AI Agents Revolutionize Marketing, But Leaders Stress Caution

The story was originally published on MartechAI.com

At MarTechAI Webinar Series 3.0, leaders from Nykaa, FNP, Furlenco, Jainam, SLIQ and TrueRE said the agentic future is compelling, but autonomy will depend on data maturity, governance, category context and human judgment.

AI agents may be the most discussed idea in marketing technology today, but the road from automation to autonomy is still being built.

That was the central theme of the latest edition of the MarTechAI Webinar Series 3.0, moderated by Brij Pahwa, Editorial Lead, e4m and MarTechAI.com. The session brought together Ambika Paliwal, Head of Marketing, TrueRE, Oriana Power; Deepanath Chinnathambi, Growth, Loyalty and Retention, Nykaa; Mithil Sejpal, Co-Founder, SLIQ; Aishwarya Gupta, Chief Marketing Officer, Jainam; Keyur Zaveri, Chief Design Officer, Furlenco; and Avi Kumar, Chief Marketing Officer, FNP.

The discussion made one thing clear. AI agents are not simply another productivity layer in the martech stack. They represent a deeper shift in how brands may eventually understand customers, make decisions, orchestrate journeys and drive outcomes. But across beauty, gifting, furniture, financial services, renewable energy and lending, the leaders also struck a note of caution: the agentic future is compelling, but brands are still testing where autonomy works and where human judgment remains non-negotiable.

Opening the discussion, Pahwa framed the shift as marketing’s movement from AI-assisted creation to AI-assisted decision-making. Unlike traditional AI tools that wait for instructions, AI agents can observe, plan, execute, learn and optimize with minimal human intervention. The bigger question, he said, is no longer whether AI will impact marketing, but how much autonomy marketers are willing to give it.

From automation to decision infrastructure

For Deepanath Chinnathambi of Nykaa, parts of agentic marketing are already familiar to performance marketers. He pointed out that platforms such as Google and Meta have had autonomous campaign systems for some time, citing examples such as Performance Max and Advantage Plus. In that sense, the concept is not entirely new for marketers.

What is changing now is the extension of autonomy into areas such as creatives, cohorting and segmentation. According to Chinnathambi, the next phase is not just about sending messages efficiently, but about using AI to understand what a customer should be encouraged to do next.

He said Nykaa is moving closer to systems that can look at customer-level data and recommend the next best action. That could mean deciding whether to prompt a customer to buy a shampoo they have purchased earlier or introduce them to a premium product they have not tried yet. Earlier, such decisions were largely heuristic or based on past analysis. Today, agents can process a much wider set of signals.

Chinnathambi described the system as having two broad layers: one that recommends actions and another that executes them. But he was clear that human validation remains important, especially because edge cases can hurt customer experience. Real-time personalization is closer than before, he noted, but cost is still a factor. Running and training systems in real time can be expensive, forcing brands to evaluate trade-offs between real-time models and pre-warmed models.

He also pointed to the rise of AEO, or answer engine optimization, as a new area of focus. Just as brands spent years optimizing for search engines, marketing teams will now need to ensure they are discoverable on AI-led search and answer platforms such as Gemini and ChatGPT.

In B2B, AI agents could compress long cycles

Ambika Paliwal brought a different perspective from the B2B and B2G energy and infrastructure space. She said the marketing context in such sectors is very different from B2C. Decisions are not impulsive and may take several months, or even close to a year.

In energy and infrastructure, marketing has to account for multiple variables, including policy dynamics, macroeconomic changes, geopolitical scenarios, vendor ecosystems and customer-specific requirements. Paliwal said AI agents can help by gathering this information, processing it in real time and giving teams sharper decision points.

For instance, an AI agent could track changing policies, identify new vendors, understand state-level market movements and help business development teams decide where to focus. Work that may earlier have taken months of market study can be compressed significantly.

However, Paliwal emphasized that correctness matters more than speed. In regulated and infrastructure-led environments, information shared with customers or published on platforms must be accurate. For her, the role of humans as checkers and governance layers remains critical.

Looking ahead, she identified account intelligence, personalization and customer lifecycle management as areas where AI agents will play a larger role. In renewable energy, she noted, the customer relationship does not end with a contract or even with delivery. It may continue for 25 to 30 years. AI agents can help organizations proactively identify customer needs and strengthen long-term relationships.

She also raised a future-facing point that could become central to marketing strategy: sellers will need to become discoverable by the buyer’s AI agent. If a company’s technical capabilities, credibility and project portfolio are not structured in a machine-readable way, it may be excluded from consideration before a human buyer even enters the process.

The design concern: what happens to the joy of shopping?

Keyur Zaveri of Furlenco introduced one of the strongest counterpoints of the discussion: not every customer experience should be reduced to speed, efficiency and conversion.

Coming from the creative and design side, Zaveri said AI-generated content and automated ad systems are still in a nascent stage. He cited examples from furniture advertising where automated systems changed product visuals incorrectly. In some cases, beds appeared to float or were shown in unsuitable visual contexts.

For a furniture and lifestyle brand, such errors are not minor. The category depends heavily on context, aspiration and visual trust. If an AI system misrepresents the product environment, it can damage the customer experience and brand perception.

