AI-Led Media Buyers Design the System, Not Just Run the Tool
Marico's Ankit Desai explains how AI changes the media buyer's role: from running tools to designing systems, writing sharper briefs, and holding accountability.
At the afaqs! Media Quest Summit in Mumbai on August 21, 2026, Ankit Desai, head of media and digital marketing at Marico, made a clear distinction: using an AI-powered platform does not make someone an AI-led media buyer. What does is the ability to design the architecture around the technology, set its guardrails, and write briefs precise enough to get useful output. As machines take over reporting and campaign execution, Desai says the role shifts from operating tools to governing systems, and accountability cannot be automated away.
What happened
| Detail | Fact |
|---|---|
| Event | afaqs! Media Quest Summit, 2nd edition |
| Date | August 21, 2026 |
| Location | Fairfield by Marriott, Mumbai |
| Speaker | Ankit Desai, head of media and digital marketing, Marico |
| Interviewer | Sreekant Khandekar, co-founder and CEO, afaqs! |
| Old campaign setup time | 10 to 12 days after planning |
| Current campaign setup time (machines) | Approximately one day |
Desai opened by pushing back on a common assumption in the industry. Simply adopting an AI platform does not make a media function AI-led. His definition is more demanding:
“For me, an AI-led buyer is someone who designs the systems, who is able to create the architecture, to create the guardrails of how the system should run, rather than just run the tool itself.”
He applied the same standard to both marketers and their agency partners. Treating the tool as a substitute for professional judgement, he warned, means you also inherit whatever the tool gets wrong, without noticing.
Which parts of media buying get automated first?
Desai broke the media function into four layers and ranked them by how quickly automation can take over.
- Reporting: The most automatable layer. It involves collecting, consolidating and formatting large volumes of data, work that rarely produces insight on its own.
- Campaign execution: Already moving fast. Setup that once took 10 to 12 days post-planning can now be done in roughly a day.
- Planning: More complicated. Machines can produce a plan quickly, but they cannot verify whether the business question driving the plan is the right one.
- Human judgement: Negotiations, platform calibrations, and cross-platform arbitration remain areas where Desai does not see machines taking full control.
The risk he flags for planning is subtle. Speed is not the same as quality. A machine that generates a plan in minutes can also lock in a flawed hypothesis in minutes, with budget behind it before anyone asks whether the brief was right.
Why the brief matters more when planning gets faster
Desai’s argument here is worth sitting with. Traditional media planning had friction: discussions, revisions, delays. That friction was frustrating, but it also gave teams time to find holes in a weak brief before money moved. Remove the friction and a bad hypothesis travels much faster.
“If you take that friction out, you can go all over the place. So, if you ask me, I think the plan-making is going to get commoditised. What will matter is, what do you guide the machine to create?”
His practical answer is to invest heavily in the hypothesis stage. A generic objective like “build awareness” gives the machine very little to work with. A brief built around a specific consumer, context and measurable business challenge gives it something to optimise against. For teams already using performance advertising platforms with automated bidding and audience targeting, this should feel familiar: the machine scales what you point it at, for better or worse.
Who is accountable when the machine gets it wrong?
As automation takes over more execution, responsibility becomes harder to trace. Desai’s answer is direct: accountability should follow control.
“Finally, it’s the marketer who signs off on what the guardrails are. If they haven’t, and if they’ve chosen to go with a black box where a publisher or an agency partner has come in and said, ‘Hey, it’s magic. It’ll happen’, then they do that at their own peril.”
He splits responsibility cleanly. The marketer owns system design and guardrails. The agency and publisher own execution within those guardrails. Neither can hand accountability to the platform.
On cross-platform measurement, he argues agencies still have a necessary role as an arbitration layer between platforms and marketers. Publishers have commercial incentives to present data in ways that favour their own yield. Without an independent layer to challenge that framing, marketers will drift toward whichever platform narrative arrives loudest.
Our take
Desai’s framing maps closely to what we see working with clients running workflow automation and media campaigns. The teams that get the most from AI tools are not the ones who hand the brief to the platform. They are the ones who spend more time upstream: defining the hypothesis, setting the guardrails, and building a process to catch the machine when it optimises for the wrong thing.
The concern about junior talent is real and largely unresolved. If automation removes the repetitive work through which planners traditionally built pattern recognition, the industry needs a deliberate replacement: structured exposure to where the machine fails, not just where it succeeds. That is harder to systematise than a dashboard, but it is the actual skill gap opening up.
The closest parallel we see in our own work: when a client asks us to automate a reporting or campaign workflow, the first question is always what decision this output is meant to support. If the client cannot answer that, no amount of automation fixes the brief. Desai is saying the same thing at scale, for an entire industry. Read our broader AI news coverage for more on how automation is reshaping marketing roles.
What to do about it
- Before briefing any AI media tool, write the business hypothesis in one sentence: who you are reaching, why they should act, and what success looks like in numbers.
- Audit which parts of your media function are pure data handling (reporting, setup) and build automation there first, where the downside of errors is lowest.
- Keep a human arbitration step in cross-platform reporting. Do not accept a publisher’s attribution numbers without an independent check.
- Document your guardrails: budget caps, audience exclusions, brand-safety rules. The marketer who cannot articulate these owns the consequences when the machine ignores them.
- Give junior team members deliberate exposure to campaign failures and edge cases, not just dashboards showing what worked.
The plan-making will get cheaper and faster. The brief will get more expensive to get wrong.
Frequently asked questions
What does an AI-led media buyer actually do differently?
According to Ankit Desai of Marico, an AI-led media buyer designs the system architecture and sets guardrails for how the AI operates, rather than simply running a platform. The focus shifts from execution to governance and brief quality.
Which parts of media buying will AI automate first?
Reporting is the most immediate candidate, followed by campaign execution. Desai notes that campaign setup that once took 10 to 12 days can now be completed in roughly one day. Strategic planning and human judgement in negotiations are harder to automate.
Who is accountable when an AI media campaign goes wrong?
Desai argues accountability follows control. The marketer is responsible for system design and guardrails; the agency and publisher are responsible for execution within those guardrails. Neither can delegate accountability to the AI platform itself.
Why is the media brief more important with AI-generated plans?
AI can generate a media plan almost instantly, which means a weak hypothesis or vague brief gets scaled and funded much faster than before. The traditional friction of revisions and delays once helped surface bad briefs before money was committed.
