Quick commerce (q-comm) is no longer just a fulfilment channel; it is becoming a significant advertising business. Advertising revenue across major q-comm platforms is expected to reach INR 4,900 crore in 2026, up from INR 3,000 crore in 2025.

But as brand spends grow, so does the complexity of optimisation.

A campaign can deliver impressions, clicks and conversions and still fall short if the product is unavailable where demand is being created. For brands operating across q-comm players such as Blinkit, Zepto and Swiggy Instamart, understanding performance increasingly means looking beyond media metrics to inventory, fulfilment and market-level data.

That is the gap Hector AI, Wondrlab’s proprietary commerce technology platform, is looking to address. Its expanding q-comm offering connects advertising, inventory, supply chain and market intelligence, with its AI interface designed to help brands move from reporting performance to making decisions and taking action.

For Meher Patel, founder, Hector AI, the move into q-comm is less about changing the fundamentals of performance marketing and more about dealing with a far more complex operating environment.

On the sidelines of Hector’s press briefing on its enhanced q-comm solutions, Manifest spoke with Patel about why conventional dashboards may no longer be enough and how AI could connect advertising, inventory and fulfilment.

Fundamentals haven’t changed. The complexity has

Patel argues that the basic mechanics of advertising optimisation remain consistent across platforms. Spend still generates impressions, which generate clicks, conversions, orders and ultimately revenue.

“The entire customer acquisition journey fundamentally remains the same,” he told us. “There is the same CTR, conversion rate, average order values, cost per acquisition, ROAS, cart pulls. Everything is the same.”

That means Hector’s move into q-comm was, in part, a natural extension of the advertising optimisation capabilities it had already developed across commerce platforms.

“The complexity of q-comm is way beyond the regular marketplace of Amazon that we managed before,” Patel shared.

That complexity also explains why the company decided it needed to build technology around the problem rather than simply add another layer of reporting.

‘Didn’t 'want' a platform. We 'needed' a platform’

Patel is clear that Hector was not born out of a desire to become a platform company.

“We are not a platform-first company. We are an agency-first company that has built a platform,” he expressed. “Building a platform was not a want. We didn’t 'want' a platform. We 'needed' a platform.”

The need, he says, became apparent when the agency began managing quick-commerce campaigns for clients and encountered the sheer number of variables involved.

“The data is spread across too many variables, so a dashboard is not going to answer it,” Patel shared.

The response was to move away from the conventional dashboard model and towards an AI interface that could allow the team to interact with the underlying data more naturally.

“How do I make it easy for my team to manage that campaign optimisation?” he questioned. “And hence, we took that hard call that there would be no dashboard. Let’s use the power of AI and let’s build an MCP where we can talk naturally.”

From reporting what happened to planning what comes next

The bigger opportunity for Hector may lie beyond campaign optimisation.

Q-comm generates highly granular data at a city, product and inventory level. Patel believes that historical data, when structured and connected to an AI system, can also be used for forecasting and planning.

“If a brand is selling popcorn, it will know that during IPL, what kind of sales am I seeing in Mumbai against Nagpur, Delhi against Ghaziabad or Gurugram?” he said.

That could allow brands to approach an upcoming event with a different set of questions. Instead of simply asking how a previous campaign performed, a marketer could ask where to allocate the next budget to achieve a particular return and how much stock should be positioned across warehouses.

“Because we have that past data, and because the data is latched onto an AI, one can use the power of AI and tell AI that if I want to plan my budget for the next event and want it at this ROAS, which city should I be investing what amount of money in to get this ROAS?” Patel added.

The same logic, he argues, can extend into supply chain planning.

“Because one has past data and because they have AI, they can now use the power of AI to do meticulous planning, which, without AI, somebody from their team had to do,” he shared.

That is where Patel sees a material difference: the data is no longer being used solely to explain past performance but potentially to inform decisions across advertising and inventory.

Why not just upload the data into an LLM?

The obvious question is whether brands actually need a specialised platform for this. If marketers can already upload spreadsheets into general-purpose AI tools such as ChatGPT or Claude, what does Hector add?

For Patel, the answer lies in the structure and context of the data, rather than simply the AI model itself.

“There is a limit to what data you can upload into any LLM,” he expressed, pointing to the difficulty of working across multiple large datasets.

More importantly, separate datasets do not necessarily understand their relationship with one another.

“If one uploads Blinkit sales data, Zepto sales data and Amazon sales data, it has to hop and skip between all three Excel sheets,” Patel said. “It doesn’t know which Amazon Standard Identification Number belongs to which product SKU (Stock Keeping Unit) on Blinkit, Zepto, Instamart.”

Hector proposes to build that connective layer before the user asks the question.

“At my end, I have done the context layer. I have done the intelligence layer,” he said. “I have structured my data in such a way that you can interpret it, and the LLM can interpret it very quickly and very easily.”

That distinction is central to the platform’s proposition: the value is not simply in putting an LLM in front of commerce data, but in making fragmented datasets intelligible to the system before decisions are made.

The harder shift is still adoption

Technology may not be the biggest hurdle. Patel believes the larger challenge is getting marketers to change how they interact with data.

“The only roadblock that we see is mindset shift,” he shared, describing the move “from the dashboard stuff to prompting stuff.”

For brands accustomed to dashboards, reports and manually interpreting multiple data sources, asking an AI system questions and acting on its answers requires a different level of trust.

Patel believes that confidence will come through use. Once brands connect their actual data and see the system respond to complex questions using that data, he said, “it just makes them feel like magic.”

For him, the industry is currently somewhere between those two models, with marketers gradually moving from dashboards towards LLMs, prompts, and eventually taking actions through those interfaces.

AI can act, but it still needs a checkpoint

Giving AI access to campaign controls introduces another question: what happens when it gets something wrong?

Hector’s answer is a system of permissions and approvals. Some users can access and make changes, while others have read-only access. Even when users can interact with the data through Claude, Patel says, they cannot necessarily write back into the system.

There is also a review stage before an action is executed.

“When you say something, it will first give you a preview,” Patel said. The preview shows what is being changed, including the budget, keywords and campaigns involved.

Only after the user approves the proposed action does Hector execute the workflow through a job ID. And if something subsequently needs to be undone, Patel stated that the same job ID can be used to reverse the transactions.

The distinction is important because Hector’s proposition ultimately goes beyond making dashboards easier to read. It is about giving AI a role in interpreting connected commerce data and, with human approval, acting on it.

That leaves the bigger question for the industry: as quick commerce becomes more data-heavy and fragmented, will marketers continue to navigate individual dashboards and platforms, or will they increasingly ask an AI system to connect the dots for them?

For Patel, the direction is already taking shape. “People will start shifting from dashboards to LLMs to prompts and taking actions through that," he shared.

The real test, then, is not whether AI can produce another answer from another dataset. It is whether connecting those datasets can help brands make a better decision before the next ad is bought, the next SKU is moved, or the next customer reaches for a product that may or may not be on the shelf.