Last week, a court in China handed down a ruling that should have landed harder than it did in marketing boardrooms across Asia. I was surprised this news was not circulated within my tech friends' circles here in Bangalore (India), arguably the GCC capital of the world, that is still reeling with the fear of AI-cited RIFs.
A technology company tried to replace a quality assurance supervisor with a large language model, offered him a 40% pay cut when he resisted, and fired him when he refused. The court ruled the dismissal illegal. AI adoption, the court said, is a deliberate business choice and not an emergency or an act of God. A decision made by management, for management's reasons, cannot be dressed up as an unforeseeable circumstance to justify breaking employment contracts.
Read that again. AI is a business choice, and I am with the Chinese court on this.
That framing ought to make marketing and advertising executives in our region deeply rethink their perspectives on AI. Between 2023 and 2025, Meta, Snap, Amazon, Microsoft, other tech players and all advertising majors from WPP, Omnicom, IPG, Dentsu slashed hundreds of thousands in headcount and did not frame AI as a choice. They framed it (or were forced to frame it) as an inevitability, and used that framing to justify layoffs that were investor-driven cost compression dressed in the language of tech-transformation.
The Kool-Aid we were sold
The story went like this: AI would automate functions wholesale. The companies that moved fastest would win. The ones that hesitated would be disrupted. Cut the headcount, deploy the models, and the productivity gains would follow.
A lot of marketing leaders in Asia bought it or felt pressure to be seen buying it.
The problem is that the story was being told largely by the people selling the infrastructure. Hyperscalers, frontier model providers and a few ‘cutting edge, AI-led’ MarTech vendors all had the same interest: get orgs to automate at scale, because scale meant volume, and volume meant revenue. The replacement narrative served them perfectly.
The companies that cut deepest are the ones now quietly rebuilding the institutional knowledge and client relationships. The judgment that told you whether the AI output was actually right. The strategist who knew when the brief was wrong before the work began. You cannot prompt your way to that. It lives in people and that is the part that got cut. And according to Forrester, more than half the employers who cut it are now saying they shouldn't have.
Talent vs. tokens
Let us talk about what tokens actually are. In the AI economy, tokens are the unit of cost - every prompt, every inference, every automated output has a price attached to it. The question is not whether to spend tokens. It is whether you are spending them wisely. And whether you have kept the human layer that knows what good output looks like.
This is the ratio that matters, as talent without tokens is slow. And token without talent is noise.
A marketing team that lays off its brand strategist and replaces the function with an AI model has not become more efficient. It has lost its calibration instrument. The model did not get smarter. The organisation got blinder. There is nobody left to look at the output and say - This is right, or " This is wrong, and here is why.
The CMOs who are getting this right are not the ones who spent the most on AI. They are the ones who asked a different question: which human functions produce judgment that a model cannot replicate, and which functions produce volume that a model can handle cheaply? The answer to the first question tells you what to protect. The answer to the second tells you where to deploy.
The real tokenomics
Here is the part of the AI story that did not get enough attention in the boardroom presentations: the price of tokens is falling faster than almost anyone predicted. And the biggest price correction did not come from Silicon Valley.
DeepSeek, the Chinese AI lab, has positioned its V4-Flash model at a disruptive $0.28 per million output tokens. For context, OpenAI’s flagship GPT-5.5 charges $30.00 for the same volume. And that's a 107-times price gap for a model that rivals top Western systems on high-volume tasks like coding and classification. Qwen, Baidu's ERNIE, and the broader wave of Chinese open-weight models tell the same story: 75 to 85% of frontier model quality at 10 to 15% of frontier model cost.
The implication for Asian CMOs is significant and underappreciated. You do not need to buy at the frontier to operate at the frontier. A thoughtful deployment of mid-tier and Chinese open-weight models across the right functions, from content production, data summarisation, creative variation, and campaign reporting, gets you the majority of the capability at a fraction of the cost. The infrastructure argument for large AI spend is weaker than it has ever been.
Spend your tokens intelligently, so pick the model that is right for the task, not the model that is most impressive in the vendor presentation. It means deploying cheaply at volume and reserving human judgment for the decisions that actually matter. It means knowing that the tokenomics of AI are moving in your favour—if you are patient enough to let them. And this is where Asian CMOs have an opportunity that is hiding in plain sight.
The message for Asian CMOs
The reflex to cut functions and call it transformation, the whole RIF culture, never took deep roots in APAC the way it did in the West. Our restraint was read in some quarters as hesitation. It was not. It was a different relationship with talent, with continuity, with the understanding that marketing functions are built from accumulated knowledge that does not live in a model.
We must not abandon that instinct now. We must not let investor pressure from San Francisco dictate headcount decisions in Singapore, Bangalore or Bangkok. The Chinese ruling named something important: technological adoption is a choice. Choices carry responsibility and we should fully own ours.
Be aggressive about AI deployment and use the affordable models, including the Chinese ones, which are now genuinely competitive. Automate what is repetitive, the templated, the high-volume, low-judgment work but spend your tokens wisely and cheaply.
But we must keep the people who know what the function is actually doing. Keep the ones who can look at an AI output and say, "this is right," or "this is wrong," and here is why. Because that judgment is what separates a marketing organisation that uses AI from one that is used by it.
Talent vs. tokens is not a competition. It is a ratio and getting the ratio right is key to winning and the CMOs who do will be the ones who look vindicated in two years, while everyone else is quietly rebuilding what they cut.
Ramakrishnan Raja regularly writes for Campaign Asia-Pacific. He is a principal at Resonant.
