An ISO/IEC27001:2013 and ISO 27018:2019 certified cloud solution
80 Key Customer Loyalty Stats & Trends to Watch in 2026 | Perx
Blogs
Banks Are Deploying AI. Almost None Are Using It to Change Customer Behaviour.
TL;DR — Read This First
The problem: BCG’s 2025 report identifies hyper-personalised customer engagement as the defining characteristic of the AI-first bank — worth $370B in profit potential by 2030. Most banks are spending AI budgets on back-office automation instead.
The gap: Deploying AI reduces cost. Using AI to engineer customer behaviour drives revenue. These are not the same thing, and most banks are only doing the first.
The three stages: BCG’s framework — Deploy → Reshape → Invent — maps exactly to this gap. Almost all BFSI institutions are stuck between Deploy and Reshape. The profit opportunity lives in Invent.
What’s required: Genuine hyper-personalisation needs four things: individual-level behavioural data, real-time triggers, mechanic design that changes behaviour, and attribution that closes at the P&L. Most banks have the first. Almost none have built the last three.
The earn-and-burn trap: Adding AI on top of a points program makes targeting more efficient but doesn’t change the underlying problem — you’re buying behaviour you don’t own.
The bottom line: The banks that win the AI era will be the ones who use it to engineer customer behaviour, not just automate customer interactions. The window to build that capability is open now.
BCG’s 2025 Global Retail Banking Report puts a number on what’s at stake: more than $370 billion in annual profit potential from AI by 2030. The industry has read it. Leadership teams have discussed it. Strategy decks have been updated.
And then most banks went back to automating their back office.
That is not a criticism of automation. Cutting cost-to-income ratios from 60%-plus down toward the 35% that well-run digital banks already operate at is a legitimate and urgent priority. AI agents handling collections, document extraction, sanctions screening — these deliver real P&L impact and the industry is right to pursue them.
But there is a second number buried inside that same BCG report, and it is the one most banks are not acting on. BCG names hyper-personalised customer engagement as a defining characteristic of the AI-first bank — not as a feature, but as the characteristic most likely to determine which banks pull away from the competition and which ones spend the next decade watching margin compress.
Fewer than 20% of banks have established quantified AI targets. Fewer than 15% have rebalanced their investment portfolios toward technology-driven revenue initiatives.
The industry knows what the destination looks like. It is not moving toward it.
This post is about why — and about what the actual path looks like for BFSI enterprises that are ready to use AI to change what customers do, not just to reduce what the bank spends.
The AI-First Bank Has a Customer Engagement Problem
Quick answer: Most banks are directing AI investment at operational cost reduction — fraud screening, document processing, back-office automation. BCG’s AI-first bank model requires a sixth dimension that most are skipping entirely: hyper-personalised customer engagement, which is a revenue driver, not a cost lever. Fewer than 20% of banks have quantified targets for this.
BCG describes the AI-first bank across six dimensions: hyper-personalised customer engagement, adaptive financial solutions, invisible embedded interfaces, autonomous operations, real-time risk and capital allocation, and a lean human core.
Five of those six are fundamentally operational. They reduce cost, compress headcount, and move decisions that humans used to make into algorithm-driven workflows. They are important. They are also where virtually all the AI investment is going right now.
Hyper-personalised customer engagement is different. It is not a cost play. It is a revenue play — and it requires something that automation cannot deliver on its own: a system designed to change what customers do, not just to respond to what they have already done.
📊 Stat Callout
“Retail bank revenue growth is forecast to slow to just 2–4% per year through 2029.” — BCG Global Retail Banking Report, 2025
At that growth rate, cost reduction alone does not close the gap. Revenue has to come from the customers you already have.
The distinction matters because the financial pressure banks face is not only a cost problem. Revenue growth across retail banking is forecast to slow to 2–4% per year through 2029. At that rate of growth, cost reduction alone does not save you. You need to grow revenue from the customers you already have — through deeper activation, higher product adoption, stronger repayment behaviours, and CLTV expansion that does not require proportionally growing the rewards budget.
That is a customer engagement problem. AI can solve it. But only if the AI is applied to the right layer.
📖 Definition
Hyper-personalised customer engagement — A customer engagement model in which every interaction is designed around the specific behavioural signals of an individual customer in real time, with mechanics built to drive the specific customer actions that produce measurable business outcomes. Distinguished from personalised marketing (which improves targeting) by its focus on behaviour change rather than message relevance.
Deploy, Reshape, Invent — Where Customer Engagement Actually Lives
Quick answer: BCG’s three-stage AI adoption framework — Deploy, Reshape, Invent — maps directly to where customer engagement value is created. Most BFSI enterprises are in Deploy (automating routines). The profit opportunity for customer engagement is in Invent — engineering customer behaviour at scale. The gap between these two stages is where most banks’ AI strategies currently fail.
