Easypaisa Chatbot ‘ello’: How Shake-to-Chat AI Reshaped Digital Banking Support in Pakistan

The Easypaisa chatbot is called ‘ello’. It launched in February 2026 inside the Easypaisa app and opens when a logged-in user shakes their phone — the “Shake-to-Chat” gesture. It was built on Ant International’s GenAI Cockpit platform and handles product questions, fee queries, account and transaction lookups, and basic troubleshooting 24/7. It does not replace the human helpline for disputes, fraud cases or transaction reversals, which are escalated. Easypaisa serves 1 in 5 Pakistanis and reports processing 4.5 billion transactions worth PKR 15 trillion in 2025.

The Easypaisa chatbot, branded ‘ello’, went live inside the Easypaisa app in February 2026 with an activation method no other Pakistani financial app had shipped: shake the phone while logged in, and a conversational assistant opens over whatever screen you are on. The deployment sits on a very large base. Easypaisa reports serving 1 in every 5 Pakistanis and processing more than 4.5 billion transactions worth over PKR 15 trillion in 2025 — roughly 13% of Pakistan’s GDP. The analytical point of this piece, and the reason it belongs in our AI coverage, is narrow and causal: support friction fell because activation moved from multi-step menu navigation to a single physical gesture, which means the cheapest support channel became the most discoverable one in the app.

About this analysis: Written by The Verge PK editorial desk, which covers fintech, digital payments and AI deployment in Pakistan and South Asia. This article is based on Easypaisa’s public launch disclosures, reporting by ProPakistani and Aaj News, and Ant International’s published platform description.

How to Build an Automated Data Analysis Pipeline for Physics Research: A Step-by-Step Guide – theverge.pk

Transparency note: All scale and volume figures in this article are company-reported by Easypaisa and have not been independently audited. Where Easypaisa or Ant International has not disclosed a metric — wait times, deflection rates, supported languages, data residency — this article says so rather than estimating. The Verge PK has no commercial relationship with Easypaisa, Telenor, Alipay or Ant International.

What Is ‘ello’? Easypaisa’s In-App AI Assistant Explained

‘ello’ is the AI-powered in-app chatbot of Easypaisa Digital Bank, launched in February 2026. The Easypaisa chatbot carries the name ‘ello’ and answers product, fee, account and transaction questions in natural conversation inside the Easypaisa app. It is activated by shaking the phone after login and operates 24/7 without a human agent.

Attribute	Detail
Assistant name	‘ello’
Launch date	February 2026
Operator	Easypaisa Digital Bank (Pakistan)
Technology platform	Ant International GenAI Cockpit
Activation method	Shake-to-Chat (shake phone after login)
Availability	24/7, in-app only
Access requirement	Updated Easypaisa app, logged-in account
Escalation path	Routed to Easypaisa human support channels

Why an in-app assistant rather than a bigger call centre? Because Easypaisa has no branch network to absorb support volume. As Pakistan’s first licensed digital retail bank under a State Bank of Pakistan digital bank licence, the institution’s entire customer-facing surface is the app, which means every question a branch teller would answer elsewhere lands either in the app or on the helpline. Support demand therefore scales with transaction volume, while call-centre capacity scales with headcount and shift hours — two curves that diverge as the user base grows. Conversational self-service is the only tier of that stack with marginal cost close to zero, so Easypaisa built it where the demand originates. Farhan Hassan, Chief Digital Officer at Easypaisa, framed the launch as deploying AI within operations to deliver instant support at scale, as reported by ProPakistani in 2026.

The interaction model is the thing to notice. An IVR helpline routes a caller through fixed numeric menus, and a scripted rule-based bot matches keywords to canned replies; both require the user to guess the vendor’s taxonomy. ‘ello’ accepts a free-text question and resolves it or routes it, which is the reason discoverability improved rather than merely the presence of a chat window.

