Why We Don’t Just Use AI — We Run on It.

Cannes Lions 2026 received 20,050 submissions — a 25% drop from 26,900 the year before. The festival chair, Phil Thomas, confirmed the reason: stricter integrity rules, mandatory proof-of-impact requirements, and compulsory executive endorsements. The new rules followed a 2025 withdrawal scandal in which an agency’s entry was found to contain AI-generated manipulation. In response, Cannes made AI disclosure mandatory. Forty per cent of 2026 entries disclosed AI use — double the proportion from 2025, nearly four times 2024.

In the same week, OpenAI made its Cannes debut. Chief Revenue Officer Denise Dresser did not pitch a feature. She pitched a reframe: “AI isn’t a tool anymore. It’s the operating system.” Her argument: the agencies and brands that will win are not the ones that added AI to their existing workflows. They are the ones that redesigned their workflows around AI.

Both signals point in the same direction. In 2026, the question is no longer whether an agency uses AI. It is what the phrase actually means — and whether a client can verify it. For Malaysian and Singaporean businesses choosing an AI marketing agency, that distinction is now a material one.

What the Cannes Drop Reveals

A 25% fall in entries is significant. However, it is not a sign that agencies are doing less. It is a sign that the industry is being forced to prove more. Cannes’s new integrity standards require documented proof of impact — not projected results, not testimonials, not AI-generated metrics dashboards. Real, verifiable, executive-endorsed outcomes.

The result: entries backed by genuine work held. Entries that relied on inflated claims or opaque AI processes fell away. Phil Thomas noted that in Creative Data specifically, “a huge amount of it was about, can you prove some of the claims that you’re making.” That scrutiny is now the standard, not the exception.

For Malaysian and Singaporean businesses, this Cannes signal matters because the same scrutiny is moving into client relationships. Clients are increasingly asking what “AI-powered” means in a proposal. They are looking for the difference between an agency that uses an AI tool and an AI marketing agency Malaysia that has genuinely structured its delivery around AI. Those are not the same thing.

OpenAI’s “Operating System” Framing: What It Actually Means

Dresser’s Cannes message was specific. OpenAI is not asking agencies to run AI as a standalone step in their workflow — brief the AI, get copy, review it, move on. That is using AI as a tool, and it produces incremental efficiency, not structural advantage.

What Dresser described is something different: multi-agent orchestration, where specialized AI systems handle research, strategy development, creative production, and optimization simultaneously rather than sequentially. “Instead of isolated AI tools performing individual tasks,” as Marketing Magazine Malaysia reported, “agencies may soon orchestrate multiple specialized AI agents across research, strategy, creative development, production and optimisation.”

The key word is “orchestrate.” An operating system does not do one thing. It coordinates many things simultaneously. An agency running AI as an operating system has connected those specializations into a working system — one where the research feeds the strategy, the strategy shapes the creative, the creative is tested and optimised in parallel, and the output is produced at a quality and pace no sequential human workflow can match.

For MY/SG clients evaluating agencies, this reframe provides a useful lens. An agency using AI as a tool will describe the tools it uses. An AI marketing agency Malaysia running AI as an operating system will describe the system — how its components connect, what each one does, and where the human layer adds judgement that AI cannot provide.

The Two Camps: Tool vs Operating System

The distinction between using AI and running on AI is not a marketing difference. It is an architectural one. In practice, it plays out in three ways.

Speed and parallelism. A tool-using agency runs tasks sequentially: write the brief, use AI to draft, review, revise, produce. An AI-operating-system agency runs tasks in parallel: research and strategy happen simultaneously, creative variants are produced and pre-screened while briefing is still being finalised, optimisation data feeds back into production in real time. The output per hour of human time is not incrementally better — it is structurally different.

Institutional memory. When AI is a tool, each conversation starts fresh. When AI is an operating system, the system builds persistent knowledge: what has worked for this client, what the brand voice requirements are, what the audience has responded to, what the competitive context is. Each new campaign starts informed by everything that came before. For clients on monthly retainers, this compounds significantly over time.

Human QC as a multiplier. This is where Cannes’s integrity signal connects directly to operational reality. AI operating systems are powerful precisely because they produce at scale. However, that scale requires a commensurate human oversight layer — not to slow the output down, but to ensure that what ships is accurate, on-brand, legally sound, and client-ready. The agencies Cannes’s new rules disadvantaged were not those using AI heavily. They were those using AI without that oversight layer. The agencies that thrive are those where human QC multiplies the quality of AI output rather than replacing it.

What Xwork’s System Actually Looks Like

Xwork operates a multi-agent ecosystem. Each agent in the system owns a specific domain: research and competitive intelligence, SEO and content strategy, paid media, organic social, email and lead nurture, creative direction, analytics, copywriting, reporting. These agents do not operate in silos. They feed each other — competitive intelligence informs content strategy, content strategy informs social scheduling, social performance data feeds back into content planning, ad performance informs landing page optimisation.

