In digital advertising, the most expensive decisions often look small on the surface. A headline variation, a new video cut, a different call to action. Creative work used to hinge on instinct and manual reviews; now, advertisers expect systems that can trial ideas quickly, measure the result, and turn what works into a repeatable advantage.
Surya Narayana Kalipattapu, Principal Product Manager at TikTok, operates squarely in that world. As the author of the HackerNoon article “Generative AI Is Transforming the Advertising Industry, a Guide for Product Managers,” he focuses on building AI products where experiments are not side projects but core behavior, so advertisers can treat every campaign as a chance to learn, not just a chance to spend.
---Advertisement---
Experimentation That Starts With Real Advertiser Friction
---Advertisement---
That philosophy begins with a simple premise: experiments only matter if they solve real pain for the people buying media. The global market for A/B testing tools is projected to reach $850.2 million in 2024, growing at about 14.0% a year through 2031, as more teams formalize testing instead of relying on gut calls. At the same time, more than 50% of marketers already use A/B testing to boost conversion, which turns experimentation into a basic expectation rather than an optional extra. When testing becomes part of everyday performance work, platforms that cannot run clean experiments fall behind.
At TikTok, Kalipattapu built TikTok’s first AI-Generated Creative Ads product from that starting point. He and his teams ran design sprints with advertisers who were struggling with creative fatigue and localization costs, then wired those findings directly into a global experimentation pipeline. Early pilots shipped only after controlled tests showed a 12% lift in click-through rates for AI-generated creatives compared to human-only baselines. He still recalls late-night review sessions with a retail client where the team walked line by line through test dashboards until everyone understood why a seemingly modest variation was winning. The experiments were not theatrical; they were a way to earn trust.
“Experiments should feel like a safety net, not a gamble,” says Kalipattapu. “When advertisers see side-by-side results, they stop arguing about opinions and start asking better questions about what to try next.”
Scaling AIGC Ads From Pilot To Global Program
As those early tests proved out, the question shifted from “does this work” to “can this scale without losing control.” Nearly90% of marketers have now used generative AI tools at work, and 88% report using AI in their day-to-day roles. That adoption curve means experimentation is no longer limited to copy tests or landing pages; entire creative pipelines are being rebuilt around AI assistance. The platforms that matter are the ones that turn that appetite into structured, safe systems rather than scattered experiments in individual teams.
On that backdrop, Kalipattapu laid out a two-year roadmap for TikTok’s GenAI Ads strategy and then delivered against it. The product moved from a May 2025 pilot with a few commerce and retail advertisers into a North America launch, then into EMEA and APAC with localized asset generation and performance-aware ranking. Within the first three months of general availability, more than 500 advertisers across commerce, retail, and gaming had onboarded to the system, using it to ship AI-generated creatives that were directly tied to targeting and bidding signals. Engineering, data science, design, and policy teams worked on the same playbook, so each new feature was evaluated as a testable hypothesis, not a feature for its own sake. For many advertisers, the shift was practical: they could ship more variants in less time without losing a handle on what was actually working.
“Scale should feel like more control, not less,” notes Kalipattapu. “When the roadmap is built around experiments and guardrails, every new market becomes another chance to learn faster rather than another risk to manage in isolation.”
Connecting Experiment Lift To Revenue And ROAS
Once experiments reach that scale, they must prove they are doing more than generating dashboards. Internet advertising revenue reached $258.6 billion in 2024, up 14.9% year over year, which makes small percentage lifts worth very real money. At the same time, the AI in marketing market was valued at $20.44 billion in 2024 and is projected to reach $82.23 billion by 2030, as more budgets flow into AI-enabled targeting and creative tools. For advertisers facing quarterly targets, experiments that only move proxy metrics are no longer enough; they need to see a clear impact on revenue and return on ad spend.
Kalipattapu treated that as a design constraint. In TikTok’s AIGC Ads product, every creative variant flowed through ranking systems tuned to CTR, conversion rate, and cost, and nothing scaled without evidence of downstream lift. Between May and September 2025, those experiments contributed to $250 million in incremental advertiser revenue, while commerce campaigns saw double-digit improvements in return on ad spend when they leaned into AI-generated creatives that had passed controlled tests. He later expanded those lessons in his research article “Data-Driven CICD for AI PM: Analytics-Powered GenAI Delivery Pipelines,” which examines how continuous testing and instrumentation keep AI releases accountable. Instead of positioning GenAI as a novelty, he framed it in tough internal reviews as a way to make media budgets more accountable: if a variant could not prove durable performance across markets and cohorts, it did not move past the experiment stage. That mindset aligned TikTok’s internal teams and advertiser success partners on what success actually meant.
“Lift is the only language that holds up in a quarterly review,” states Kalipattapu. “If an AI-powered idea cannot show revenue impact, we treat it as a draft, not a result.”
Guardrails That Keep AI Experiments Safe For Brands
As experiments touch more creative surface area, risk rises alongside opportunity. In 2024, EU data protection authorities imposed €1.2 billion (~$1.395 billion) in GDPR fines, bringing the total value of penalties to €5.88 billion (~$6.83 billion) since the regulation took effect, with technology and social media companies bearing much of the impact. Brand owners remain cautious about AI-generated content: only 40% currently use such assets in marketing, even though many intend to adopt them. For any AI-generated creative product, that combination of enforcement and caution makes governance as important as raw model quality.
Kalipattapu built TikTok’s AIGC Ads product with that reality in mind. His teams implemented automated copyright and IP filtering, advertiser usage-rights checks, and review workflows that tied AI-generated assets to clear terms of use. A transparency dashboard helped advertisers understand why the system suggested particular creatives, which variants were winning, and how safety filters were applied. Internal legal reviews ensured alignment with copyright standards and industry guidelines, while engineers worked on efficiency so that governance did not slow down experimentation. Advertisers saw roughly 30% reductions in creative production costs when they used AI-generated variants as a first draft, yet they still retained the choice and control needed to protect their brands. For many of them, that balance between speed and safety was the difference between experimenting in a sandbox and committing meaningful budgets.
“Trust is built in quiet details like rights screens and logs,” explains Kalipattapu. “If brands know the system respects their constraints, they are much more willing to push into bolder experiments.”
Experiments As Everyday Product Behavior
As these systems mature, the economics around them continue to shift. The generative AI in marketing market is expected to reach about $19.1 billion by 2030, while the broader AI for sales and marketing market is projected to grow from $57.99 billion in 2025 to $240.58 billion by 2030. Across industries, generative AI software spending is forecast to approach $220 billion by 2030, with retail and e-commerce capturing a large share of that value. In that context, experimentation is not a specialist function; it is becoming standard product behavior for any platform that touches media budgets.
Kalipattapu’s trajectory aligns with that future. He continues to apply the same discipline he brings to experimentation in his broader leadership, including his role as a Globee Awards Judge for Leadership, where he evaluates programs on measurable impact rather than promises. His goal at TikTok is clear: treat experiments as a living part of the product, so advertisers can move faster without guessing and AI systems can prove their value in numbers, not adjectives.
“We are still early in learning what AI-generated creative can do,” reflects Kalipattapu. “If we keep the experiments honest and the guardrails real, growth follows.”