How Shake-to-Chat Works Inside the Easypaisa App

Shake-to-Chat replaces menu navigation with a single physical gesture. That matters because every removed tap reduces the share of users who abandon self-service halfway and call the helpline instead — the most expensive outcome for the bank and the slowest one for the customer.

1. Update the Easypaisa app to the latest version from Google Play or the App Store.
2. Open the app and log in to your Easypaisa account.
3. Gently shake the phone — the ‘ello’ chat window opens over the current screen.
4. Type or select your query in plain language; escalate to human support from inside the chat if the issue needs it.

The design logic is about literacy, not novelty. A gesture is discoverable without reading English menu labels, and a large share of Easypaisa’s base consists of first-time formal-finance customers for whom menu hierarchies in a second language are the actual barrier to self-service. Women account for 31% of the customer base, a figure Easypaisa cites as evidence of inclusion progress and one that exists because the company has pursued users with limited prior banking exposure. Driven by that mandate, a motion-triggered, free-text interface removes two obstacles at once: finding the support entry point, and phrasing the request in the system’s vocabulary. Easypaisa has not disclosed adoption rates for the gesture, so the inclusion benefit is a design argument supported by the product’s structure rather than a published result.

One security consequence follows from the login requirement. The gesture works only post-login, therefore the assistant inherits an authenticated session rather than asking the user to identify themselves in chat — which means no account numbers, PINs or OTPs need to be typed into a conversation at all.

If Shake-to-Chat Is Not Working

• App version is outdated — install the current Easypaisa app update.
• You are not logged in — the gesture only works inside an authenticated session.
• The shake was too gentle or too slow — use a firm, short side-to-side motion.
• Device motion sensors are restricted by battery-saver or accessibility settings.
• Feature rollout has not reached your account yet — use the in-app Help Centre or helpline 3737 as the fallback.

If none of the above resolves it, the in-app Help Centre remains the documented route to a human agent.

From Chatbot to Financial Assistant: What ‘ello’ Can Actually Do

‘ello’ handles information and navigation, not adjudication — therefore it answers and routes, while disputes over money that has already moved stay with human agents.

• Product knowledge hub — fees, limits, eligibility and service details answered without contacting an agent.
• Help Centre and use-case redirection — conversational routing into the correct in-app journey instead of menu hunting.
• 24/7 self-service — support available outside call-centre staffing hours, which shifts off-peak demand away from the helpline.

User request	Handled by ‘ello’	Where it goes next
Account balance and transaction history lookup	Yes, within the logged-in session	Stays in chat
Fees, charges and transaction limits	Yes, from the product knowledge hub	Stays in chat
Product eligibility and sign-up questions	Yes, with routing into the sign-up journey	In-app journey
Failed or stuck transfer — status check	Status check and issue logging	Human agent if unresolved
Transaction reversal or refund decision	No	Human support / complaint process
Reported fraud or unauthorised debit	No	Human agent, priority escalation
Account blocked or KYC rejection	Explanation and routing only	Human support channels
Complaint registration and ticket status	Registration and status retrieval	Human agent for resolution

Page not found – theverge.pk

The boundary is drawn there for liability reasons, not technical ones. If a generative assistant tells a customer a reversal has been approved, that statement is a financial representation by a licensed bank — an incorrect one is a mis-selling or misrepresentation exposure, not a text error. Consequently Easypaisa constrains the assistant to retrieval, explanation and routing, and keeps adjudication with agents who operate inside a documented complaints process. The documented roadmap points toward digital lending, wealth management and fraud prevention, and each of those expands the liability surface rather than merely the feature list, which means guardrail design becomes the binding constraint on scope long before model capability does. On language, the assistant is built for natural conversational input, but Easypaisa has not published a supported-language list or confirmed Roman Urdu handling depth, so that coverage should be treated as undisclosed.

The Tech Stack: Why Easypaisa Built on Ant International’s GenAI Cockpit

‘ello’ runs on GenAI Cockpit, Ant International’s AI-as-a-Service platform for building generative and agentic financial applications. Easypaisa bought the platform rather than building one because the scarce input in Pakistan’s market is financial-domain-tuned model infrastructure, not engineering headcount.

