If AI predicts what will work, should creatives listen?

Every few months, a new dashboard promises to tell agencies whether their next big idea will land before a single rupee is spent on media. Predictive AI tools that scan social conversations, ad libraries and consumer sentiment at a scale no planner could match are increasingly being pitched to Indian creative agencies as insurance against the one thing this business has never been able to guarantee: audience response. The pitch is seductive: test the idea before you shoot it, spend less on campaigns that were never going to work, and put more behind the ones that might.

But advertising was not built on certainty. It was built on the industry's willingness to back an idea nobody could prove would work, from a jingle with no precedent to a film that broke every convention of its category. So, when a machine claims it can tell a creative director what the market will respond to before the market has even seen the work, the real question is not whether the tool is accurate. It is whether an industry that has always run on instinct should let a probability score anywhere near the room where ideas are approved or killed.

The numbers suggest the shift is already well underway. Globally, 41% of marketing and advertising decision-makers said they planned to use AI for campaign activation in 2025, up from 31% the year before, according to DoubleVerify's 2025 Global Insights report, based on a survey of nearly 2,000 industry professionals across Asia Pacific, Europe, Latin America and North America. Closer to home, Adobe's 2025 AI and Digital Trends India study found that 23% of Indian businesses were already reporting measurable returns from generative AI, the highest share in the Asia Pacific region, with executives crediting faster content ideation and greater production speed as the biggest gains.

Separately, Forrester research cited by WPP's global chief technology officer earlier this year found that three in four ad industry executives said their companies were using generative AI tools in 2025, up from 61% a year earlier. The appetite for prediction, in other words, is not a Silicon Valley preoccupation alone. It has arrived squarely on Indian agency floors, inside media plans and, increasingly, inside pitch decks.

When agencies backed the gut
There was a time when the pitch itself was the risk. Agencies walked into boardrooms willing to lose the account because they believed in a script, a line, a piece of casting that no test could have validated in advance. That willingness to be wrong in public is, depending on who is asked, either the founding myth of Indian advertising or its last surviving instinct.

Rohit Dhamija, Founder and Creative Director at Goodness of Design and Digital, a boutique creative and design studio, remembers it as neither myth nor nostalgia but method. “Once upon a time, agencies were willing to take the risk of losing a client because they believed in their instincts,” he said. “Experimentation was welcomed. We were taught that our job as creatives was to find a solution, not seek approval.” He pointed to a run of campaigns that became part of the industry's folklore precisely because nothing like them had existed before, work that made audiences uncomfortable before it made them loyal, custom-built for a moment rather than assembled from what had already tested well elsewhere.

The detective and the forensics lab
What predictive AI tools are genuinely good at is compression. A human planner scanning hundreds of hours of content across platforms, languages, and formats to spot an emerging pattern would take weeks to do it properly; AI can do a version of that work in hours, surfacing a spike in conversation before it becomes obvious to anyone watching a single feed. That speed changes when a brand notices a shift, not necessarily whether it understands what the shift means.

Rachita Goel, Strategy Director at 22feet Tribal, draws a sharp line between the two. “AI can get you early signals, but it can’t predict culture on its own,” she said. “We like to think of it this way: humans are the detective, AI is the forensics lab. The lab can process evidence at scale, but it can’t decide what the evidence means without the detective’s judgment.” According to her, setting AI to track early indicators is not prediction in any mystical sense; it is simply spotting a trend's earliest shape, after which a human still has to interpret the context and decide which two or three signals are likely to actually break out.

That distinction matters more in India than in most markets, if only because of scale. Sundeep Sehgal, Senior Vice President and Executive Creative Director at VML India, argues that the country's diversity is precisely what makes cultural prediction unreliable. “India is not one culture. It is many cultures moving at the same time: different languages, regions, age groups, internet habits,” he said. “Something can trend in Coimbatore and not matter in Chandigarh. AI can detect rising conversations and unusual spikes, but it cannot fully understand the context of why people care, what emotion is driving it, or whether it will fade in two days.” His formulation is blunt: AI can see signals, it cannot judge significance, and it is far stronger at explaining what is happening than why it is happening or what happens next.

