The column was originally published in .
CXM chief business officer Devang Shah argues that in Indian boardrooms, AI ambitions are often reshaped by the pressure for quick returns, leading organisations to favour short-term, measurable wins over long-term, compounding intelligence, resulting in adoption that looks productive on paper but rarely builds lasting competitive advantage.
The real problem with AI adoption in Indian marketing isn’t resistance - it is the quiet distortion that happens when organisations demand returns before the foundations are even in place.
There is a familiar ritual in Indian boardrooms. Someone proposes an AI investment. Heads nod. A slide deck gets approved. And then, almost inevitably, comes the question that reshapes everything: What is the ROI, and when exactly will we see it?
The project rarely dies. It simply gets reshaped, rescoped, and redirected - trimmed down to whatever can be measured before the next quarterly review, like a long journey cut short because the first milestone is mistaken for the destination.
This is India’s real AI problem. Not rejection, but distortion.
The pressure to show quick returns does not kill ambition. It bends it - gently but decisively - toward what is visible, defensible, and immediate. Chatbots that deflect routine queries. Dashboards that present yesterday’s numbers more elegantly. Automation that quietly trims a headcount line. All useful. All sensible. And yet, none of them the kind of investment that grows stronger with time. None of them the kind of advantage a competitor cannot replicate over a weekend.
Asking for return on investment is not the problem. It is, in fact, the instinct that has helped Indian businesses navigate decades of real uncertainty. The difficulty lies elsewhere - in the horizon we attach to that question, and in the quiet confusion between a milestone and an outcome.
A system that reduces call centre volume by fifteen percent in ninety days is a milestone. An AI that learns from every customer interaction, refines its understanding, and improves decisions over time is an outcome. The two belong to entirely different clocks. Treating them as the same is a little like pulling a plant out of the ground every few weeks to check whether the roots have taken hold. The act of checking becomes the reason nothing grows.
This tension becomes clearer when you look at retail lending. Many large lenders have invested in AI to assess credit, flag risk, and segment customers. The initiatives that moved fastest were those that could show results within a quarter or two - lower processing times, marginal improvements in default rates. The slower, less visible work - building systems that learn continuously from repayment behaviour and get sharper with every data point - received far less support.
And yet, over time, it is this slower work that compounds. The lenders who stayed with it are now making better credit decisions than their peers - faster, cheaper, and more precise. The difference is not access to better technology. It is simply the willingness to wait long enough for the system to learn.
A similar pattern is visible in consumer goods. AI has been widely adopted for demand forecasting, and the gains are real. But the larger opportunity lies elsewhere - in understanding why consumers switch brands, what builds loyalty, and how emerging needs can be detected before they appear in sales data. That kind of capability does not arrive fully formed. It requires time, proprietary data, and the patience to treat early signals as direction rather than verdict. Most organisations, understandably, stop short of that journey.
Experienced marketers recognise this tension instinctively. No one expects a brand campaign to pay for itself in a quarter. Brand equity is built through repetition, consistency, and the willingness to invest before the outcome is obvious. It often looks unjustifiable in the short term and inevitable in hindsight.
The same is true of AI - with one important difference. The cost of falling behind is not just a media budget that can be increased next year. It is time that cannot be recovered. It is years of learning that cannot be bought back once lost.
The advantages that endure are, almost by definition, the ones that cannot be rushed. Training AI on proprietary customer data. Embedding years of institutional knowledge into how systems make decisions. Redesigning workflows so human judgement and machine intelligence reinforce each other rather than operate in parallel. None of this produces a tidy number by the next board meeting.
A competitor can license the same AI model in a matter of days. What they cannot replicate is eighteen months of your data, your customers, and the quiet accumulation of organisational learning built into how your business operates. That is the moat - and it is often left unbuilt, not because organisations do not see it, but because they cannot justify waiting for it.
Short-term milestones still have a role to play, but perhaps not the one they are often given. They are not proof that an AI investment is working. They are signals that it is moving in the right direction. Is adoption spreading across the organisation? Are workflows genuinely changing? Is the underlying data becoming richer and more reliable? These are useful questions to ask early on. What they cannot be is the final verdict.
For a business culture that has long celebrated jugaad - the ability to find a clever workaround for almost any problem - this creates a subtle tension. Some advantages cannot be improvised. They have to be cultivated, patiently, often without immediate evidence that they are taking root.
India has what it takes to build meaningful AI capability. The engineering talent is here. The scale of market data is here. The willingness to invest is clearly present. What is less visible, perhaps, is the comfort with treating AI the way the best marketers treat a long-term platform - something you commit to before the returns are obvious, measure through signals rather than outcomes, and allow to compound over time.
The organisations that develop that discipline will not simply use AI. They will be shaped by it.
The ones that keep asking for the bill before the meal is served may find themselves well-fed in the short term - but still wondering, years later, why the deeper advantage never arrived.
Because the truth, as it has always been, is simple.
The food takes time to cook.
