An ISO/IEC27001:2013 and ISO 27018:2019 certified cloud solution
80 Key Customer Loyalty Stats & Trends to Watch in 2026 | Perx
Blogs
What Is Agentic AI in Loyalty Programs?
A Complete Guide for BFSI and Telco Leaders
Perx Technologies is a B2B SaaS loyalty and customer engagement platform serving BFSI and telco clients across APAC, redefining what a loyalty platform can be — from campaign execution to Autonomous Revenue Intelligence — powered by agentic AI and purpose-built for the revenue intelligence needs of BFSI and telco organisations across APAC.
IN BRIEF
- Agentic AI goes beyond generative AI — it takes autonomous, goal-directed action without requiring a human prompt at every step.
- For BFSI and telco leaders, agentic AI transforms loyalty from a cost centre into an autonomous revenue growth engine — detecting spend declines, reactivating lapsed customers, and optimising incentives in real time.
- 67% of loyalty programme operators say they are comfortable using AI-powered agents to manage their programmes (Antavo Global Loyalty Report 2025). The window to build first-mover advantage is now.
For most of the last decade, loyalty programmes in banking and telecoms operated on a familiar loop: plan a campaign, build the rules, launch, wait, review results, and repeat. Smart marketing teams learned to get faster at this loop. Platforms got better at automating parts of it. But the fundamental model — human-designed, human-triggered, human-reviewed — stayed intact.
That model is now obsolete.
Not because the people running loyalty programmes lack talent. But because the competitive environment has shifted so dramatically, a monthly or even weekly campaign cadence is no longer fast enough to keep pace with individual customer behaviour. A customer who begins showing signs of spend decline on a Tuesday doesn’t need a campaign next week. They need a relevant, personalised intervention on Wednesday — before the behaviour becomes a habit.
Agentic AI makes that possible. And for BFSI and telco organisations operating at scale, it is rapidly becoming the defining capability that separates platforms that drive revenue from platforms that merely report on it.
This guide explains what agentic AI is, how it differs from the AI tools your teams are already using, what it means specifically for loyalty programmes in banking and telecoms, and how to think about readiness for this transition.
What Is Agentic AI? A Clear Definition
Agentic AI refers to artificial intelligence systems designed to pursue goals autonomously — planning, deciding, and acting across multi-step workflows without requiring a human to prompt each individual step.
This is meaningfully different from the AI most marketing teams encounter today. To understand why, it helps to map the three generations of AI capability:
| Generation | What it does | Loyalty example |
|---|---|---|
| Rule-based automation | Executes predefined if/then logic. Fast but rigid. | "Send birthday email if customer birthday = today" |
| Generative AI | Produces outputs (text, images, summaries) based on a prompt. Requires human direction. | "Write three subject line variations for our points expiry campaign" |
| Agentic AI | Pursues goals autonomously — detecting signals, making decisions, taking action, and learning from outcomes. No human prompt required per action. | Detects a customer's spend declining, identifies the optimal intervention, launches a personalised spend-booster challenge, adjusts the incentive in real time, and reports the revenue outcome — all autonomously. |
The critical distinction: generative AI is a tool you use. Agentic AI is a system that works.
When a bank’s marketing team asks generative AI to help with a retention campaign, they are still doing the thinking — writing the brief, reviewing the output, deciding what to launch, monitoring results. Agentic AI flips this. The system observes customer signals, reasons about what action will maximise the desired outcome, executes that action, and improves its own decision logic based on what worked. The human defines the goal. The agent handles the execution.
Why Agentic AI Matters Specifically for BFSI and Telco
Every industry will be affected by agentic AI. But BFSI and telco have three structural characteristics that make the opportunity — and the urgency — particularly acute.
1. Transaction volume creates the signal density that agentic AI needs
Agentic AI systems require rich, continuous behavioural signals to make good decisions. Banks and telcos sit on exactly this kind of data: card transactions, payment patterns, data usage behaviours, channel interactions, product uptake signals. A retail loyalty programme might see a customer interact a few times a week. A bank sees them dozens of times a day across channels. That signal density is what makes agentic AI interventions precise rather than generic.