Zaveri’s deeper question was emotional. If AI agents do all the discovery, comparison and purchase decisions, where does the joy of shopping go? Furniture and home products are not purely functional purchases. Customers often imagine how they want to live, how they want their rooms to look and how a product will shape their personal space.

He argued that brands which preserve this sense of involvement and joy in an AI-led environment will stand apart. If every company eventually has access to similar AI capabilities, differentiation may come not from the technology itself, but from the way a brand designs the experience around it.

Financial services draws a firm line

Aishwarya Gupta of Jainam placed the debate firmly in the context of regulation, trust and investor protection.

She said AI is already helping in areas such as content creation, design, videos and internal models that help brands move faster. Jainam has also worked on understanding consumers in real time, moving beyond periodic research to sharper, more current insight.

But when it comes to financial decision-making, Gupta drew a clear line. In her view, AI agents cannot be allowed to independently make decisions for customers in a regulated environment. Investor interest has to be protected, and human judgment remains essential.

She said AI agents can play a role across onboarding, KYC, document uploads, customer nudges and support during the investment journey. They can help customers complete processes, understand friction points and make informed decisions. But they cannot simply recommend a particular stock or financial instrument unless the regulatory framework allows it.

Gupta said AI can help mitigate risk, encourage disciplined investing and guide customers up to a point, but it will not replace advisors or research professionals in the near future. Later in the discussion, she reiterated that AI agents will not take over marketing. They will act as enablers, making decision-making faster and giving marketers better visibility into consumer behaviour, data and campaign performance.

In gifting, intent is emotional and real time

For Avi Kumar of FNP, the next stage of AI will be shaped by both brand-side agents and consumer-side agents.

He said the last decade was about automation and AI, while the last two years made AI more native to marketing across analytics, design and customer lifecycle management. The agentic world, however, is still a work in progress. Brands are testing use cases, costs, implementation models and areas where decision-making can be handled autonomously.

Kumar said the future could see a consumer’s AI agent interacting with a brand’s AI agent. A buyer may simply ask their agent to find the best product, while the brand’s agent recommends options based on its own data and systems.

In a category like gifting, this becomes especially relevant because customer intent is often emotional and event-driven. Customers may know they want to send a gift but may not know what gift to choose. Kumar said AI agents can help customers make that decision faster by reading real-time browsing behaviour.

If a customer is browsing birthday products, for instance, the next page can dynamically change to show more birthday-led options instead of anniversary-led recommendations. But he also warned that poorly trained agents can create frustrating customer service loops. The agent has to be trained well and must have enough data, otherwise the experience can become a problem rather than a solution.

Responding to an audience question on when organizations are ready to move from copilots to agents, Kumar said readiness depends on category and the success of copilot use cases. He described three phases: fully human, copilot-assisted and agentic. Brands should move toward agentic systems only when copilot use cases have proven successful and can be scaled with clear success criteria.

AI as an advisor amplifier

Mithil Sejpal of SLIQ said AI agents are turning marketing into decision infrastructure. Earlier, marketing automation focused on sending the right message to the right customer. Now, AI agents are beginning to influence intent generation, product fitment and lead conversion.

In financial services, however, Sejpal said agents are not replacing advisors. Their more immediate role is to assist advisors and make them more productive. AI can process multiple data points and products, curate relevant offerings and help the advisor communicate better with the customer. The final layer of judgment, he said, still rests with the human.

He described this as an amplifier for advisors. A person who may have earlier completed five sales could potentially complete ten with AI support. But he strongly cautioned against assuming that financial advisors will disappear.

The cost of bad judgment in financial services is too high. Buying the wrong consumer product may be manageable, but making a wrong long-term financial decision can affect a customer for years. In such categories, AI can improve productivity and intent quality, but human oversight remains essential.

Sejpal also said that marketers will need to move from campaign ownership to outcome ownership. In an AI-agent-led world, marketing may no longer stop at communication. It will increasingly extend into customer journeys, next-best-actions, sales enablement and business outcomes.

Proprietary data becomes the moat

The audience questions pushed the discussion toward what will differentiate brands when AI capabilities become widely available. On whether proprietary data, proprietary models or proprietary workflows matter most, Zaveri said the operating model will become crucial when everyone has access to similar AI capabilities.

Kumar added that proprietary data will remain extremely valuable. An AI agent can create workflows, but without enough proprietary data to understand customer behaviour, it will not be as effective. For FNP, occasion and gifting behaviour data can make agents far more relevant. Similar logic applies in categories such as healthcare, where proprietary diagnostic and patient data could shape stronger use cases.

The session ended with a balanced view of the agentic future. AI agents represent a powerful vision for marketing, but the industry is still in the early innings. The opportunities are significant, but so are the challenges around data readiness, governance, category maturity, cost, adoption and trust.

The biggest takeaway from the discussion was not that marketers are about to be replaced. It was that marketing itself is being redefined.

The future may be agentic, but it will not become fully autonomous overnight. The brands that win will not be those that rush to remove humans from the loop. They will be the ones that understand where AI can accelerate intelligence, where humans must retain judgment, and how both can work together to create stronger customer outcomes.