A banking assistant needs five layers working together: retrieval over product and policy content so answers trace to approved sources; session-aware access to account data so a balance query resolves without re-authentication; intent routing that drops the user into the correct in-app journey; guardrails that block unsupported categories such as investment advice or credit decisions; and audit logging that makes each response reconstructable after the fact. An AI-as-a-Service layer supplies the orchestration and guardrail tooling, which means the deploying bank’s remaining work shifts to content quality, escalation design and oversight — three disciplines that are organisational rather than technical. The trade-off sits in portability, and this is where open standards in AI stop being an abstraction: the more the assistant’s behaviour is defined inside a vendor’s orchestration layer, the harder it is to move the deployment without rebuilding it.

Jiang-Ming Yang, Chief Innovation Officer at Ant International, has publicly positioned the launch as a step in supporting Easypaisa’s AI transformation and wider innovation agenda, backed by Ant International’s financial domain expertise. That characterisation is a paraphrase of his reported comments rather than a verbatim quote, because the full wording is not in the public record used for this article.

theverge.pk – AI Governance and Data Standards

The build-versus-buy consequence is concrete. Vendor platforms compress time-to-deploy from years to months, and that is precisely why Easypaisa shipped in February 2026 rather than later. They also concentrate dependency: model updates, prompt behaviour and training data decisions sit partly outside the bank’s direct control. The model provenance challenges documented in collaborative AI environments apply directly here — when model lineage sits with a vendor, the deploying institution inherits an evidence gap it must close contractually, through disclosure clauses and retention guarantees, because the regulator will ask the bank, not the platform.

Easypaisa’s Scale: 1 in 5 Pakistanis and PKR 15 Trillion in 2025 Volume

Easypaisa reports processing over 4.5 billion transactions worth more than PKR 15 trillion in 2025 — equivalent to approximately 13% of Pakistan’s GDP — across a customer base it describes as reaching 1 in every 5 Pakistanis.

Metric	Figure	Period	Source
Transactions processed	Over 4.5 billion	2025	Easypaisa (company-reported)
Total transaction value	Over PKR 15 trillion	2025	Easypaisa (company-reported)
Share of Pakistan’s GDP	Approximately 13%	2025	Easypaisa (company-reported)
Population reach	1 in 5 Pakistanis	2026	Easypaisa (company-reported)
Women as share of customer base	31%	2026	Easypaisa (company-reported)
‘ello’ chatbot launch	February 2026	2026	Easypaisa launch announcement

Convert that volume into support economics and the deployment reads as a cost-structure decision. A base of that size generates informational contacts — fee questions, limit checks, “where is my transfer” queries — at a rate no call centre scales to linearly, because each additional million users requires additional trained agents, additional shifts and additional quality assurance. Automating the informational tier therefore changes the slope of the cost curve rather than trimming a line item. The 31% women share deserves the same analytical treatment: it is high by regional standards for mobile financial services and low in absolute terms, which means the inclusion gap is narrowing without being closed. A gesture-activated, free-text interface lowers the navigation and literacy barrier that keeps first-time users dependent on agents, relatives or shopkeepers to complete basic tasks. All of these figures come from Easypaisa’s own launch disclosures as reported by Aaj News and other outlets; they are company-reported and not independently audited.

The reason such figures are published in banking terms at all is regulatory. Easypaisa’s State Bank of Pakistan digital bank licence brings disclosure and reporting obligations that a pure wallet operator does not carry, which means its scale claims sit inside a supervised reporting framework rather than purely in marketing material.

How ‘ello’ Changed Easypaisa Customer Service: Channels Compared

The assistant did not replace the helpline; it reordered the funnel. Informational contacts now resolve in-app, which means the helpline inherits a harder, lower-volume mix dominated by disputes, fraud reports and account-status cases.

