From Agency to Creative Engine
My argument was that the traditional agency model — departments, job titles, long timelines, pitch-for-everything — was breaking under pressure from multiple directions: clients building internal teams, AI automating the doing, and creative talent demanding flexibility over hierarchy. I proposed that the future agency would be productised, embedded, flexible and IP-led — selling outcomes and systems, not hours and deliverables.
What actually happened
The directional argument has been validated, comprehensively. AI adoption is no longer early-stage — McKinsey's 2025 State of AI research shows 88% of respondents reporting regular use across at least one business function, and that normalisation has accelerated exactly the pressures I identified. But the 2026 evidence reveals a twist the original framing didn't fully capture: the bottleneck isn't production. It's judgment.
Production has been commoditised, but value hasn’t. AI can generate content at scale. It can create variants, personalise messages, draft copy. What it can’t reliably do is decide what’s worth making, verify what’s safe to ship, or maintain distinctiveness across hundreds of assets. That’s editorial judgment — and it’s the scarcest resource in the 2026 market.
Scaling without governance is expensive. Gartner predicts that at least 30% of GenAI projects will be abandoned after proof-of-concept, citing poor data quality, inadequate risk controls, escalating costs and unclear value. The real cost of AI is the data readiness, compliance infrastructure and governance overhead that most organisations underbudget for.
Homogenisation is a real and measurable risk. Academic research published in 2024–25 provides the evidence base that was largely absent in 2025. Studies on text-to-image AI show that while productivity and peer-evaluated value can increase, average novelty tends to decline. Research on LLM-generated writing shows that each additional AI essay contributes fewer new ideas than human writing, with the diversity gap widening at scale. A 2025 meta-analysis confirms the pattern: GenAI can improve task performance, but idea diversity drops sharply.
The “creative engine” metaphor I used in 2025 still holds. But the engine’s most important component is the ability to choose: what to make, what to kill, what to verify, and how to stay coherent when production is essentially free.
One of the strongest market signals of this shift is the WPP/Adobe expanded partnership announced in early 2026. It’s explicitly about orchestrating the content supply chain end-to-end, keeping content “on-brand from the start” with models trained on a client’s own IP, and deploying what they call “creative AI forward-deployed engineers.” Instead of selling hours, that’s a system selling governed output. Exactly the direction I predicted — but with governance and editorial authority at the centre, not just speed and flexibility.
What this means now
For any creative or digital team: the question to ask isn't "how do we produce more?" It's "how do we decide better?" That means codifying taste into constraints (voice libraries, do/don't patterns, prompt guardrails), installing editorial gates (factual checks, bias review, brand distinctiveness), and rewarding reduction (fewer variants shipped, higher confidence decisions).
For healthcare and pharma: you already have a structural analogue for this. MLR review is, at its core, an editorial gate. The question is whether it's been updated for the AI-enabled content pipeline. If AI is generating or iterating materials that enter review, the process needs to account for AI-specific risks. If it hasn't been updated, you have a governance gap that's widening every time someone uses a generative tool.
Verdict: Confirmed
The prediction was right in direction and largely right in the specifics. The creative engine model is happening. What the 2025 analysis underestimated was how quickly the “engine” would be defined not by its production capacity, but by its editorial judgment. The teams that decide best — not produce most — are the ones winning in 2026.