2. The cost of customer inaction is high
In most retail settings, a lapsed loyalty member is an inconvenience. In banking, a customer who quietly reduces their primary account usage, transfers savings to a competitor, or stops using their credit card represents significant lifetime value at risk — often in the tens of thousands of dollars before the trend is even visible in a monthly report. Agentic AI detects these shifts in days rather than weeks, enabling intervention before the behaviour becomes a pattern.
3. Regulatory and compliance constraints make autonomous AI governance critical
BFSI organisations operate under strict regulatory requirements around customer communications, incentive disclosures, and data usage. Well-designed agentic AI systems respect these constraints by operating within defined guardrails — they do not override compliance rules but instead optimise within them. The key is selecting platforms where governance and audit trails are built into the agent architecture from the ground up.
The Agentic AI Market Opportunity
44%
of finance teams will use agentic AI in 2026
Wolters Kluwer
$3.50
average return for every $1 invested in agentic AI
KPMG 2025
67%
of loyalty programme operators comfortable using AI agents
Antavo Global Loyalty Report 2025
The Agentic Loyalty Stack: A Framework for BFSI and Telco
Understanding where agentic AI fits within a loyalty platform requires thinking in layers. At Perx, we describe this as the Agentic Loyalty Stack — a four-layer value architecture that maps the progression from basic execution to fully autonomous revenue intelligence.
| Layer | Name | What it includes | AI Maturity |
|---|---|---|---|
| 1 | Execution Layer |
Perx Core Platform Campaign orchestration, rewards management, gamification mechanics, engagement journeys, APIs and integrations |
Rule-based automation |
| 2 | Data & Signal Layer |
Transaction Evidence Engine Card transactions, purchase events, app engagement signals, reward redemptions, product usage data — real behavioural intelligence, not survey data |
Predictive analytics |
| 3 | Intelligence Layer |
Customer Command Centre Real-time ROI dashboards, behavioural analytics, campaign performance intelligence, predictive forecasting, and revenue attribution — turning raw signals into decisions |
Generative AI + ML |
| 4 | Autonomous Revenue Layer |
Agentic Automation Autonomous AI agents that detect opportunities, execute interventions, and continuously optimise without human campaign management: Spend Acceleration, Customer Reactivation, Cross-Sell, Reward Optimisation, Campaign Auto-Pilot, and Gamified Engagement agents |
Agentic AI ✔ |
Most BFSI and telco organisations today operate firmly in Layer 1, with some use of Layer 2 predictive analytics. The competitive advantage over the next two to three years will be determined by how quickly they progress to Layers 3 and 4.
What Do Agentic AI Agents Actually Do in a Loyalty Context?
The best way to understand agentic AI is not through definitions but through examples. Below are the six autonomous revenue agents that sit at the top of the Agentic Loyalty Stack, and what each one does in practice.
1. The Spend Acceleration Agent
What triggers it: A statistically significant decline in a customer’s spending velocity — detected in real time against their historical baseline.
What it does: Identifies the customer as high-priority, selects the optimal intervention (a personalised spend-booster challenge, a limited-time reward tier, a category-specific bonus), launches it without waiting for a human campaign review, adjusts incentive value based on engagement signals, and closes the loop with an attributed revenue outcome.
The vision: A spend acceleration agent operating on Perx’s existing transaction signal infrastructure could target spend decline patterns and deploy personalised interventions in real time — the kind of outcome that currently requires significant manual campaign effort to approximate.
What this makes possible: Jenius (part of SMBC Indonesia) used Perx’s gamified loyalty mechanics to lift customer credit card spend 67% above Indonesia’s national average — US$460 per customer per month versus the national average of US$275 (source: Global Data). The rules engine triggered 13.4 million spend interactions over eight months, generating US$599 million in actual customer transactions. That volume of real-time transaction signal is exactly the data environment a spend acceleration agent would need — detecting individual spend declines as they happen and intervening before they become a pattern, rather than waiting for a human to spot the trend in a monthly report.
