Fundle.ai: Agentic AI for Retail Loyalty: India's Next Competitive Edge

Current State of Retail Customer Experience in India

Walk into any Select CITYWALK store corridor on a Saturday afternoon and you will see the paradox of Indian retail loyalty in full display. Footfall counters tick upward, POS terminals at Lifestyle, Manyavar, and FabIndia process hundreds of transactions per hour, and yet the average retailer has clean purchase data on fewer than 30% of those customers. The remaining 70% are ghosts — transacting in cash or through guest checkouts, invisible to the CRM system, unreachable for retention. This is the data foundation problem that sits upstream of every loyalty conversation.

Where data does exist, it lives in silos. The mall operator's Wi-Fi analytics platform knows dwell time and zone visits. The anchor brand's POS system — often running Petpooja, POSist, GoFrugal, or Wondersoft — knows the SKU-level transaction. The loyalty vendor, typically a legacy player like Capillary or EasyRewardz, holds points balances. The marketing automation tool — MoEngage, WebEngage, or Xeno — holds campaign response data. None of these systems talk to each other in real time. A customer who just spent ₹12,000 at a Reliance Trends outlet and is standing 40 metres away from a Cafe Coffee Day gets no contextually relevant nudge. The mall's marketing team simply does not have the infrastructure to generate that nudge in under 90 seconds, which is the window that matters.

The consequence is brutal at the commercial level. India's average loyalty programme churns roughly 55% of its enrolled base within the first 18 months. Redemption rates hover between 22% and 35% for mid-market programmes. Customer Acquisition Cost in organised retail has risen to between ₹380 and ₹620 per new loyalty member depending on category, while the average lifetime value of an unengaged member barely recovers that cost. Mall operators running 15–25 brands across a single property are spending ₹8–14 crore annually on loyalty infrastructure that generates provably low incremental revenue.

The answer is not more points. It is not a better-designed app. It is not another push notification. The answer is an intelligence layer that can see the whole customer, reason across their behaviour in real time, and act on their behalf in ways that feel genuinely useful — which is precisely where Agentic AI for retail loyalty enters the frame.

How Agentic AI Redefines Customer Loyalty in Indian Retail

The term 'agentic' carries a specific technical meaning that gets diluted in vendor marketing. An AI agent, in the rigorous sense, has four properties: it perceives its environment through sensors or data feeds, it maintains an internal state or memory across interactions, it selects actions from a goal-directed policy rather than a fixed rule set, and it learns from outcomes to update its behaviour. When you apply that architecture to retail loyalty, you get something qualitatively different from any campaign tool currently deployed at scale in India.

Consider a concrete scenario. A member of a Phoenix Marketcity loyalty programme — let us call her Priya — visits the mall on a Tuesday evening. The Fundle Agentic AI platform picks up her geofence entry, cross-references her purchase history (last visit was 47 days ago, lapsed from her usual 18-day cadence), notes that her last three purchases were all in the kids' apparel segment, sees that her profile indicates a birthday for her child in 11 days, and observes that the toy anchor store is running a 20% discount campaign that ends tomorrow. The Fundle AI Agent does not wait for a human CRM executive to notice this confluence of signals. It reasons: intervention opportunity is high, discount urgency is real, category relevance is strong, channel preference from past response data is WhatsApp. It sends a WhatsApp message in Hindi with a personalised nudge within 90 seconds of geofence entry. Priya redeems. The agent logs the outcome and recalibrates its confidence weights. No human touched this workflow.

This is not hypothetical. Fundle AI Workflow and Fundle AI Agents are built precisely to execute this kind of multi-step, multi-signal, goal-directed action at scale. The contrast with a rules engine — where that same scenario would require a CRM manager to have pre-programmed a segment called 'lapsed members with kids aged 4–10 near a toy store with a 48-hour campaign window' — should be obvious. Rules engines can handle a few dozen segments. Agentic AI handles millions of micro-contexts simultaneously.

For mall operators, the commercial upside is substantial. Programmes that have moved from rules-based to AI-agent-driven engagement report 25–40% uplift in redemption rates, 18–28% improvement in repeat visit frequency, and 12–22% increase in average basket size among AI-engaged members. These are not marketing projections — they are outcomes from AI-first loyalty deployments tracked over 12-month cohorts in Indian retail contexts.