Channel	Availability	How you reach it	Best suited to
‘ello’ in-app chatbot	24/7	Shake phone after login	Fees, limits, balances, status checks
Helpline 3737	Not stated	Phone call	Fraud, disputes, blocked accounts
In-app Help Centre / complaint form	24/7 submission	App menu navigation	Formal complaints, ticket tracking

“At easypaisa, our mission is to make digital financial services simple, accessible, and inclusive for every Pakistani. The launch of ello helps us deploy AI within our operations to deliver instant support, streamlining everyday digital banking, and enhancing the overall customer experience at scale.”
— Farhan Hassan, Chief Digital Officer, Easypaisa

Easypaisa has not published before-and-after wait times, deflection rates or cost per contact, therefore the verifiable change is in channel design rather than in demonstrated performance. Reasoning from the design is still informative: 24/7 availability removes the overnight queue that previously accumulated and discharged into the morning call peak; gesture activation removes the menu-hunting stage entirely; and the authenticated session removes identity verification steps that consume the opening minute of most helpline calls. Together those three removals compress a routine informational contact from minutes to seconds, which is the mechanism behind any efficiency gain the company later reports. On the broader question of whether AI displaces bank call centres in Pakistan, the evidence points one way: AI shifts agent workload toward exception handling rather than eliminating it, because disputes carry liability a model cannot assume and regulators expect a human accountable owner for each complaint. Readers should treat any uplift claim made without published metrics as a design argument, not as evidence.

Easypaisa vs JazzCash: Who Has the Better In-App Support in 2026?

On in-app support specifically, Easypaisa leads in 2026 because it ships a named generative assistant with gesture activation and documented 24/7 availability, while JazzCash’s in-app support remains menu-and-ticket driven as of publication.

Dimension	Easypaisa	JazzCash
Named AI assistant	‘ello’, launched February 2026	None publicly documented
Activation method	Shake-to-Chat after login	In-app menu navigation
Availability hours	24/7 in-app	Not publicly documented
Escalation to human agent	From inside chat, plus helpline 3737	Helpline and in-app complaint form
Regulatory status	Licensed digital retail bank (State Bank of Pakistan)	Not publicly documented as a digital bank licensee

The competitive consequence is straightforward. Core transfers, bill payments and mobile top-ups are effectively commoditised between the two wallets, therefore support experience is one of the few remaining dimensions on which either can differentiate — and the firm that resolves queries without an agent reduces cost and churn in the same move. Place this in regional context and the advantage looks temporal rather than structural: banks and wallets across South Asia are deploying conversational assistants on vendor platforms comparable to GenAI Cockpit, which means Easypaisa’s lead is a timing lead that compresses as rivals ship equivalents. We are not asserting market share figures for either platform, because current, reliable 2026 share data is not in the public record used here, and the JazzCash rows above describe publicly documented behaviour only.

Governance Questions: Data, Provenance and Accountability Behind ‘ello’

A generative assistant inside a licensed bank inherits the bank’s accountability, not the vendor’s. Easypaisa therefore owns the governance obligations for ‘ello’ regardless of who trained, hosted or tuned the underlying model.

• Data residency — where conversation and account-linked data is processed and stored.
• Model provenance — which model version, fine-tuning data and content sources produced each response.
• Hallucination liability — who is accountable when the assistant states an incorrect fee, limit or eligibility rule.
• Audit trails — retrievable, timestamped logs of chat sessions linked to authenticated accounts.
• Escalation guarantees — enforced hand-off thresholds for fraud, disputes and vulnerable-customer signals.
• Scope guardrails — blocked categories such as investment advice, credit decisions and account credentials.
• Vendor exit — portability of conversation data and content assets away from the platform provider.