BCG frames AI adoption across three stages. Most banks are in the first one.
Deploy is where you automate daily routines. Chatbots answer account queries. AI screens transactions for fraud. Documents get extracted and summarised without a human reading them. Cost comes down. Throughput goes up. This is real value — but it is value extracted from existing processes, not value created from customer behaviour.
Reshape is where AI starts to touch how the bank works with customers end-to-end. Personalised offers surface in the right channel at the right moment. Credit processes run faster because customer context is available in real time. Campaigns adapt to individual behaviour rather than segment averages. This is where most banks aspire to be, and where the minority of leading institutions are beginning to operate.
Invent is where entirely new value is created. The bank stops responding to customer behaviour and starts engineering it. Customer journeys are designed around the specific actions the bank needs customers to take — and the mechanics, triggers, and incentives are built to make those actions happen, at scale, without IT touching every campaign. MAU lifts. Cross-sell conversion moves. CLTV expands. And every outcome is traceable back to the engagement investment that produced it.
The gap in the BFSI market right now is not between Deploy and Reshape. The gap is between Reshape and Invent — specifically at the customer engagement layer.
Banks are spending on AI infrastructure. They are not spending on engagement design. And without engagement design, the infrastructure has nothing valuable to act on.
Definition
Behavioural loyalty — A loyalty model in which customer retention is built through repeated, designed behaviours rather than transactional rewards. The bank engineers the habit first; the reward, if any, reinforces an action the customer already values. Distinct from earn-and-burn loyalty, where the reward is the entire incentive and behaviour stops the moment the promotion ends.
What "Hyper-Personalised Engagement" Actually Requires
Quick answer: Hyper-personalisation in banking is not a recommendation engine or a chatbot with a customer’s name in it. It requires four things operating simultaneously: individual-level behavioural data, real-time triggers connected to journey mechanics, mechanic design that drives specific behaviours, and outcome measurement that traces engagement spend back to revenue. Most banks have the data. Almost none have built the other three.
The term hyper-personalisation has been absorbed into every vendor pitch in the market, which means it has started to mean very little. Let’s be precise about what it actually requires — because the requirements are more demanding than most loyalty and engagement programs are currently set up to meet.
Genuine hyper-personalisation in a banking context means four things working together simultaneously.
Behavioural data at the individual level. Not segment averages. Not demographic proxies. Real signals — what this customer did, in what sequence, at what frequency, and what that pattern predicts about what they will do next. Most banks have this data. Most banks are not using it at the engagement layer.
Real-time triggers connected to journey mechanics. The moment the signal fires is the moment the engagement needs to land. A customer who just made their third cross-border payment in a week responds to a travel insurance nudge differently than a customer who has never used international transfer. Static campaigns cannot capture this. Rules-engine-driven journey orchestration can.
Mechanic design that drives the behaviour the bank needs. This is the piece that is almost entirely absent from the current AI conversation in banking. It is not enough to know what a customer is likely to do — you need mechanics that make them more likely to do the thing that matters to the business. That is behavioural design. It is a discipline. It is learnable. And it is the difference between a personalised notification and a behaviour change.
Outcome measurement that closes the attribution loop. If engagement spend cannot be traced to a revenue outcome — activation, repayment, cross-sell conversion, MAU — then it is not defensible at board level.The measurement layer is not a reporting add-on. It is the mechanism that tells the next campaign what worked.
These four requirements do not describe a technology stack. They describe an engagement operating model. AI enables all four — but it cannot substitute for the design layer underneath them.
📊 Stat Callout
“Fewer than 20% of banks have established quantified AI targets. Fewer than 15% have rebalanced their investment portfolios toward technology-driven revenue initiatives.” — The Financial Brand, 2025
The gap is not ambition. It is execution at the engagement layer.
Why Earn-and-Burn Cannot Scale into the AI Era
Quick answer: Banks adding AI on top of legacy points programs are optimising the wrong system. Earn-and-burn trains customers to engage only when the reward is large enough — which means the bank is buying behaviour it doesn’t own. AI personalisation makes this more efficient but doesn’t change the logic. The AI-first bank runs on habit, not cashback. Behaviour change is the cause; the reward, if any, is the output.
Banks that are adding AI capability on top of legacy points programs are making a specific and expensive mistake. They are optimising the wrong system.
An earn-and-burn program is built on a transactional logic: do something the bank wants, receive a reward. The reward is the incentive. The problem is that this logic trains customers to engage when the reward is large enough and disengage the moment it is not. The result is a loyalty liability on the balance sheet, a rewards budget that grows without growing CLTV, and a customer relationship defined entirely by the last promotion the bank ran.
