Fundle.ai: AI-Powered Customer Loyalty Agents Transforming Indian Retail Personalisation

The Importance of Personalisation in Loyalty Rewards

Walk into any major Indian mall on a Saturday afternoon and ask the marketing manager what their top loyalty pain point is. Nine times out of ten the answer is some variation of: 'We have the data but we cannot act on it fast enough.' That gap — between data availability and actioned personalisation — is precisely where loyalty programmes haemorrhage value.

Personalisation in loyalty is not about calling a customer by their first name in an SMS. That is table stakes. Real personalisation means understanding that a customer who buys ethnic wear at Manyavar every October is almost certainly shopping for Diwali, and therefore deserves a reward communication in late September — not a generic 'festive offer' blast on October 20th when she has already converted or moved on. It means recognising that a Lenskart buyer who has bought two frames in 18 months is overdue for an eye-test reminder, and that the right incentive is not a percentage discount but a complimentary service. It means knowing that an Apollo Pharmacy customer buying diabetes management products responds to health milestone rewards, not random points multipliers on unrelated categories.

The business case for this level of personalisation is well-documented. McKinsey's retail personalisation research consistently shows that personalisation at scale can deliver a 10–15% revenue uplift for omnichannel retailers. In the Indian context, where average basket sizes in fashion retail sit between ₹1,200 and ₹2,800 and visit frequency for mid-market shoppers is 3–5 times per year, even a single additional annual visit per active member is worth ₹1,200–₹2,800 in incremental revenue per member. Multiply that across 5 lakh active members and the programme-level impact runs to ₹60–₹140 crore in incremental topline — without acquiring a single new customer.

Mall operators face an additional complexity layer: they must personalise across a portfolio of 150–300 brands simultaneously, with each brand having its own margin structure, redemption economics, and customer acquisition priorities. A Phoenix Marketcity or an Inorbit Mall cannot treat a luxury jewellery anchor and a quick-service restaurant the same way in their loyalty engine. The personalisation layer must be smart enough to optimise for mall-level footfall and tenant-level revenue simultaneously. This is a multi-objective optimisation problem that no spreadsheet and no legacy CRM can solve — but AI loyalty agents can.

How AI Loyalty Agents Tailor Offers to Individual Customers

The architectural difference between a traditional loyalty platform and an agentic AI loyalty system is the difference between a vending machine and a seasoned retail advisor. A vending machine dispenses the same item when you press the same button. A retail advisor listens, observes, and adjusts their recommendation in real time based on what you said, what you browsed, what you bought last time, and what you seem to be in the mood for today. AI-powered customer loyalty agents are built on the second model.

At the technical level, AI loyalty agents operate through a continuous sense-reason-act loop. They ingest signals from multiple sources — POS transaction feeds from systems like Petpooja, POSist, GoFrugal, and Wondersoft; in-app browse events; geolocation triggers within a mall; customer service interactions; and even external signals like local weather or a nearby competitor promotion. The agent then reasons over this multi-signal context against a member's historical profile, RFM segment, and predicted next-best-action. Finally, it executes — dispatching a WhatsApp message, adjusting the homepage banner in the loyalty app, triggering a push notification, or queuing a personalised email — all without a human marketer configuring a new campaign rule.

The 'agentic' quality matters enormously here. Agentic AI for retail loyalty means the system has the autonomy to pursue a goal — say, re-engaging a lapsing member who has not visited in 47 days — by chaining together a sequence of actions: first a soft 'we miss you' push notification, then three days later a personalised bonus-points offer on the category she last bought in, then on day nine a time-limited reward that expires in 48 hours to create urgency. No human marketer at a Pantaloons or a Lifestyle store chain has the bandwidth to orchestrate that sequence for each of five lakh lapsing members individually. An AI loyalty agent does it simultaneously for all of them.

Competitors in this space — Capillary, Antavo, and to some extent MoEngage and WebEngage — offer campaign automation and segmentation. But campaign automation is still human-initiated; a marketer defines the segment, writes the message, and triggers the flow. Agentic AI is self-initiating: the agent identifies the opportunity, composes the intervention, and executes it — logging the outcome to continuously refine its own decision model. That is a qualitatively different capability, and it is the core of what Fundle AI Agents deliver.

Using Behavioral and Transactional Data Effectively

Data is not the differentiator in 2025 — most retailers of scale have it. The differentiator is the speed and granularity with which you act on it. Indian retailers sitting on POS data from GoFrugal or Wondersoft integrations, app engagement data from their loyalty apps, and WhatsApp opt-in databases are data-rich but often insight-poor, because the transformation layer between raw event data and personalised action is missing or too slow.

Effective use of behavioural and transactional data for loyalty personalisation starts with RFM segmentation — Recency, Frequency, Monetary — as the foundational layer. A customer who bought at a Pantaloons outlet 12 days ago, has shopped four times in the last six months, and has a basket value averaging ₹3,400 is a fundamentally different loyalty investment than a customer who last visited eight months ago, has two lifetime transactions, and a ₹800 basket average. These two members need different reward structures, different communication cadences, and different reactivation incentives. Any platform that treats them identically is burning marketing budget.

Beyond RFM, category affinity modelling adds the next layer of precision. A Cafe Coffee Day loyalty member who consistently orders cold beverages in the evening versus one who buys hot beverages at 8 AM is signalling different consumption contexts, and a context-aware AI loyalty agent designs different offers accordingly — perhaps a 'evening combo reward' for the former and a 'morning ritual' bonus for the latter. Apparel retailers like FabIndia can use purchase category signals — kurtas versus home furnishings versus personal care — to determine which communication should lead with which product universe in a particular member's journey.