5 Critical AI Governance Challenges in Multi-Institution Research Labs – theverge.pk

That checklist maps onto the four core functions of the NIST AI Risk Management Framework — govern, map, measure and manage. Govern covers ownership and escalation policy; map covers the deployment context and affected customer segments; measure covers response accuracy and guardrail breach rates; manage covers remediation when the assistant answers wrongly. Provenance logging carries particular weight in a financial deployment because a wrong fee answer is a mis-selling exposure, which means the institution has to reconstruct which model version and which content source produced a specific response months after the conversation ended. That is an engineering requirement before it is a compliance one, and it is the practical dividing line in AI versus traditional data governance: deterministic systems log transactions, generative systems must also log lineage. Institutions building this capability from scratch can work from a structured AI governance framework. Easypaisa has not published its model governance documentation, so the checklist above states what should be verifiable, not what has been verified.

What ‘ello’ Signals for Digital Banking in Pakistan

The significant shift is not the chatbot itself but the precedent: a licensed Pakistani bank now runs a generative model directly in the customer path, which means supervisory attention moves from contained pilots to live consumer-facing AI.

• Digital lending — eligibility screening and loan servicing conversations.
• Wealth management — guided savings and investment product journeys.
• Fraud prevention — anomaly alerts and in-chat verification prompts.

The driver is arithmetic. Transaction volume at the 4.5 billion level grows support contacts faster than revenue per user grows, therefore automating the informational tier was an economic necessity rather than a marketing exercise, and the February 2026 launch timing follows directly from that pressure. What comes next is the consequential part: the same retrieval-and-routing stack that answers a fee question becomes the natural interface for lending eligibility checks and fraud alerts, consequently the next deployments carry decision consequences rather than informational ones. At that point the State Bank of Pakistan’s supervisory posture — what it requires in model documentation, explainability and complaint handling for AI-mediated decisions — becomes the gating factor on scope, not the vendor’s roadmap. The competitive dynamic reinforces the timeline: because Easypaisa’s current in-app support advantage over JazzCash is a timing advantage, rivals have an incentive to ship equivalents quickly, and a cluster of simultaneous deployments is what typically triggers formal supervisory guidance. Our continuing technology analysis tracks that sequence.

Frequently Asked Questions About the Easypaisa Chatbot

What is ‘ello’ on the Easypaisa app?

‘ello’ is Easypaisa’s AI-powered in-app assistant, launched in February 2026. It answers product, fee, account and transaction questions in natural conversation inside the app and routes users into the correct in-app journey. It exists because Easypaisa operates as a branchless digital bank, which means app-based support carries the volume a branch network would otherwise absorb.

How does Easypaisa’s Shake-to-Chat feature work?

Log into the updated Easypaisa app and shake the phone — the chat window opens over your current screen. The gesture reads the device’s motion sensor and functions only inside an authenticated session. As a result, the assistant inherits your verified identity instead of asking you to confirm it in chat, which removes the verification step that normally opens a support contact.

Who built the Easypaisa chatbot?

Easypaisa built it on GenAI Cockpit, Ant International’s AI-as-a-Service platform for generative and agentic financial applications. Jiang-Ming Yang, Chief Innovation Officer at Ant International, has described the launch as supporting Easypaisa’s AI transformation agenda. Easypaisa bought rather than built because financial-domain-tuned model infrastructure was the scarce input in Pakistan’s market, not engineering capacity.

Is the Easypaisa chatbot available 24/7?

Yes — ‘ello’ operates around the clock inside the app. Consequently, overnight informational queries no longer wait for call-centre opening hours, which removes the morning queue that previously built up between shifts. Cases needing human adjudication, such as fraud reports or reversal decisions, still follow Easypaisa’s human support channels and their operating hours.

Can ‘ello’ resolve a failed Easypaisa transaction?

It can check the status and log the issue, but it does not authorise reversals or refunds. A reversal is a financial decision carrying liability, therefore Easypaisa routes those cases to human agents. For a stuck transfer that the assistant cannot clear, use the escalation option inside the chat, the in-app complaint form, or helpline 3737.

How many people use Easypaisa in 2026?

Easypaisa reports serving 1 in every 5 Pakistanis. The company also reports processing over 4.5 billion transactions worth more than PKR 15 trillion in 2025 — roughly 13% of Pakistan’s GDP — and states that women make up 31% of its customer base. These figures are company-reported and have not been independently audited.