The human layer — Reno’s quality control and strategic oversight — sits above this system. It does not duplicate what the agents do. It provides the judgement, the client relationship, the contextual calibration, and the accountability that AI systems cannot provide. Every deliverable that leaves Xwork has passed through that layer. The agents produce. The human layer decides.

This is what OpenAI described at Cannes. It is also what Cannes’s integrity rules reward: AI as a production infrastructure, human expertise as the accountability layer, verifiable outcomes as the proof. Xwork has operated this architecture since its inception. This is, in precise terms, what CODE/RAVEN delivers — not individual tools applied to individual tasks, but an integrated system that compounds across every protocol. To see how that system is structured, visit coderaven.my.

Why Human QC Matters More, Not Less, in an AI-Run Agency

A common concern about AI-heavy agencies is that human expertise is being diluted. In practice, the opposite is true — when the architecture is correct. As AI production capacity increases, the value of accurate human judgement increases alongside it. An AI system producing 50 campaign variations is only valuable if the human reviewing it knows which five to use and why. An AI system drafting 20 blog articles is only valuable if the human editing them can distinguish expert insight from plausible-sounding noise.

Cannes’s 2026 integrity rules made this explicit. Forty per cent of entries used AI. The entries that won were not the ones that used AI least — they were the ones with the most rigorous human-oversight layer. As Xwork’s own analysis of AI-assisted versus AI-generated content demonstrates, the human QC layer is not a compromise. It is the variable that separates content that builds brand authority from content that merely fills a calendar.

For MY/SG clients, this means the right question is not “how much AI does this agency use?” It is “where does the human expertise sit in this agency’s process — and can they show me?” An AI marketing agency Malaysia with a credible answer to that question is offering something materially different from one that describes AI as a feature.

The Questions to Ask Any “AI-Powered” Agency

The Cannes integrity rules now provide a useful template for client due diligence. Before signing with any agency making AI-powered claims, ask these questions:

What is the human oversight layer? Who reviews AI output before it reaches the client? What are their qualifications? At what stage in the process does human judgement apply?

What AI systems are connected — and how? An agency with an AI operating system can describe how its research, strategy, creative, and production functions connect. An agency using AI as a tool can describe the tools it subscribes to. The answer reveals which category the agency belongs to.

Can you show a verifiable outcome? In the Cannes framing: can you prove the impact, with evidence an independent observer could validate? For marketing services, this means organic traffic data, lead volume, revenue attribution, or similarly concrete metrics. Impressions and engagement rates are a starting point, not a proof.

What does the output look like six months in? An AI tool produces consistent output from day one. An AI operating system produces improving output over time, because institutional memory compounds. Ask to see how a client’s results have developed across a retainer engagement, not just at the first deliverable.

For more on what Xwork’s system produces for clients across content marketing, see the content marketing ROI framework for Malaysian businesses.

The AI Marketing Agency Malaysia Checklist: What to Look For

The Cannes integrity rules now provide a useful template for evaluating any agency claiming AI-powered delivery. For MY/SG clients, these questions are becoming standard due diligence.

Ask about the system, not the tools. “What AI tools do you use?” is the wrong question. The right question is: “How do your research, strategy, creative, and production functions connect?” An agency with an operating system can describe the connections. An agency using a tool can name the subscription.

Ask where the human layer sits. Every output that reaches a client should have passed through a defined human review step. If the answer is vague — “we review everything, of course” — press for specifics. Who reviews? At what stage? What criteria do they apply? These questions separate accountability from assurance.

Ask for a before-and-after on an active client. Not a case study from three years ago. A live account where you can see results across a retainer period. An AI marketing agency Malaysia that is genuinely running on AI will have compounding results over time, not a strong first month followed by a plateau.

Ask what the AI cannot do in their system. Agencies that have actually built AI systems know their limits precisely — the steps that still require human judgement, the questions AI cannot reliably answer, the decisions that need contextual knowledge AI does not have. Agencies that claim AI can do everything have not actually built a system. They have a sales script. The most credible answer to this question is a specific, honest list of where human expertise remains irreplaceable.

This Year, the Architecture Becomes the Story

OpenAI’s Cannes message was not aimed at agencies. It was aimed at clients. The argument: stop evaluating agencies by their tool stack. Start evaluating them by their operating model. That shift in framing will move through client conversations in MY/SG over the next twelve months, as the Cannes signal ripples into the Southeast Asian market.

For Xwork, this is the year to make the architecture visible in public-facing content. Not as a sales pitch — as a demonstration. Clients who understand what an AI operating system looks like will be better equipped to evaluate the agencies claiming to offer one. And the agencies with the architecture will benefit from that scrutiny rather than being harmed by it.

We don’t use AI as a tool. We run on it. And the Cannes integrity bar now exists precisely to separate those two things from a distance. If you want to understand what that means for your specific marketing objectives, book a strategy call and see the system in action.