Built to reward the familiar
The more uncomfortable problem sits underneath the accuracy debate altogether. Every prediction model is trained on precedent, on what has already worked, which means it has no real frame of reference for the idea that has never been tried. A campaign that looks nothing like anything the model has seen will, almost by definition, score lower than a safe, familiar one, regardless of whether it is actually the better idea.

Gopa Menon, Co-Founder and Chief Operating Officer at TheBlurr, a brand strategy and innovation consultancy, makes the case plainly. “Prediction tools are trained on what's worked before. So, by design, they'll always score a safe, familiar idea higher than a genuinely new one, because the new one has no historical pattern to match against,” he said. “An agency that leans too hard on these scores ends up optimising for what won't fail instead of what could break through. Those aren't the same thing.” His warning points to a specific failure mode: AI, he said, is reasonably reliable at flagging an idea that will underperform, but far weaker at identifying the one that will break out, since breakout ideas by definition do not resemble anything the model has already seen.

Dhamija sees the same mechanism from the agency floor rather than the model's architecture, and describes its cultural cost in blunter terms. “AI is built on historical data and patterns of consumer behaviour,” he said. “But advertising is not built on what has already happened. Advertising is built on new ideas. And that is where the danger lies.” He worries less about AI itself than about what it lets agencies get away with.

Talent, in his view, is the real shortage behind the prediction obsession: when an agency does not have enough people with the instinct and the courage to give an idea the right direction, a low prediction score becomes a convenient excuse to play safe, and data ends up validating mediocre thinking rather than challenging it. He points, by way of example, to how unexpected slogans and a clear sense of what was at stake have repeatedly cut through entrenched messaging in India's own public discourse, arguing that no volume of automated content can substitute for people who understand exactly what they are fighting for.

Dashboards are hygiene, not strategy
None of this has stopped the demand for these tools from growing inside client conversations. If anything, the request has become routine enough that agencies now build it into the pitch itself.

“Yes, and rightly so,” said Sehgal. “Marketing today runs at internet speed. Clients want early signals, volatility alerts, meme tracking, regional pulse checks.” But he separates the tool from the value an agency actually adds on top of it. “The smartest clients aren't asking what's trending,” he said. “They're asking, 'What should we do about it. Dashboards are hygiene. Interpretation is advantage.” His sharper worry is less about accuracy and more about conformity: once every brand in a category is reading the same trend dashboard, the industry's real risk stops being a wrong call and starts being sameness, since a hundred brands reacting to the identical signal tend to end up saying the identical thing.

Goel has heard the same appetite from clients, though she frames it as a genuine operational gain rather than a threat. “We've been meeting clients who are increasingly curious about this, but what they're most excited about is the speed and agility benefit that these dashboards provide,” she said. “They're cutting down research time and shrinking the lag between a trend emerging and a brand spotting it.” What clients are really asking for, by her account, is not certainty but a head start: enough context and confidence to decide whether to lean into a moment, sit it out, or have a response ready before the conversation is even fully formed.

Where that leaves the industry is less a resolved question than a live one, and neither side of it is arguing that the tools should be switched off. The disagreement is over how much weight a probability score should carry once it reaches the room where the actual decision gets made, and how many agencies still have the conviction to overrule it when it says no.

Dhamija, for his part, is unwilling to let AI carry that decision at all. “AI should be used as a tool, not as a creative conscience,” he said. “It can analyze, challenge, test, and identify patterns. But the decision to break a pattern still has to come from creatives, planners, strategists, and ultimately people who understand that culture doesn't always move according to historical data.” Agencies of the past, he argues, were not built on efficiency; they were built on unforgetfulness, and that is precisely the quality a dashboard cannot manufacture. “So to answer the question,” he said, “good agencies will listen to AI. The best will choose when to listen and when to take creative risks.”