I’ve largely stopped answering the question of whether AI will replace agencies, not because it doesn’t matter, but because it’s the wrong way to frame what’s actually happening. An agency was never really one thing that either survives this or doesn’t. It’s a supply chain, a sequence of decisions running from brief to insight to idea to execution to measurement, and AI is moving through that chain unevenly. In some parts of it, the work stays but what a person spends their time on changes completely. In the parts that made agencies worth hiring in the first place, the bar for what counts as good is simply getting higher. The question worth asking isn’t whether the agency survives. It’s which parts of the work are changing shape, and whether you’ve worked that out before your P&L works it out for you.
Here’s the part that’s already shifted, quietly, over the last couple of years. A large share of what used to fill an agency’s day was throughput dressed up as craft, things like format variants, adaptation sizes, first-pass copy, rough visual options put together just to react to. That kind of work existed because producing even one version of an idea used to be expensive and slow, so the whole industry built itself around moving that work through a sequence, strategy to creative to media to production, each stage handing off carefully to the next. AI has made producing a version of an idea nearly free. What’s really under pressure isn’t the agency itself. It’s the assembly line built inside it, the handoffs, the waiting, the same brief being re-explained to every new team along the way.
I’ve said for a while that the agency that wins is the agency that packs integration, and this is exactly the reason why. That department-by-department structure made a lot of sense when every stage needed a different, scarce, expensive kind of expertise. AI takes away much of the reason for that sequence to exist at all. A strategist can now sit with an actual creative option in the room instead of just a written brief. A media plan can respond to early creative signal instead of waiting for a finished asset to land. The walls between disciplines were always a little artificial to begin with, held up mostly by how hard it used to be to move quickly across them. AI is the first thing I’ve come across that genuinely dissolves those walls, rather than just promising to someday.
The easier path with AI adoption is to bolt a faster tool onto each department exactly as it already exists, a quicker copy tool for the writers, a quicker visual tool for the designers, everyone a little faster at their one step, still handing off work the same way they always have. That isn’t really integration. It’s the same assembly line, just moving a bit quicker. Real integration means the same context, the same brand memory, the same client history, sitting in front of every discipline at once, so the handoff stops being the place where ideas quietly lose something. That’s what we’ve spent a little over a year building inside Schbang AI Labs. An internal system we call Second Brain holds the institutional memory an agency usually loses the moment someone leaves or a client relationship changes hands, things like brand language rules, past campaign insights, meeting notes, and audience behaviour, and puts all of it in front of strategy, creative and media teams at the same time, instead of letting it sit in one department’s inbox. Alongside it, a set of assistants we call The Edge takes on the first-pass mining and drafting that used to eat into a junior team’s week. Neither of these replaced anyone. Both exist simply to make the walls between departments a little thinner.
What all of this changes, underneath the surface, is an agency’s relationship with its own output. Once production stops being the bottleneck, the real constraint moves upstream to briefing and strategy, and downstream to measurement. An agency that used to sell “we’ll make this for you” now needs to be able to say “we know what’s worth making, and we can tell you what it actually did once it was out in the world.” I’ve made a version of this argument before, specifically about media measurement. India’s ad landscape has changed enormously, digital now carries well over half of it, and yet we’re still largely measuring against exposure metrics built for a completely different era. What the industry actually needs is a measurement standard that’s cross-platform, always-on, AI-driven, and built around attention rather than just impressions. That isn’t a footnote about media buying. It’s the same discipline this whole argument rests on, just applied at the very last mile
There’s a market-specific piece of this too, one I don’t think gets nearly enough attention. Most of the global conversation around AI and marketing assumes a fairly uniform consumer, one search behaviour, one language, one shared set of cultural references. That isn’t India. A brand’s customer in Bandra and its customer in Raipur are, for AI’s purposes, almost different markets altogether. Building for that isn’t really a prompt-engineering problem. It’s a strategy and localisation problem, and it rewards an agency that’s genuinely integrated over one that’s simply stitched together from separate specialists, because nobody in a stitched-together chain can afford to sit around waiting for the next handoff when the market itself is this fragmented.
There are, I think, two honest ways to spend this moment. You can bolt AI onto the structure you already have and end up modestly faster at being the same agency you were before. Or you can use it to actually take down the walls between departments, and become something genuinely different, an agency where an idea can move from insight all the way to output without losing anything of itself along the way. AI didn’t create the case for the second option. It simply took away the last good excuse for choosing the first. That’s the agency we’re trying to build. I think it’s the only one worth building right now.
