If agencies are now technology vendors, the real discourse revolves around how should advertiser data be handled?
New Delhi: Advertising agencies have been known to sell ideas, strategy, planning and execution. Technology sat in the background, supporting campaigns rather than defining agency relationships.
Today, agencies are bringing proprietary AI platforms, data tools, and automated workflows into client relationships. What was once a service business is steadily becoming a technology business. For brands, the real question is no longer whether an agency has an AI platform. Almost every large network claims one today.
If agencies are now technology vendors, the real question revolves around how should advertiser data be handled? Where does a brand's marketing intelligence end and an agency's intellectual property begin? And when an advertiser decides to change partners, what exactly should move with it?
Agencies are building on AI. Okay, but why?
AI cuts down the time and manpower needed for time-intensive tasks such as reporting, audience analysis, creative testing, planning and campaign optimisation. Over the past year, agencies have built AI into their existing workflows, identifying where it can take over repetitive processes while leaving strategic decision-making to people. Their day-to-day experience managing marketing campaigns gives them a clear understanding of which functions can be standardised, automated and scaled. As a result, agencies are increasingly developing AI-led processes that solve universal marketing challenges rather than one-off client requirements.
But if agencies sell AI only as an efficiency tool, it could weaken a business model that has always run on retainers, commissions, and people based fees. A proprietary platform gives agencies another way to earn. It lets them package their processes as technology, use it across clients and create revenue that does not depend only on headcount.
Prasenjit Basu, Head of Marketing at Voltas Beko, said the agency business itself is changing. Agencies are increasingly bringing proprietary tools and technology to clients as value-added services. .
He added that these tools often start free and then move to a paid model as the requirement grows. “The first layer of these tools is often available free of cost. As you move higher and require deeper datasets or more sophisticated analysis, you start paying for it. That's similar to how any research or data company operates,” he said.
According to Basu, agencies are essentially monetising their understanding of client businesses, and this becomes a revenue stream that sits outside regular client servicing.
The logic is fairly simple. Technology can be reused across accounts, while agency manpower has to be deployed account by account. A whitepaper by 3C Ventures raises a similar tension, asking whether advertisers end up paying for technology on top of headcount, or instead of it.
Vivek Bhargava, Co-Founder of consumr.ai, believes specialist technology firms could eventually outspend agencies in this race. “AI platforms like ours are valued at 30 times revenue, agencies are valued at 10 times profit. That gap tells you where capital markets believe compounding value sits, and it means agencies structurally cannot outspend technology companies on R&D,” he said. His point is not that agencies will stop using tech. He believes many will end up licensing it instead of owning every layer.
Haresh Nayak, Founder and CEO of Connect Network Inc., feels the platform's accumulated knowledge matters more than the technology itself. “Technology will increasingly become commoditized. What will differentiate platforms is not the AI engine itself, but the quality of intelligence they accumulate over time,” he said.
What exactly are brands paying for?
These tools usually cover performance monitoring, consumer insight, creative testing, A B testing, campaign tracking and eye tracking. Their biggest value is speed, since they cut down the time needed to answer a marketing question. “Sometimes you need a quick dipstick study or immediate insights on a product. These tools help agencies respond much faster,” Basu said.
So a brand ends up paying for a mix of things at once. This includes access to the technology, automation of repetitive work, agency people who interpret the output, and a decision making process that gets smarter with time.
This is why the buying decision is shifting. It is no longer about whether a platform exists, but about the actual value it adds. Does it just show existing data faster? Does it improve recommendations? Does it hold on to past campaign learning? And is the platform fee paying for something new, or repackaging work that is already covered under the agency's fee.
Bhargava expects the market to eventually settle into three clear layers. Technology owned by a specialist platform, brand intelligence held by the advertiser, and strategy provided by the agency.
In this setup, the agency stays the expert operator but no longer controls the infrastructure the client's learning actually sits on.
Who owns what the platform learns?
A marketing platform learns by constantly processing campaign signals. It picks up which audiences respond, which creative works, where the budget performs best, and how past optimisation decisions shaped the outcome.
Over time this builds into something bigger than a campaign report. It creates audience segments, scoring systems, prompts, optimisation histories, and recommendations built specifically for one advertiser.
The legal question starts with separating the raw input from what gets built out of it. Kaushik Moitra, Partner and Practice Lead for Regulatory, IP and TMT at Bharucha and Partners, explained the basic split. “The agency's platform, code and cross-client models are its intellectual property, the brand's own data is the brand's,” he said.
The tricky part is what Moitra called the single brand derived layer. This includes models fine tuned on one brand's data, along with its optimisation history and prompts. Indian law does not automatically hand this entire layer to either side.
Copyright can apply to an original compilation or database, while trained models often depend on the contract and on the common law protection around breach of confidence. Personal data processing separately falls under the Digital Personal Data Protection Act, 2023.
Moitra said ownership of the input does not automatically extend to what is built from it. “Ownership of the input data, which lies with the brand, confers ownership of a model or output built from it only where the contract or the general law so provides, it does not follow automatically,” he said.
Nayak looked at the same question from an industry lens. “The platform itself, its AI architecture, algorithms, optimisation models and reasoning systems, is the agency's intellectual property. The outputs generated through a brand's investment, campaign insights, audience learnings, optimisation history, planning intelligence and performance data belong to the brand,” he said.
He added that while platforms do learn patterns across multiple campaigns, this learning should stay anonymised and aggregated. “A brand's competitive intelligence should never become another client's advantage,” he said.
Basu also pointed out that not every agency tool depends heavily on client data. Many are built around broader consumer insight or planning support. “Sometimes we provide our own business intelligence to the agency, but our brand intelligence remains our intellectual property. We share customer-level data only when it is genuinely required,” he said.
What happens when a brand switches agencies?
This is where the difference between owning data and owning the intelligence built from it becomes real.
Moitra explained that contracts decide everything here. “What moves is what the contract provides for. First-party data and campaign reports are ordinarily the brand's and return to it. Audience models, optimisation history, AI learnings and workflows are a separate, derivative asset, and move only if the agreement says so, there is no default rule of law that transfers them,” he said.
A standard clause on work product or deliverables may not cover this. Unless the contract clearly mentions model weights, prompts, audience segments and optimisation histories, the ownership position stays unclear.
Moitra also separated what a brand is entitled to from what it can actually take away. “Where the contract is silent on derived assets, the brand cannot compel their transfer, and even where it is entitled to an asset, there may be no interoperable format in which to hand it over,” he said.
Even a clause asking the agency to return all data may not help much. The contract needs to spell out exactly what gets transferred, in what format, within what time, and whether it can even be used elsewhere.
Bhargava believes independent platforms solve part of this problem, since the technology relationship stays intact even if the agency changes. “When the AI platform is a licensed third party rather than agency-proprietary, the brand's audience models, optimisation history and learnings sit on the platform, and the brand can retain that relationship directly even as agencies change,” he said.
Moitra's advice is to deal with raw data, derived models, prompts and optimisation histories separately right at the start of the engagement. Exit clauses should clearly state what gets returned or deleted, in what format, and within what time. Contracts should also cover audit rights, confidentiality, data protection, and how the agency is allowed to use aggregated or anonymised client data.
The agency of the future will probably still be hired for its people and its ideas. But it will be judged the way any technology vendor is judged. By what its system learns, who gets to use that learning, and whether the advertiser can walk away with it.