Every impression is now a machine decision. Is that a promise or a problem?

Advertising has always lived in the space between instinct and intelligence. For most of its modern history, it was a craft of calculated bets that included media plans drawn in boardrooms, budgets locked in quarters, creative tested in focus groups before anything went live. The feedback loop was slow, the optimisation was human, and the gap between intention and outcome was wide enough to hide in. That gap is closing fast, and the engine closing it is not a new channel or a new format. It is a new cognitive layer entirely.

Global programmatic ad spends reached $595 billion in 2024 and is forecast to approach $800 billion by 2028, with over 90% of all display ad buying now transacted programmatically. These are not simply large numbers. They represent the near-complete institutionalisation of machine-led decision-making as the default operating model for advertising at scale. What began as automation of the mundane, the buying and selling of impressions through real-time auctions, has mutated into something far more consequential: an always-on, self-correcting, intelligence layer that decides, in milliseconds, who sees what, when, and at what price. The human hand in this process is increasingly upstream, setting goals and guardrails. The machine runs the rest.

This is the defining structural shift in modern marketing, and it is accelerating. The question is no longer whether advertising will be machine-led. It already is. The question is whether the industry truly understands what kind of system it has built, and whether brands, agencies, and platforms are equipped to govern it wisely.

According to Nikhil Kumar, Chief Growth and Marketing Officer for India and Emerging Markets at Affle, “We are entering a world where machines are increasingly interacting with other machines, and the role of advertising systems is to ensure that, within this machine-to-machine ecosystem, brands are still meaningfully connected to real consumers."

This shift underscores how automation is redefining delivery and optimisation, even as the responsibility for meaningful connection stays rooted in human insight.

Dhruv Dhawan, VP Revenue, India at The Trade Desk, explains that AI has reshaped media buying into a probability-led discipline, where every impression is assessed for its potential to drive a specific outcome. Yet the art of persuasion continues to rest with humans—marketers still shape the brand story, creative vision, and messaging that truly resonate with audiences.

The Speed Problem Is Actually a Quality Problem

The real-time bidding ecosystem operates in a window most humans cannot meaningfully perceive. A bid decision is made, evaluated, won or lost, and served in under 100 milliseconds. In that fraction of time, an AI system must evaluate audience signals, contextual fit, bid price, fraud risk, and historical performance outcomes simultaneously. Campaigns today benefit from continuous, real-time decisioning, meaning buying decisions are evaluated impression by impression, a reality that has compressed the optimisation cycle from weeks to seconds.

But speed, as it turns out, is the easier part of the problem. Tejas Rathod, Founder and CTO at Mobavenue AI Tech Limited, is direct about where the real complexity lies. "AI systems separate true growth signals from short-term noise by focusing on signal quality rather than signal volume. In fast-moving categories, a sudden rise in clicks, views, or installs may look positive, but not every spike reflects real customer intent," he says. "At Mobavenue, our AI decision layer processes more than 200 crore data signals every day across the consumer journey. The objective is not to chase short-term spikes, but to identify high-intent cohorts that are more likely to convert and stay engaged."

Rathod's point about response time has a practical ceiling. "We see clear gains where sub-15-millisecond response times help brands engage users in real time and convert demand more effectively. That said, speed alone does not deliver unlimited returns. Once execution is already near-instant, the bigger differentiator becomes decision quality. The strongest results come from combining fast infrastructure with intelligent decisioning, not speed in isolation."

Siddharth Jhawar, Country Manager, Moloco India, frames the same tension from a modelling perspective. "The quality of machine learning models, the ability to consume large amounts of data, and the number of repetitions for improvement determine how well a system can choose the right signals and avoid noise," he explains. "As the system gets tuned and learns over time, it becomes better at these predictions."

India and Southeast Asia are the fastest-growing programmatic markets within APAC, which as a whole is expected to see over 80% of digital ad transactions conducted programmatically by 2026. India's programmatic advertising market, valued at approximately $4.9 billion in 2025, is projected to reach $17.6 billion by 2033, registering a compound annual growth rate of around 17.4%. India's growth is underpinned by affordable data plans, UPI payment ubiquity, and government digital-service portals that continuously onboard new internet users. In this context, the stakes of getting decisioning right are not abstract. They are competitive, financial, and increasingly structural.

The Mid-Funnel Is Not Missing. It Has Been Absorbed

One of the most persistent anxieties in performance marketing circles is that the relentless push toward bottom-funnel outcomes, cost per acquisition, return on ad spend, is hollowing out the middle of the funnel. The concern is legitimate on its surface: if a system is only rewarded for conversions, what happens to brand consideration, reach quality, and engagement depth?