Adding AI personalisation to this model produces a more efficient version of the same problem. The bank now knows which customers to target with which offer at which moment — and is still using that knowledge to buy behaviour it does not own. The moment the cashback stops, the behaviour stops.
📖 Definition
Earn-and-burn — A loyalty program model where customers accumulate points or rewards by completing transactions, then redeem them for benefits. The engagement is entirely incentive-driven: without the reward, the behaviour typically stops. Earn-and-burn programs create loyalty liabilities on the balance sheet rather than genuine retention, and scale the rewards budget proportionally with program growth rather than with business outcomes.
The AI-first bank BCG describes does not run on cashback. It runs on habit — on customer journeys designed to make specific behaviours routine, intrinsically valued, and independent of the promotional cycle. That is a fundamentally different design problem from improving targeting. And it requires a fundamentally different platform to solve.
The Engagement Operating Model for an AI-First BFSI Bank
Quick answer: Moving from Deploy to Invent at the customer engagement layer is an operating model problem, not a technology acquisition. In practice, it means four things: defining behavioural targets before building mechanics, building journey triggers around real-time signals rather than campaign calendars, running campaigns without IT bottlenecks, and measuring ROI at the revenue line — not in clicks or open rates.
Moving from Deploy to Invent at the customer engagement layer requires an operating model — not a technology acquisition. Here is what that model looks like in practice for Tier 1 BFSI enterprises in APAC markets.
Start with the behaviour, not the reward. Define the specific customer actions that produce measurable business outcomes — first product activation within 30 days, second loan repayment on time, cross-sell conversion from savings to investment products. These are the behavioural targets. Every mechanic is built to move one of these numbers. [How Perx and UOB engineered banking engagement that moved MAU in 90 days]
Build the journey around the trigger, not the calendar. Campaign calendars are a Deploy-era artefact. Real-time behavioural triggers — a customer checks their credit score for the second time in a week, a wallet user makes their first merchant QR payment, a savings account holder crosses a deposit threshold — are the moments where engagement lands with precision. The platform needs to respond in real time, without an IT ticket.
Run without IT bottlenecks. Bank-grade compliance and no-code campaign execution are not in conflict — but most legacy systems make them behave as if they are. A marketing team that needs six weeks of IT resourcing to launch a behaviour-triggered campaign cannot operate at AI speed. The engagement layer needs to move as fast as the data does.
Close the attribution loop at the revenue level. Engagement ROI measured in clicks and open rates does not survive a CFO review. The metric that matters is the behaviour change — did activation go up? Did cross-sell conversion move? Did repayment rate improve? — and the revenue that behaviour change produced. Every engagement investment should produce a number that belongs in a board deck.
This is what the Reshape → Invent movement looks like at the engagement layer. It is not a technology project. It is a design and measurement project that technology enables.
📊 Stat Callout
“BCG estimates AI implementation can slash banks’ costs by as much as 40% — while the $370B profit opportunity from AI by 2030 remains largely untapped at the engagement layer.” — BCG Global Retail Banking Report, 2025
The cost story is being captured. The revenue story is still available.
Conclusion
BCG’s $370 billion number is real. The AI-first bank is not a concept — it is a competitive position that a small number of institutions will occupy by 2030, and the rest will compete against.
The banks that get there will not be the ones with the most sophisticated AI infrastructure. They will be the ones who directed that infrastructure at the right problem: engineering customer behaviour that drives revenue, not just automating customer interactions that reduce cost.
Hyper-personalised customer engagement is not a feature of the AI-first bank. It is the revenue engine of the AI-first bank. And it requires an engagement operating model — behavioural design, real-time orchestration, compliance-grade execution, and attribution that closes at the P&L — that most banks do not yet have in place.
The window to build it is open. The institutions moving now are already pulling ahead on MAU, activation, and CLTV. The ones waiting for the AI infrastructure to mature before addressing the engagement layer are building the wrong foundation.
Key Takeaways — What You Just Read
- BCG’s $370B AI profit opportunity is real — but most banks are chasing the wrong part of it. Back-office automation delivers cost reduction. Customer engagement engineering delivers revenue growth. These require different strategies.
- The Deploy → Reshape → Invent ladder maps directly to where engagement value lives. Most banks are in Deploy. The engagement profit opportunity is in Invent. The gap between the two is the Behaviour Gap.
- Hyper-personalisation requires four layers — not one. Individual behavioural data + real-time triggers + mechanic design + P&L-level attribution. Most banks have the data. Almost none have built the other three.
- AI on top of earn-and-burn is the wrong investment. It improves the efficiency of a system that is already failing. The foundation needs to change before the AI can do useful work on it.
- The engagement operating model is not a technology project. It is a design and measurement project that technology enables. Banks that treat it as a procurement decision will still be in Deploy by 2030.
