Retail has crossed a quiet but consequential threshold. Real-time decisioning is no longer a differentiator reserved for elite operators – it is table stakes for survival. The gap between what customers expect and what most retail systems can deliver has widened to the point where latency itself has become a liability.
That tension sits at the heart of a recent conversation with Balaji Balasubramanian, President and Chief Product Officer for SAP Customer Experience. While much of the industry still frames AI as a set of tools – pricing engines here, personalization widgets there – the discussion pointed to something more structural: a shift from destination-based retail to instant demand, where intent is fleeting, signals are fragmented, and value is created (or lost) in moments, not funnels.
The end of the destination model
For decades, retail was organized around destinations – stores, sites, campaigns – designed to pull shoppers into predefined journeys. That model assumed time, patience, and linearity. None of those assumptions hold today.
Demand increasingly originates outside owned channels: in search results, social feeds, answer engines, and conversational interfaces. By the time a shopper lands on a site or opens an app, intent may already be formed – or gone. The competitive advantage shifts from attracting attention to recognizing intent early and acting on it immediately.
This is where real-time intelligence stops being an optimization layer and becomes an operating requirement. Retailers must sense signals as they emerge, predict likely outcomes, and coordinate responses across merchandising, pricing, fulfillment, and service – often in the span of a single interaction.
Unified context beats isolated intelligence
A recurring theme in the conversation was that AI is only as effective as the context it operates within. Many retailers have invested heavily in point solutions – best-in-class tools that optimize narrow slices of the business. The result is local intelligence trapped in global blindness.
What’s missing is unified context: the ability for models to understand the customer, the product, the inventory position, the margin profile, and the operational constraints as parts of the same system. Without that, “personalization” is little more than educated guessing.
When context is unified – combining transactional, behavioral, and engagement data – new capabilities emerge. Merchandising becomes responsive rather than seasonal. Service becomes proactive rather than reactive. Journeys can be orchestrated end-to-end instead of stitched together after the fact. Most importantly, decisions that once required manual reconciliation across teams can be made algorithmically, with human oversight focused on exceptions and strategy rather than data wrangling.
AI as augmentation, not substitution
One of the more pragmatic insights from Balasubramanian’s perspective is the framing of AI as augmentation, not replacement. Retail organizations are not short on effort; they are short on time and coherence.
Teams burn hours reconciling dashboards, exporting reports, and translating insights between systems that were never designed to talk to one another. Embedded AI assistants change that equation – not by making decisions autonomously, but by collapsing friction in the flow of work. When questions can be asked in natural language and answered with live, contextual data, velocity improves without sacrificing judgment.
Equally important are deeper research capabilities that sit alongside assistants. These systems correlate signals across channels and timeframes to surface patterns humans are unlikely to spot: margin erosion tied to fulfillment routes, category cannibalization across channels, or loyalty decay linked to post-purchase friction. When assistants and deep analysis operate on the same semantic foundation, retailers can test and learn at speed – adjusting price, placement, bundles, or service levels with confidence rather than intuition.
Personalization becomes operational
Perhaps the most consequential shift is how this architecture reframes personalization itself. Historically, personalization lived in marketing – a recommendation engine, a targeted offer, a customized message. In an instant-demand world, that definition is dangerously incomplete.
True personalization must be operational. Before an offer is shown, the system needs to verify inventory at the right node, evaluate margin impact, and select a fulfillment option that aligns with both customer preference and sustainability goals. That orchestration extends from discovery through post-purchase, turning casual questions into guided journeys that respect intent, context, and constraints simultaneously.
Retailers who connect these dots move beyond “relevant content” to relevant outcomes: fewer stockouts, lower returns, faster cycles, and experiences that feel bespoke without adding headcount or complexity.
Enablers that reduce friction, not add it
The technology stack matters – but only insofar as it reduces friction rather than introducing new layers. A business data cloud that harmonizes information across SAP and non-SAP systems provides the substrate for consistent context. On top of that, conversational co-pilots like Joule translate complexity into action – answering questions, triggering workflows, and summarizing insights in language teams actually use.
WalkMe plays a complementary role by addressing a chronic blind spot in digital transformation: adoption. By guiding users through complex workflows and surfacing context-aware prompts at the moment of need, it ensures that intelligence doesn’t stall at the interface. Together, these layers shorten the distance between insight and execution – a gap that has historically swallowed much of AI’s promised value.
Why the timing matters now
Two forces amplify the urgency. First, the expectation gap continues to widen. Consumers want relevance, responsiveness, and values alignment in real time, not as brand promises. Second, traditional loyalty is eroding. Points and perks cannot compensate for friction, delays, or broken experiences.
AI changes the equation by decoding emotional signals in feedback, predicting needs, and aligning operations to meet them consistently. Trust becomes the currency of loyalty, and loyalty becomes a function of delivery rather than incentives.
Retailers that pilot now – especially during peak periods – gain compounding advantage. They learn which use cases scale: margin-aware personalization, proactive service triage, dynamic bundling, or sustainable last-mile optimization. As those experiments accumulate, real-time retail stops being a vision and becomes the default operating model.
The takeaway from this conversation is not that AI will save retail. It’s that retail that fails to operate in real time will increasingly fail to operate at all. The winners will be those who treat intelligence not as a feature, but as the connective tissue of the enterprise – linking intent to action, and action to trust, in the moments that matter most.