The practitioners building these systems argue the framing is outdated. Dhruv Dhawan, VP Revenue, India at The Trade Desk, puts it squarely: "Brands are not losing visibility into mid-funnel metrics. Rather than treating them as end goals, they are increasingly used as inputs into the optimisation process. AI-powered media buying platforms factor these signals into decision-making but ultimately prioritise outcomes defined by the advertiser, such as conversions or return on investment. This shifts the focus from tracking multiple disconnected metrics to aligning all signals toward a single outcome, making measurement more cohesive while still preserving transparency."

Rathod of Mobavenue echoes this but goes further in describing the architectural shift. "A video view, repeat site visit, product interaction, or qualified reach may not be the final goal, but each helps indicate future conversion value. At Mobavenue, we see mid-funnel behaviour as an important input to growth outcomes, not a secondary metric. Our platform brings awareness, engagement, and conversion data into one decision framework." The implication is significant: the funnel has not been abandoned. It has been internalised inside the model.

As many as 83% of senior brand marketers now use artificial intelligence to target digital ads, and the shift is moving from simply using AI for targeting to letting it control entire campaigns, including real-time adjustments to ad placements, targeting, and spending based on live data and predictive insights. This is not incremental automation. It is a fundamental redefinition of what a campaign is.

Nikhil Kumar identifies a dimension of this shift that is still underappreciated: the rise of machine-to-machine interactions within the advertising ecosystem itself.

"As AI agents begin to act on behalf of users, browsing, shortlisting, even transacting, the surface layer of engagement can quickly become misleading. Not every click, interaction, or signal represents human intent in its purest form."

This is where Affle's model becomes instructive. Kumar explains that the company's Cost Per Converted User framework is built around authenticating intent rather than simply measuring it. "Mid-funnel signals like engagement and consideration don't disappear. They are continuously stress-tested against their ability to translate into verified outcomes. Instead of being standalone KPIs, they become inputs within a larger intelligence layer that is always learning which signals truly matter."

Persuasion at Machine Speed

The deepest question this moment raises is a philosophical one wrapped in a commercial skin: if every impression is now a machine-made decision, has advertising stopped being about persuasion? Has the art form that built the twentieth century's greatest brands become, in the twenty-first, a system of statistical probability management at scale?

Agencies increasingly gain leverage with smarter AI tools, advertisers secure higher campaign performance, and the boundary between ad creatives and media buying is fading into one fluid, AI-driven system. The creative and the algorithmic are converging. But the experts are careful not to declare one victorious over the other.

Jhawar remaked, "Advertising has always been a mix of art and science. While the importance of the science of optimisation and measurement has increased exponentially in the last two decades, the advertiser's customer insights and creativity are also important. It takes a smart marketer to decide and refine the brand positioning, creative strategy, and full-funnel optimisation approach."

Dhawan draws a distinction between the layers of the system. "Advertising is now a combination of both, operating at different layers. AI has transformed media buying into a system of probability-driven decision-making, where each impression is evaluated based on how likely it is to deliver a desired outcome. However, persuasion remains firmly human-led. Marketers still define the brand narrative, creative direction, and messaging that connect with audiences. AI simply ensures that these messages are delivered more efficiently, to the right audience, at the right moment. In that sense, advertising hasn't become less creative. It has become more precise."

Rathod frames it as an evolution of the operating model rather than the underlying purpose. "The shift in advertising is not a replacement of persuasion with probability, but a refinement of how persuasion is delivered. AI now helps determine the right moment, audience, channel, and frequency for each message, using live signals rather than fixed plans. Creative judgment still matters most because data can improve delivery, but it cannot define a brand's story, tone, or emotional value. The real evolution is in execution, which is now continuous, adaptive, and responsive to behaviour as it happens."

India alone recorded over 220 billion programmatic impressions in 2024, largely driven by mobile commerce platforms, a number that contextualises the scale at which these machine-led decisions are already operating in the market. The India programmatic advertising market is expected to grow from approximately $2.29 billion in 2024 to $30 billion by 2035, representing a CAGR of over 26%, making it one of the fastest-growing advertising economies in the world.

Kumar of Affle distils the trajectory cleanly: "Advertising is entering a phase where intelligence, not inventory, is the primary lever of growth. The real shift isn't just toward always-on execution, but toward outcome-anchored intelligence systems that learn continuously from verified human intent. This evolution doesn't dilute the essence of advertising. It sharpens it. Persuasion still drives intent, but AI ensures that persuasion is delivered with contextual precision and validated impact."

The industry built a machine. Now it must learn to think alongside it. The brands that will win in this environment are not those that cede the most to automation, but those that understand most clearly what the automation is actually doing, and why. That distinction, between the discipline of setting the right goals and the mechanics of chasing the wrong signals at speed, may well be the most important competitive advantage of the decade ahead.