2. The Customer Reactivation Agent
What triggers it: A customer crossing a dormancy threshold — a defined period of inactivity relative to their own historical engagement pattern, not a generic 30-day rule applied uniformly.
What it does: Deploys a personalised win-back sequence — not a mass email, but an individually tailored offer based on the customer’s transaction history, product preferences, and past response patterns. Sequences escalate if the first intervention doesn’t produce a signal.
Why it matters for telco: In markets with high SIM churn, a customer who stops using mobile data above a certain threshold is often in the consideration phase for switching providers. The reactivation agent detects this window and intervenes before the switch happens.
3. The Cross-Sell Opportunity Agent
What triggers it: Behavioural signals that indicate a customer is ready for an adjacent product — a savings account customer whose transaction patterns suggest they are carrying a credit balance with another bank, or a telco customer whose data usage patterns suggest they would benefit from a different plan.
What it does: Surfaces a contextually relevant cross-sell offer at the moment of highest receptivity — not at the end of the month when it’s convenient for the business, but when the customer’s own behaviour signals readiness. Connects loyalty incentives directly to product adoption.
The distinction from traditional cross-sell: Traditional cross-sell campaigns are broad, scheduled, and segment-level. The Cross-Sell Opportunity Agent is individualised, event-triggered, and outcome-attributed.
4. The Reward Optimisation Agent
What triggers it: Continuous monitoring of reward redemption rates, incentive cost-per-engagement, and behavioural response patterns across customer segments.
What it does: Adjusts incentive levels in real time to maximise engagement at minimum cost. If a segment is responding strongly to lower-value rewards, the agent reduces incentive spend for that segment and reallocates budget to segments that require higher motivation. This is yield management applied to loyalty economics.
5. The Campaign Auto-Pilot Agent
What triggers it: Ongoing campaign performance monitoring — tracking open rates, conversion rates, attribution outcomes, and comparative performance across variants.
What it does: Manages scheduling, targeting adjustments, and budget allocation autonomously. Identifies underperforming campaign elements and either replaces them with better-performing variants or escalates for human review when performance falls outside expected thresholds. Keeps campaigns optimised between manual review cycles.
6. The Gamified Engagement Agent
What triggers it: A drop in engagement metrics — session frequency, challenge participation rates, reward catalogue browsing — that signals a customer’s interest in the programme is waning.
What it does: Dynamically designs and launches gamification mechanics — challenges, missions, leaderboards, surprise rewards — tailored to the specific engagement patterns of the individual customer. Sustains programme momentum between major campaign cycles without requiring a marketing team to design each intervention.
What this makes possible: Singapore’s top neo bank — one of the world’s fastest-growing digital banks, with over 700,000 customers — used Perx’s gamified rules engine and reward-led in-app journeys to generate $6.6 million in campaign-driven transactions, delivering a 2x ROI on Perx platform costs. The same engagement mechanics drove a 72% returning customer rate and a 70% average campaign engagement rate per user. A gamified engagement agent would take this foundation further — autonomously detecting when individual engagement signals drop and dynamically launching the next best mechanic, rather than waiting for a team to plan the next campaign cycle.
How Agentic AI Differs from What Your Platform Does Today
BFSI and telco marketing teams are sophisticated. Most are already using some form of AI or automation in their loyalty operations. It is worth being precise about what agentic AI adds that existing approaches do not.
| Capability | Traditional / Generative AI | Agentic AI |
|---|---|---|
| Trigger mechanism | Human-initiated campaign or scheduled automation | Self-initiated from live behavioural signals |
| Decision-making | Human reviews recommendations, decides action | Agent reasons, decides, and acts autonomously |
| Speed to intervention | Days to weeks (campaign cycle) | Hours to real time |
| Personalisation depth | Segment-level targeting | Individual-level, dynamically composed |
| Optimisation | Manual A/B testing, monthly review | Continuous, in-flight, without human intervention |
| Revenue attribution | Modelled, often delayed | Direct, real-time loop between action and outcome |
| Scale | Limited by team bandwidth | Unlimited — agents work across all customers simultaneously |
Is Your Organisation Ready for Agentic AI? A Readiness Checklist
Agentic AI is not a plug-in. It requires a certain level of data and platform maturity to operate effectively. Before evaluating agentic AI solutions, BFSI and telco organisations should assess their readiness across four dimensions.