Technology Trends Driving AI Loyalty Agents in India

Three technology shifts have converged in 2023–2024 to make agentic AI deployable at Indian retail scale — rather than just theoretically attractive. Understanding these shifts is essential for any CRM head evaluating whether now is the right time to move, or whether waiting another 12 months is a defensible call (it is not).

First, large language model inference costs have fallen roughly 90% in 18 months. Running a reasoning step — the kind of contextual interpretation that allows an AI agent to read Priya's behavioural signals and produce a relevant action — now costs fractions of a paisa per customer interaction. At 1 crore monthly active loyalty members, the compute cost of full agentic reasoning per interaction is commercially viable in a way it simply was not in 2021. This cost inflection is what separates the current generation of AI loyalty tools from the 'AI-powered' badge that vendors slapped on top of their 2018 rules engines.

Second, India's digital infrastructure has matured to support real-time data pipelines at mall scale. UPI transaction data is available near-instantly. WhatsApp Business API penetration among Indian retailers has crossed 78%. POS integration via systems like POSist and GoFrugal now supports webhooks and event-streaming rather than nightly batch exports. This means an agentic system can actually receive the signals it needs to reason within the contextual window that matters — the 90-second window of opportunity during a customer's in-mall visit.

Third, customer data platforms (CDPs) have commoditised identity resolution. Stitching together a customer's in-store POS data, app behaviour, WhatsApp interactions, and mall Wi-Fi footprints into a single profile — the unified customer graph — is no longer a 12-month integration project. Platforms built for Indian retail contexts can now deliver a resolved identity layer within weeks of deployment. Without this unified graph, agentic AI is reasoning on partial information; with it, the quality of agent decisions improves non-linearly as more signal types are connected.

For mall operators and multi-brand retail chains, the combined effect of these three shifts means the technical barriers to deploying Agentic AI for retail loyalty have dropped below the level where a 'wait and see' posture is commercially rational. Competitors who move in the next two quarters will accumulate behavioural data assets and model training advantages that are genuinely hard to replicate 18 months later. First-party data moats compound, and in Indian retail — where third-party cookies never really worked and DTC is still maturing — the brand or mall that owns the deepest customer graph wins.

Consumer Expectations and Behavioural Shifts Reshaping Loyalty

The Indian consumer of 2024 is not the same person who signed up for a points card at a Pantaloons counter in 2015. Three generational and behavioural shifts are rewriting what 'loyalty' means at the consumer level, and any technology investment that does not account for these shifts will underperform regardless of its AI pedigree.

Shift one: personalisation has moved from delight to baseline expectation. KPMG's India Customer Experience Excellence report consistently shows that personalisation is now the top driver of loyalty programme satisfaction for consumers under 35, ahead of discount depth and point value. A 25-year-old in Bengaluru who uses Swiggy, Nykaa, and Zepto daily receives algorithmic personalisation in every scroll. When she walks into a mall and receives a generic '500 bonus points this weekend' broadcast, the cognitive dissonance is jarring. She does not just ignore it — she actively lowers her trust in the brand. The bar for what counts as relevant communication has been permanently reset by consumer internet apps.

Shift two: channel fatigue is real and accelerating. India now has the world's second-highest SMS spam rate. Open rates for retail marketing SMS have fallen from roughly 35% in 2019 to below 18% in 2024 in most urban markets. Email is worse. App push notifications are being turned off by 60%+ of users within 30 days of download. WhatsApp remains the highest-engagement channel — 85–92% open rates for transactional messages in retail — but only when the content is genuinely relevant. An agentic system that chooses the right channel and times the message to a contextually appropriate moment does not just get better click-through; it avoids the burnout that degrades the channel for everyone.

Shift three: customers increasingly expect loyalty programmes to know them, not just track them. There is a meaningful difference between a programme that stores your purchase history and one that demonstrates awareness of your preferences in real time. Apollo Pharmacy's loyalty programme, for instance, has experimented with health-profile-aware communications — reminding members about recurring prescription refills before they run out, rather than pushing generic discount offers. The programmes that treat data as a tool for genuine customer service, rather than as a targeting list for promotional spam, are seeing NPS scores 15–20 points higher than category averages. Agentic AI makes this kind of service-first loyalty executable at scale.

