of publisher-AI deal outcomes (traffic, revenue, subscriber retention) as more of the '70+' deals mature past their first year or two.

Closes gaps elsewhere

The latest AI news we announced in May 2026 · g3 — That gap flags the unaddressed question of publisher/open-web economics under generative AI — licensing, attribution, impact on the publisher economy. This document is a direct, detailed case study of exactly that: a real publisher-AI content-licensing deal (the Atlantic and OpenAI), its financial terms, motives, and internal fallout.