Data Infrastructure Readiness
Agentic AI needs real-time or near-real-time access to transactional and behavioural data. Ask:
- Can your loyalty platform ingest card transaction data, not just reward redemptions?
- Is customer data unified across channels, or siloed by product line?
- How long does it take for a customer action to appear in your analytics environment — hours, days, or weeks?
Platform Architecture Readiness
Agentic AI requires a platform architecture that supports event-driven workflows and API-level integrations. Ask:
- Is your loyalty platform API-first, or does it require custom development for every integration?
- Can your platform trigger actions based on real-time events, or only scheduled campaigns?
- Does your rules engine support dynamic, personalised incentive construction or only pre-configured reward types?
Governance and Compliance Readiness
For BFSI organisations especially, autonomous AI systems must operate within defined regulatory guardrails. Ask:
- Can you configure compliance constraints — communication frequency caps, incentive disclosure rules, regulatory limits — at the platform level?
- Does the platform maintain full audit trails of autonomous decisions for regulatory review?
- Can human oversight be triggered when agent confidence falls below a defined threshold?
Strategic Readiness
Agentic AI changes how marketing teams operate — not by replacing them, but by shifting their focus. Ask:
- Does your leadership have clarity on which customer outcomes agentic AI should optimise for — revenue uplift, churn reduction, product adoption, or a combination?
- Are your marketing teams ready to shift from campaign management to goal definition and performance governance?
- Is there executive buy-in for a phased investment in intelligence and automation layers, not just the execution layer?
The Competitive Landscape: Where Agentic AI Is Heading in Loyalty
Agentic AI in loyalty is not a future concept. It is being deployed now, and the market is moving fast.
Antavo launched Timi AI — described as the world’s first agentic AI for loyalty programmes — in 2025, positioning it as a virtual loyalty assistant capable of autonomous programme management. Capillary Technologies introduced multi-agent architectures for campaign orchestration under their AI-First Loyalty framework, building on their aiRA assistant and Nudge Framework. Salesforce launched Agentforce — a horizontal agentic AI platform — with banking-specific role-based agents.
What the market has not yet produced is an agentic AI system built specifically around the revenue intelligence needs of BFSI and telco organisations in APAC — where transaction signal density, regulatory complexity, and the specific economics of banking and telco loyalty create a distinct set of requirements that general-purpose platforms were not designed to serve.
That is the gap Perx is actively exploring — and we’d like to think through what it means for BFSI and telco organisations alongside you.
Key Takeaways
What BFSI and Telco Leaders Need to Remember
- Agentic AI is not a future trend — it is being deployed by loyalty platform vendors now, and the market is moving fast.
- For BFSI and telco, the combination of rich transaction data and high customer lifetime value makes agentic AI interventions both highly effective and economically essential.
- The Agentic Loyalty Stack provides a clear progression path: from execution to signals to intelligence to autonomous revenue — organisations should assess where they sit today and plan accordingly.
- Readiness requires data infrastructure, platform architecture, compliance governance, and strategic clarity — not just a technology selection decision.
- The organisations that build agentic AI capability now — before it becomes standard — will define the loyalty ROI benchmarks that everyone else is measured against in 2027 and beyond.
Frequently Asked Questions
What is the difference between agentic AI and generative AI in loyalty programmes?
Generative AI creates content — campaign copy, subject line variations, personalised messages — when a human asks it to. Agentic AI takes action — detecting behavioural signals, making decisions, launching interventions, and adjusting them in real time — without waiting for a human prompt. Generative AI is a tool your team uses. Agentic AI is a system that works independently toward your defined revenue goals.
Is agentic AI safe to deploy in a regulated banking environment?
Yes, when implemented on a platform designed with compliance as a first-class requirement. Agentic AI systems in BFSI operate within configurable guardrails — communication frequency caps, incentive disclosure rules, and regulatory limits — and maintain full audit trails of every autonomous decision. The key is selecting a platform where governance architecture is built in, not bolted on.