Ready to map your own engagement gap? Run the Tier 1 Engagement Audit — it takes 10 minutes and shows you exactly where your program is leaking revenue.
If you are ready to understand exactly where your current engagement model is leaving revenue behind — and what the path to Invent looks like for your business — talk to the Perx team.
Frequently Asked Questions
What is AI-driven customer engagement in banking?
AI-driven customer engagement in banking uses real-time behavioural data, predictive models, and automated journey mechanics to change what customers do — not just to respond to what they have already done. The distinction is critical: AI applied to customer engagement should increase activation, product adoption, and CLTV by engineering specific customer behaviours, not just by automating service interactions. A bank using AI to flag a cross-sell opportunity is doing service automation. A bank using AI to trigger a journey that makes a customer 40% more likely to activate a second product within 30 days is doing customer engagement.
How is hyper-personalisation in banking different from personalised offers?
Personalised offers target the right customer with the right product at the right moment — this is improved targeting. Hyper-personalisation goes further: it uses real-time behavioural signals to design the entire customer journey, not just the offer. It requires four things working together: behavioural data at the individual level, real-time trigger architecture, mechanic design that changes behaviour, and outcome measurement that traces engagement spend back to revenue. Most banks have achieved the first. Almost none have built the last three. The difference between the two is the difference between a smarter notification and a behaviour change
What is the difference between deploying AI and using AI to change customer behaviour?
Deploying AI reduces the cost of existing operations — automating document processing, fraud screening, and customer service routing. This is the Deploy stage in BCG’s three-stage model and is where most banks currently operate. Using AI to change customer behaviour requires a different design layer entirely: journey mechanics built around specific behavioural targets, real-time triggers connected to those mechanics, and measurement that closes the loop between engagement investment and revenue outcome. BCG calls this the Invent stage. The gap between Deploy and Invent is where most banks’ AI strategies currently stall — and where the $370B profit opportunity largely sits.
What is the Deploy, Reshape, Invent framework in banking AI?
BCG’s Deploy–Reshape–Invent framework describes three stages of AI maturity in banking. Deploy means automating existing routines — fraud detection, document processing, chatbots. Reshape means optimising customer-facing processes end-to-end — personalised offers, faster credit decisions, real-time campaign adaptation. Invent means engineering entirely new value — designing customer journeys that change behaviour, creating habit loops independent of promotional budgets, and building engagement infrastructure that traces every customer action to a revenue outcome. Most banks are in Deploy. The Reshape-to-Invent movement at the customer engagement layer is where competitive advantage will be determined through 2029.
Why do earn-and-burn loyalty programs fail in an AI-first banking model?
Earn-and-burn programs train customers to engage when the reward is large enough and stop when it is not — creating a loyalty liability rather than genuine retention. Adding AI personalisation to this model improves targeting efficiency but does not change the underlying logic: the bank is still buying behaviour it does not own. When the cashback stops, the behaviour stops. The AI-first bank requires a behaviour-led design layer where customer actions become habit, not incentive-response. Points are the output of loyalty. Behaviour change is the cause. Banks that build AI personalisation on top of earn-and-burn are optimising the wrong system.
How do banks measure the ROI of AI customer engagement?
The correct measurement framework connects engagement investment directly to revenue-line outcomes: activation rate within a defined window (typically 30–90 days), cross-sell conversion lift, repayment behaviour improvement, MAU growth, and CLTV expansion over a 12-month period. Engagement metrics measured in clicks, open rates, or points issued are not board-level numbers — they measure activity, not outcome. The metric that survives a CFO review is the behaviour change and the revenue it produced. Every engagement investment should be traceable from the mechanic that triggered the behaviour to the P&L line it moved.
What does a real-time behavioural trigger look like in banking?
A real-time behavioural trigger is a specific customer action — or pattern of actions — that automatically initiates a personalised engagement journey. Examples: a customer checks their credit score twice in one week (trigger: home loan journey); a wallet user completes their first QR payment (trigger: merchant rewards activation); a savings account holder crosses a deposit threshold for the third consecutive month (trigger: investment product cross-sell). The trigger fires the moment the signal appears — not at the next campaign cycle, not in the next weekly batch. The platform responds in real time, without an IT ticket, with a mechanic designed specifically for that behaviour.
How does behavioural loyalty differ from traditional loyalty programs?
Traditional loyalty programs — typically earn-and-burn point systems — are transactional: customers receive rewards for completing specific actions. The engagement is entirely incentive-driven and stops the moment the reward is removed. Behavioural loyalty is designed differently: the goal is to make specific customer actions routine, intrinsically valued, and independent of the promotional cycle. Journey mechanics, gamification, and real-time triggers build the habit first; any reward reinforces an action the customer already values. The result is a customer relationship that does not need a cashback budget to sustain it — and a loyalty investment that appears on the right side of the P&L.
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