Channel preference data is equally critical and chronically underused. Indian consumers in the 25–40 demographic are predominantly WhatsApp-first; open rates on WhatsApp Business messages run at 68–80% versus 18–24% for email and 12–15% for SMS. Yet most loyalty CRM platforms default to SMS because it is the cheapest channel to operate. An AI loyalty agent that analyses per-member channel response rates and dynamically routes messages to the highest-engagement channel for that individual will consistently outperform a platform that blasts on a single channel. This is not a sophisticated insight — it is basic operational hygiene that agentic AI enforces automatically.

Customer Feedback and Continuous Optimisation

The most dangerous assumption in loyalty programme management is that a well-configured AI system, once deployed, can be left to run unattended. AI loyalty agents improve dramatically over time — but only if they are fed a continuous stream of outcome signals that allow them to distinguish what worked from what did not. Building that feedback architecture is not a technical afterthought; it is a strategic programme management discipline.

The primary feedback signals for loyalty AI are explicit and implicit. Explicit signals include member ratings on reward redemption experiences, in-app satisfaction prompts after a personalised offer is delivered, and customer service ticket themes ('I never use these points', 'Why do I keep getting offers for a category I never buy in?'). These are gold. Indian mall operators and brand loyalty teams systematically under-collect explicit feedback because they fear it — they worry the NPS scores will be embarrassing. They should collect it anyway, because a 3.2 NPS on a loyalty programme is vastly more valuable than ignorance.

Implicit signals are equally powerful and entirely passive to collect: offer open rate by channel, click-through rate to the reward redemption page, redemption completion rate (started but not completed versus fully redeemed), and time-to-next-visit after a reward is received. An AI loyalty agent that delivers a bonus-points offer to a member and sees zero click-through three times in a row should autonomously downgrade that offer type and experiment with an alternative — a complimentary service upgrade, a charity donation option, or a different category reward — without waiting for a campaign manager to spot the pattern.

In India's multi-brand retail context, aggregate anonymised feedback from cross-tenant interactions creates additional optimisation surface. If data from Lifestyle stores in a particular mall cluster shows consistently higher redemption rates on 'experience rewards' (beauty consultations, styling sessions) versus 'discount rewards', that signal should propagate across the entire AI system and influence how agents construct offers for similar member profiles across other fashion tenants in the same ecosystem. This is the network effect that platform-level loyalty AI creates — and it is an advantage that a single-brand in-house loyalty solution simply cannot replicate.

Continuous optimisation also means deprecating what does not work. One of the most value-destructive habits in loyalty management is 'offer inflation' — adding new reward mechanics on top of old ones until members are confused and programme complexity kills engagement. AI agents that track member cognitive load signals — declining login frequency, increasing time-to-redemption, rising unsubscribe rates — can flag when a programme's complexity has outpaced its member base's willingness to engage, prompting a simplification intervention before churn accelerates.

How Fundle solves this

Fundle was designed from first principles to address the exact gap that legacy loyalty platforms — Capillary, EasyRewardz, Almonds.ai, Customer Capital — cannot close: the gap between data volume and personalisation velocity. The Fundle AI Platform is not a campaign automation tool with a machine learning layer bolted on. It is an agentic AI system in which AI-powered customer loyalty agents operate autonomously across the full member lifecycle, from first-visit activation through long-term advocacy, without requiring a campaign manager to manually configure every intervention.

The Fundle Loyalty Platform's core is the AI Personalisation Engine — the system that already engages 1.33 Cr+ members with dynamic rewards calibrated to individual behavioural profiles. Fundle Mall Loyalty is the vertical-specific deployment that enables mall operators at the scale of Phoenix, Inorbit, or any Tier-1 or Tier-2 mall cluster to run a single unified loyalty programme across 150–300 tenants, with AI agents optimising simultaneously for mall-level footfall and individual brand-level revenue. Fundle Brand Loyalty serves standalone retail brands — an Apollo Pharmacy, a Manyavar, a Lenskart — that need a dedicated, AI-native loyalty programme without the overheads of building and maintaining a data science team.

Fundle AI Agents are the execution layer: discrete, goal-directed AI workers that handle specific tasks — lapsing member reactivation, new member onboarding, post-purchase upsell orchestration, cross-tenant reward recommendation — each optimising for its assigned outcome while sharing a unified member intelligence graph. Fundle Agentic AI is the architectural philosophy that ties these agents together: a multi-agent orchestration system in which agents collaborate, hand off context, and learn from each other's outcomes. When the reactivation agent successfully brings a member back, it passes enriched context to the post-purchase agent, which picks up the conversation without any member-facing discontinuity. Fundle AI Workflow is the operational layer that allows mall marketing directors and CRM heads to configure, monitor, and audit agent behaviour — maintaining human oversight without requiring human execution of every decision.

Vineet Narang's founding vision for Fundle was specific: build the loyalty intelligence infrastructure that makes every Indian retailer — from a 5-store ethnic wear chain in Surat to a 200-brand mall in Hyderabad — capable of the same quality of personalised member engagement that only the largest global retailers could previously afford. That vision is now live at scale. For retail CRM heads and mall marketing directors evaluating their next engagement stack, the relevant question is not whether AI loyalty agents deliver ROI — the evidence is unambiguous. The question is which platform gives you agentic AI depth, Indian retail context, and the integration breadth across POS ecosystems and DPDP-compliant data infrastructure to operationalise it within a quarter.