How long does it take to see results from agentic AI loyalty agents?
Initial results from individual agents — such as spend acceleration or customer reactivation — can be visible within weeks of deployment, as these agents act on live customer signals rather than waiting for a campaign cycle. Broader revenue intelligence outcomes, including cross-sell uplift and programme-level ROI improvements, typically materialise within one to two quarters as the agents accumulate decision-making experience.
Does agentic AI replace loyalty programme managers and marketing teams?
No. Agentic AI replaces the most time-consuming and repetitive aspects of campaign management — the constant monitoring, adjusting, and re-launching — freeing marketing teams to focus on strategy, creative direction, and the high-judgment decisions that genuinely require human expertise. The role of the loyalty professional shifts from campaign operator to goal architect and performance governor.
What data does agentic AI need to function effectively in a banking loyalty context?
The richer the signal, the better the decisions. At a minimum, agentic AI systems benefit from real-time or near-real-time access to transaction data (amount, merchant category, frequency), product usage signals (account types held, feature adoption), channel interaction data (app sessions, branch visits, call centre contacts), and historical campaign response data. Platforms that only ingest reward redemption events provide insufficient signal for effective agentic operation.
How does agentic AI handle customers who should not receive automated interventions?
Well-designed agentic AI systems include suppression logic that excludes customers from automated interventions based on defined criteria — recent service complaints, open disputes, regulatory exclusion lists, or manual marketing holds. Compliance teams can configure these suppression rules at the platform level, and agents respect them across all their decision-making without requiring human review of each individual case.
What is the Autonomous Revenue Intelligence Maturity Model?
The Autonomous Revenue Intelligence Maturity Model is Perx’s framework for describing the four stages of loyalty platform evolution: Stage 1 (Execution Layer — campaign operations), Stage 2 (Signal Layer — behavioural data capture), Stage 3 (Intelligence Layer — real-time analytics and revenue attribution), and Stage 4 (Autonomous Layer — agentic AI-driven revenue optimisation). Most BFSI organisations are currently at Stage 1 or Stage 2. The model provides a structured progression path toward full autonomous revenue intelligence.
Which agentic AI agent should a bank or telco deploy first?
For most BFSI organisations, the Spend Acceleration Agent or Customer Reactivation Agent provides the clearest path to measurable early ROI. Both are highly suited to the rich transaction signal environment of banking and telco, and both address high-value business problems — spend decline and customer churn — that create direct revenue impact. Starting with one agent allows teams to build confidence in autonomous decision-making before expanding to a full multi-agent deployment.
How is agentic AI different from marketing automation platforms like Salesforce or HubSpot?
Marketing automation platforms are excellent at executing pre-defined customer journeys at scale. They automate the delivery of campaigns that humans have designed. Agentic AI goes further — it designs and optimises the interventions itself, based on real-time signal interpretation. An automation platform sends a birthday email because it was configured to. An agentic AI system identifies that a specific customer is at elevated churn risk, determines the optimal intervention, launches it, and adjusts based on response — without any human having configured that specific scenario.
What should BFSI organisations look for when evaluating an agentic AI loyalty platform?
The most important evaluation criteria are: (1) real-time data ingestion capability — can the platform act on live transaction signals, not batched data? (2) governance architecture — are compliance guardrails configurable and auditable? (3) agent specificity — are the agents designed for loyalty and BFSI use cases, or are they general-purpose? (4) attribution clarity — can the platform produce direct causal attribution between agent actions and revenue outcomes? (5) APAC regulatory alignment — does the platform vendor have deep experience with the regulatory environments of your operating markets?
Recommended for you
An ISO/IEC27001:2013 and ISO 27018:2019 certified cloud solution
SAY HELLO
Perx Technologies Pte Ltd
20A Tanjong Pagar Road
Singapore 088443
An ISO/IEC27001:2013 and ISO 27018:2019 compliant cloud solution
An ISO/IEC27001:2013 and ISO 27018:2019 certified cloud solution