AI Makes Creative Production Cheaper. Companies Respond by Producing More Than Ever.
Every major AI tool launch was supposed to reduce headcount. The data says the opposite. Open engineering roles in the US are at their highest level in over three years — and the curve bends upward precisely where AI milestones appear on the timeline. The mechanism has a name: the Jevons Paradox. And it applies to creative production just as much as it applies to code.
- Engineering job postings are up 78% from their 2024 low, reaching 67,665 open roles globally
- Microsoft, Anthropic, and Citadel are all hiring more engineers, not fewer, as AI capabilities advance
- The same dynamic is hitting creative teams: cheaper production doesn't mean less production — it means more
In January 2025, after the Chinese startup DeepSeek unveiled an AI model that matched OpenAI's capabilities at a fraction of the cost, Microsoft CEO Satya Nadella posted a single phrase on LinkedIn: "Jevons paradox strikes again."
He was referencing a 160-year-old observation by English economist William Stanley Jevons. In 1865, Jevons noticed that as steam engines became more efficient, Britain didn't burn less coal. It burned dramatically more. Efficiency didn't suppress demand. It detonated it. Cheaper energy made new applications viable, and the economy expanded into territory that had been economically impossible before.
Fourteen months later, the data confirms Nadella was right — and the pattern extends far beyond software engineering.
67,665 Open Engineering Roles, and Climbing
TrueUp's tracking data, visualized in Lenny Rachitsky's newsletter this month, shows a striking chart. Software engineering job postings collapsed from over 100,000 in early 2023 to a low of 37,982 by mid-2024. Then they reversed. Hard. As of March 2026, there are 67,665 open engineering roles globally — a 78% increase from the trough.
The inflection points are telling. The curve begins bending upward around the launches of Claude Code and Opus 4.5 — precisely the tools that were supposed to make engineers less necessary. Instead, they made software cheaper to build, which made more software projects economically viable, which created more demand for the people who define problems, architect systems, and manage complexity.
Citadel's 2026 Global Intelligence report found that software engineer postings are up 11% year over year, even as overall job postings across the economy are flat or falling. Anthropic — the company that builds the AI tools automating code — has 448 open positions, 146 of them in software engineering. They grew from 500 employees to over 1,000 and plan to reach 2,000+.
The displacement narrative got the first act right and the second act completely wrong. AI did automate the repetitive, narrowly scoped coding tasks. But the falling cost of execution made more ambitious projects viable. Companies are not building the same amount of software with fewer people. They're building things they couldn't have justified before.
The Same Paradox Is Playing Out in Creative Production
Now apply the same logic to content, campaigns, and creative assets.
Generative AI tools have made it dramatically cheaper and faster to produce visual assets, copy, video variants, and campaign adaptations. The initial assumption — shared by consultants, CFOs, and holding company CEOs — was that this would reduce creative headcount. Produce the same output with fewer people.
That's not what's happening. What's happening is that brands are producing more. More formats. More variants. More localized versions. More test-and-learn iterations. The surface area of creative output is expanding because the unit economics changed — exactly as Jevons predicted for coal, and exactly as the engineering data now confirms for code.
Forrester's 2026 agency predictions forecast a 15% reduction in agency jobs. But read the detail: the roles disappearing are execution-heavy, repetitive positions. The roles growing are the ones that involve judgment, coordination, integration, and strategic decision-making — the skills that become scarcer as AI handles more of the mechanical work.
The parallel is exact. In engineering: writing for-loops is automated; defining the right system architecture is more valuable than ever. In creative production: resizing banners is automated; deciding what the campaign should say, to whom, in what sequence, is more valuable than ever. The question isn't whether AI replaces the project manager — it's whether human judgment becomes the scarcest resource in the room.
Cheaper Execution Doesn't Reduce Complexity — It Compounds It
Here's where the optimistic Jevons narrative needs a dose of operational reality.
More production means more assets to review, more versions to track, more approvals to route, more stakeholders to coordinate. The bottleneck was never creation. It was — and increasingly is — governance: knowing what was produced, whether it's on-brand, who approved it, and where it shipped.
When a team goes from producing 50 assets per campaign to 200 because AI made it possible, the production cost dropped. But the coordination cost didn't. If anything, it increased. More assets mean more review cycles. More variants mean more version confusion. More localized content means more handoff friction between global and local teams. The organizations that try to do more with less without rethinking the workflow underneath discover that "more" arrives immediately and "less" never materializes.