KPIs That Actually Measure AI Loyalty Performance

Most retail loyalty programmes in India are measured on enrolled member count and points issued — the two metrics most divorced from commercial reality. A programme can show 2 million enrolled members and ₹40 crore in outstanding points liability while delivering near-zero incremental revenue. If your CFO is asking 'what is the loyalty programme worth?', and the answer involves member count rather than incremental revenue, you have a measurement problem before you have a technology problem.

The right KPI framework for an AI-driven loyalty programme has three tiers. The first tier covers engagement quality: monthly active member rate (target 35–45% for a healthy programme), first-redemption rate within 90 days for new enrollees (target 55%+), and channel response rate by segment. These tell you whether the AI's communications are landing. The second tier covers commercial impact: incremental revenue per active member per quarter versus a matched non-member cohort, basket uplift percentage for AI-touched transactions versus baseline, and lapse-rate recovery percentage for members who were flagged as at-risk by the churn prediction agent. These tell you whether engagement is converting to revenue.

The third tier covers programme health and efficiency: cost per incremental transaction (total loyalty programme OpEx divided by transactions genuinely attributable to loyalty intervention), redemption rate as a percentage of issued points, and Net Promoter Score delta between loyalty members and non-members. For context, well-run Indian retail loyalty programmes operating on AI-driven engagement show cost per incremental transaction in the ₹18–35 range; rule-based programmes running heavy SMS and discount spend often land at ₹80–120 per incremental transaction — a 3–4x efficiency gap that compounds at scale.

It is also worth tracking the data asset quality metric — the percentage of your transaction volume with a resolved customer identity — as a leading indicator of AI performance. As your identified transaction rate rises, your agent's decision quality rises with it. This metric closes the loop between the data infrastructure investments you made in step one of the playbook and the commercial outcomes you are reporting to the board.

How Fundle solves this

Fundle was built from first principles for the Indian retail context — not adapted from a Western loyalty stack designed for supermarket frequent-flyer economics. The Fundle AI Platform integrates across the full loyalty value chain: data ingestion, identity resolution, AI agent deployment, omnichannel execution, and closed-loop measurement. Every component is designed to work together, which is why the platform can deliver the 90-second contextual intervention described earlier rather than the 72-hour campaign cycle that legacy tools require.

At the core of the platform are Fundle AI Agents — purpose-built intelligent agents trained on Indian retail behavioural patterns. Fundle Mall Loyalty deploys agents specifically calibrated for multi-anchor mall environments, where the key commercial objective is driving cross-brand visitation and increasing dwell time. A member who visits the food court but never enters the fashion zone becomes a specific agent goal: understand why, identify the relevant offer or experience hook, and create a path to cross-zone trial. Fundle Brand Loyalty, by contrast, is configured for single-brand or mono-category retail chains — a jewellery brand like Tanishq, a lens retailer like Lenskart, or a pharmacy chain like Apollo Pharmacy — where the agent's goals centre on purchase frequency, category cross-sell, and lapse prevention.

Fundle Agentic AI and Fundle AI Workflow are the orchestration layers that allow multiple agents to collaborate on complex, multi-step customer journeys. A customer approaching a high-value anniversary purchase, for instance, might trigger a sequence involving a product recommendation agent, a personalised offer agent, a service upgrade agent (complimentary gift wrapping, priority billing), and a post-purchase experience agent — all coordinated through the Fundle AI Workflow without requiring a human campaign manager to stitch together four separate tools. The workflow self-monitors, self-corrects if a step underperforms, and surfaces the outcome data back to the CRM team in a readable dashboard.

Vineet Narang's founding vision for Fundle was straightforward: India's retail operators deserve AI infrastructure that is as sophisticated as what the consumer internet companies use internally, delivered at a price point and integration depth that works for mall developers and mid-market retail chains — not just billion-dollar e-commerce platforms. That vision is operationalised in every product decision: native WhatsApp and UPI integration built for India's actual digital stack, pre-built POS connectors for GoFrugal, Petpooja, POSist, and Wondersoft, and AI models fine-tuned on Indian purchase behaviour rather than imported from Western retail datasets. For the CRM head evaluating this space today, Fundle AI Platform is the platform built specifically for the problem you are actually trying to solve.