One of the most common attributes that connect people and AI systems is the ability to detect patterns. Sitting next to a regional CSO at an event last week, we started pulling apart each presenter’s content, swiftly identifying each title, image and sentence as AI slop. The sentences were short. Constantly defining tensions between opposing ideas: speed vs. depth, scale vs craft, data vs. creativity. Not because the tension was real, but because it sounds like strategy—where the f*** is that extended hyphen coming from? It’s obvious, and it’s jarring.
How did we get here?
In the world of marketing, creative thinking has been under attack for a long, hot minute. In recent years, strategy has taken a back seat to data flow charts, platform performance metrics, versioning, and optimisation loops that favour the predictable over the distinctive. It’s no surprise the AI systems we’re now using are stuck in the same performance loop.
General-purpose LLMs are trained to tell you what you want to hear, statistically consistent with the patterns in their training data, and usually the one most likely to “feel” right.
That’s useful for lots of high-touch, low-value tasks, such as summarising research, editing emails, and rewriting headlines. Suddenly, everything becomes punchy, polished, platform-perfect prose…
Strategy is about uncovering what’s distinctively true about a brand and turning it into an advantage no one else can exploit. You don’t get there by averaging the internet. You get there by creating the conditions for something left-field to emerge. Cadbury Gorilla didn’t come from someone prompting for “high-attention confectionery creative with emotional resonance.” It came from a brand truth, a feeling, and a leap.
The problem is, the most “probable” answer to a brand challenge is by definition the most generic one. Standardised audiences and territories that sound the same are interchangeable for everyone in the category.
Harvard Business Review recently tested this across multiple strategic scenarios. Different briefs. Different contexts. Same answers. They called it “trendslop”—coherent, well-structured, and completely detached from anything uniquely true. We’ve trained the system to reward familiarity and it’s feeding it straight back to us.
We’re living in a world that is being exponentially homogenised by algorithms. It’s not lifting the work; it’s making it look and feel the same.
Specialist agents to the rescue
Agents help because they more closely reflect how real strategic thinking actually works. Specialist agents outperform a single model because no decent strategist approaches every problem the same way. No decent strategist looks at audience, culture and category through the same lens. They’re completely different problems, with different inputs, different frames, and different ways of judging what’s actually good.
When you force a single model to do it all in one pass, it just smooths everything out. Averages it and sands off anything sharp or specific. It’s not a team of experts, but one overconfident generalist having a crack at everything and calling it “strategy."
The more specific and focused the task, the better the output, but, more importantly, in an agentic system, it becomes far easier to orchestrate with other specialists to break down, build, and properly complete complex work in collaboration.
The over-reliance on prompting
One of the less talked-about problems with general AI for strategy is how much it relies on prompting. The burden sits with the user and you literally get what you ask for. The quality of the output is limited by the person writing the prompt and the patterns that are surfaced.
That’s fine when you already know what you’re looking for, though it falls apart in strategy, where the whole point is to surface something you might not be aware of. Prompting tends to surface what’s obvious with clean, logical, black and white answers that look right, but rarely push into anything interesting.
What’s needed are systems trained to spot the less obvious signals; the edges, the tensions, the things that wouldn’t be asked for in the first place. Strategy isn’t about retrieval but lateral structured thinking. The methodology shouldn’t live in the prompt; instead, it should be embedded in the system, with consistent thinking regardless of who’s running the brief.
AI to sharpen, not to replace
The goal isn’t AI that replaces thinking, but using it as a tool to sharpen it. Strips out the grunt work and brings the right context into the moment. Leave the actual craft to the people who know what they’re doing.
In practice, that means moving beyond a single model trying to do everything, toward systems made up of specialists, different lenses applied to audience, culture, category and creative that are working together while constantly improving, and building on what’s come before.
It means connecting the dots properly. Qualitative and quantitative signals, brand knowledge and past performance all work together, not sitting in separate decks unused and being applied directly to the work in front of you as a clear, actionable direction.
The result of these systems would be sharper thinking, earlier, with fewer loops and less rework. More ideas that actually make it through.
In a world of AI-driven output, the advantage isn’t more of it. It’s having the context, clarity and consistency to make better decisions, faster.
Ben Crawford is the co-founder and chief commercial officer at Brainwaves.
Source: Campaign Asia-Pacific