This is the operating gap that most discussions of the Jevons Paradox skip entirely. The paradox predicts that efficiency expands demand. It doesn't predict that the infrastructure to manage that expanded demand magically appears. In engineering, the answer was CI/CD pipelines, code review systems, and deployment automation. In creative production, the equivalent is production infrastructure that scales with volume: traceable workflows, centralized asset management, structured review, and version discipline that doesn't depend on someone's memory.
Master The Monster exists at this inflection: a creative project management platform where AI-augmented production plugs into governed workflows — briefing, versioning, annotation, approval, delivery tracking — so that the expanded volume AI enables doesn't overwhelm the team's ability to control it. L'Oréal Paris, which runs hundreds of campaigns annually across dozens of markets, doesn't use AI to produce less. They use it to produce more — with the infrastructure to ensure that "more" actually ships on-brand, on-time, and with full traceability. The shift from tools that execute to agents that decide only works when the decision architecture exists.
The Skill Mix Question, Not the Headcount Question
The newsletter spotlight framing is exactly right: "The risk is not in the size of the team — it is in the skill mix."
Organizations planning their creative workforce around headcount reduction are asking the wrong question. The right question is: which capabilities remain structurally valuable as the cost of production drops?
The answer, visible across both engineering and creative production, is consistent. What stays valuable: defining the problem (the brief, the strategy, the campaign architecture). What stays valuable: integrating AI outputs into coherent, brand-consistent work (the judgment layer). What stays valuable: coordinating complexity across stakeholders, markets, and formats (the operational layer). What gets commoditized: execution of narrowly defined, repeatable production tasks.
The organizations that understand this fast enough will gain a compounding advantage. Those that mistake the Jevons Paradox for a reason not to hire will discover, two years from now, that their competitors used cheaper production to do more things — not the same things with fewer people.
FAQ
What is the Jevons Paradox and how does it apply to creative production? The Jevons Paradox, first described in 1865, observes that when efficiency makes a resource cheaper to use, total consumption increases rather than decreases. Applied to creative production: AI makes assets cheaper and faster to produce, so organizations produce more of them — more formats, more variants, more localized versions — rather than the same amount with fewer people.
Are engineering jobs really growing despite AI automation? Yes. TrueUp data shows 67,665 open engineering roles globally as of March 2026, up 78% from the 2024 trough. Citadel reports software engineer postings up 11% year over year while overall postings are flat. Anthropic, which builds AI coding tools, has 448 open positions including 146 in engineering.
Does this mean creative teams won't shrink? The execution-heavy roles will. Forrester forecasts 15% agency job losses in 2026, concentrated in repetitive production tasks. But the judgment, coordination, and strategic roles are growing — because more production volume requires more governance, not less. The risk isn't team size. It's having the wrong skill mix.
What should Creative Ops leaders do differently based on this? Stop planning around headcount reduction. Start planning around volume expansion and the infrastructure it requires. If AI makes your team 3x more productive, the question isn't how many people you can cut — it's what new capabilities become viable and whether your workflow can handle the expanded output without breaking.
Sources
TrueUp / Lenny's Newsletter — Engineering Jobs Are Growing Despite AI (March 2026): https://www.lennysnewsletter.com/
Citadel — 2026 Global Intelligence Report: Software Engineer Demand (2026): cited via Alfa BCN analysis at https://www.alfabcn.ai/post/ai-creates-more-jobs-than-it-destroys
Satya Nadella — Jevons Paradox post on LinkedIn/X (January 2025): https://x.com/satyanadella/status/1883753899255046301
GeekWire — Microsoft CEO says AI use will 'skyrocket' with more efficiency (2025): https://www.geekwire.com/2025/microsoft-ceo-says-ai-use-will-skyrocket-with-more-efficiency-amid-craze-over-deepseek/
Newsweek — Machines Do, Humans Decide: When Jevons Paradox Hits the Workforce (2026): https://www.newsweek.com/jevons-paradox-about-hit-workforce-opinion-11747919
INSEAD Knowledge — Is AI Really Going to Take Your Job? (2026): https://knowledge.insead.edu/career/ai-really-going-take-your-job
Forrester — Predictions 2026: Marketing Agencies (2025): https://www.forrester.com/blogs/predictions-2026-marketing-agencies-resign-their-agency/