Does the Easypaisa AI assistant support Urdu?

‘ello’ is designed for natural conversational input, but Easypaisa has not published a full list of supported languages or confirmed Roman Urdu coverage. Test your preferred language inside the app and fall back to the Help Centre if responses degrade. Language coverage is material because it determines whether first-time formal-finance users can self-serve at all.

Is it safe to share account details with a banking chatbot?

Never share your PIN, password or OTP with any chatbot, agent or caller. ‘ello’ runs inside an authenticated session, therefore it already has the context needed to answer account questions and does not require credentials. Any request for an OTP or PIN in a chat should be treated as a fraud attempt and reported through Easypaisa’s official channels.

How is an AI chatbot different from a traditional IVR helpline?

An IVR routes you through fixed numeric menus; a generative assistant interprets a free-text question and answers or routes directly. The consequence is fewer steps to resolution for informational queries, because the user no longer has to guess the system’s categories. The trade-off: IVR output is deterministic, while a model requires guardrails, monitoring and audit logging.

What does ‘ello’ mean for the future of digital banking in Pakistan?

It establishes the precedent of a licensed bank running a generative model in the live customer path. Easypaisa has signalled expansion into digital lending, wealth management and fraud prevention, which means the next wave of deployments will carry decision consequences rather than informational ones — and therefore attracts closer State Bank of Pakistan supervisory attention.

Limitations of This Analysis and What Easypaisa Has Not Disclosed

This article analyses a deployment whose performance data is not public. Easypaisa has not published deflection rates, before-and-after resolution times or cost per contact, which means no claim about measured support improvement can be verified from the record; the efficiency case here rests on the channel’s design, not on reported outcomes. All scale figures are company-reported and have not been independently audited. Model governance documentation, data residency arrangements, conversation retention periods and supported-language coverage are also not public.

• Before-and-after customer support wait times and deflection rates
• Cost per contact for the chatbot versus the human helpline
• Full supported-language list, including Roman Urdu handling
• Data residency and conversation retention periods
• Model version, guardrail policy and audit log retention
• Independent audit of the 2025 transaction volume and value figures

For disputes, fraud and reversals, helpline 3737, the in-app Help Centre and formal complaint registration remain the documented escalation routes. The JazzCash rows in the comparison table reflect publicly documented behaviour as of October 2026 and can change without notice.

Conclusion

The Easypaisa chatbot matters because it moved the cheapest support channel into the most discoverable position in the app, not because the underlying model is technically novel. The causal chain is tight: PKR 15 trillion in 2025 volume across a base covering 1 in 5 Pakistanis generated informational support demand that agent headcount cannot track linearly, therefore Easypaisa licensed Ant International’s GenAI Cockpit to compress time-to-deploy, and shipped ‘ello’ with Shake-to-Chat activation in February 2026. That speed came at the cost of dependency on a vendor’s model lineage, which is now the governance question the bank owns.

Two things are worth watching, both with a clear basis. First, whether Easypaisa publishes deflection, resolution and governance metrics — absent those, the performance case remains a design argument. Second, whether JazzCash ships a comparable named assistant, because that would convert Easypaisa’s timing lead into a baseline expectation. Follow our AI coverage and technology analysis for the next milestones. Read more.

References

1. ProPakistani — “easypaisa Launches In-App Chatbot ‘ello’ to Redefine Customer Experience for Millions,” 2026 reporting on the February 2026 launch, the Shake-to-Chat activation and Farhan Hassan’s statement. Link
2. Aaj News — 2026 coverage of the ‘ello’ launch, including Easypaisa’s company-reported 2025 transaction volume, value and customer-base figures. Link
3. ContactCenterWorld — industry report on the Easypaisa and Ant International GenAI Cockpit deployment and the assistant’s customer-support scope, 2026. Link
4. National Institute of Standards and Technology — AI Risk Management Framework (govern, map, measure, manage functions), used here as the governance reference standard. Link