Insights

AI Strategy

Most AI Business Cases Measure Potential, Not Value

Part 2

Most AI business cases begin with a large number. They take the cost of a function or the hours spent on a task, apply an assumed percentage improvement and call the result value. The arithmetic can be correct while the conclusion is wrong. That number measures an opportunity AI might influence, not how much can be captured, how long capture takes or what it costs to achieve. Exposure, uplift and realization must replace hours saved and addressable value.

Payments

Payment Providers Are Competing to Become the Merchant's Operating System

Part 1

Merchant acquiring was built around a narrow promise. Accept the payment reliably, settle the funds accurately and charge a competitive price. That promise remains essential but it is no longer sufficient. Payment providers are moving deeper into the merchant's operation through smart terminals, commerce software, wallets, working capital and loyalty, and AI is turning those services into a connected decision system. The contest is shifting from who processes the transaction to who helps the merchant run and grow the business.

Banking

The Banks That Win Will Not Take More Credit Risk. They Will Understand It Better

Part 2

Banking interfaces are becoming easy to imitate. A competitor can copy a digital application journey, strip out paperwork and present a prequalified offer in seconds. The experience around a credit decision is far easier to reproduce than the decision itself. AI is reshaping that decision, but the opportunity is not to relax standards or approve more applicants because a model is more complex. It is to separate customers who present genuine risk from customers who only appear risky because the bank does not understand them well enough.

Government

The Border Cannot Inspect Everything

Part 1

Border agencies operate under a constraint they can never remove. The movement of people, goods, money and information is always larger than the capacity available to examine it. A border cannot inspect everything, and trying to would not make a nation safer. AI is not a way to inspect more. It is a way to decide better where scrutiny belongs, what form it should take and when it must happen. The shift is from universal inspection to intelligent, proportionate decisions.

Government

The Cargo Manifest Is Not the Shipment

Part 2

Every cargo decision begins with a representation of reality. A manifest describes the goods, a declaration assigns their value and classification, an image shows what a scanner can see. None of these records is the shipment itself. Each is a partial claim, produced by a different participant for a different purpose. AI is reshaping customs control by comparing those claims with trade patterns, entity relationships, physical observation and the outcomes of past interventions, turning cargo targeting from document screening into a live assessment of what is actually moving.

Healthcare

The Claim Is Not Suspicious in Isolation - The Provider Pattern Makes It So

Part 2

A healthcare claim is built to be adjudicated as a transaction. It identifies a member, a provider, a service, a diagnosis and an amount, and on its face nothing necessarily explains why a payer should hesitate. But many payment integrity problems are invisible on a single claim. They emerge in repetition, concentration and comparison across a provider's history. AI is reshaping this by connecting retrospective provider intelligence with the near-real-time decision about the next claim, before the money moves.

Government

The Cost of a False Identity Is Paid Everywhere Else

Part 3

A national identity credential may be issued by one authority, but its consequences extend across an entire economy. Banks, tax agencies, telecoms, employers and hospitals all rely on it to decide whether the person in front of them is who they claim to be. When an identity is established incorrectly, the error becomes portable and every institution inherits it. AI is reshaping this foundational decision, but the opportunity is larger than better document checks. It is to manage identity as a continuously assured public asset.

Banking

The First Missed Payment Is Already Too Late

Part 1

Banks have traditionally treated financial difficulty as something customers declare or something accounts reveal through arrears. By then the decisive change has often already happened. Income has weakened, essential costs have risen, savings are gone and short-term borrowing has become structural. AI is moving the bank's response upstream, from managing contractual failure to recognizing deterioration early enough to change the customer's trajectory.

Payments

The Lesson from China Is Not the Digital Wallet. It Is the Merchant Ecosystem.

Part 2

China is often treated as proof that the future of payments belongs to the digital wallet. That reading is too shallow. Alipay and WeChat Pay became powerful because payment sat inside broader systems of commerce, communication, customer discovery and financial services. The wallet reduced friction at checkout. The ecosystem made the relationship valuable before and after it. For payment providers elsewhere the opportunity is not to reproduce a Chinese super-app. It is to understand how payment can become the connective tissue of a merchant ecosystem.

Payments

The Smart Terminal Is Becoming the Distribution Channel for Merchant Services

Part 3

The payment terminal used to have one job. It accepted a payment securely and reliably. Smart terminals now sit at the intersection of orders, customers, staff, inventory, settlement and financing, giving payment providers a powerful route into the daily operation of the merchant. Yet the device itself is not the advantage. It appears when the terminal becomes a trusted distribution channel for services that improve a merchant decision and when the outcome of that decision returns to a shared intelligence layer.

Public Sector

The VAT Refund Is Becoming a Real-Time Decision

Part 3

A VAT refund is often treated as the final administrative step after a return is filed. In reality it is one of the most consequential decisions a tax authority makes. Pay too slowly and legitimate businesses finance government uncertainty. Pay too readily and fabricated transactions become public money. Digital transaction evidence and AI are turning that blunt trade-off into a precise decision about what can be released, what must be verified and what warrants investigation.

insights  ·  7 min read

The Global Shift in APP Scam Liability

APP scam regulations are evolving around the world, but the real story is not reimbursement. As liability shifts toward financial institutions, investment priorities are changing. Fraud prevention is becoming an economic imperative, driving new investment in real-time decisioning, mule account detection, behavioural analytics, and Transactional AI.

APAC Banking

AI Governance Built Right Gets Regulatory Sign-Off in Weeks

Part 3

AP regulators have moved from issuing guidance to imposing penalties. Banks that built AI governance as an afterthought are finding that the cost of remediation far exceeds what it would have cost to build it correctly from the start. The institutions that will compound advantage are those that treat governance as the mechanism by which model quality improves — not as the overhead that slows it down.

South African Banking

AI-Enabled Fraud Is Growing Fastest and Receiving the Least Attention

Part 2

Most South African banks are targeting AI at the wrong decisions. The ones that identify the right decisions are failing for three consistent reasons. The SARB/FSCA joint report found governance frameworks uneven across the sector — which means the failure patterns documented globally are also the failure patterns active in South Africa right now.

APAC Banking

APAC Banks Pay $4.59 Per Dollar Lost Without AI Detection

Part 1

APAC financial institutions spend $4.59 for every $1 lost to fraud. Regional AML penalties surged 266% in H1 2024. The 2023 Singapore money laundering case has shifted every major AP regulator from consultation to enforcement. Three pressures are converging on APAC banks simultaneously — and all three are backed by primary evidence.

European Banking

Banks That Built AI Governance First Are Already Ahead

Part 3

The EU AI Act, DORA, and PSD3 together create a compliance architecture that most EMEA banks have not yet built. The institutions that treat that architecture as a design requirement will satisfy regulators faster, deploy to production faster, and compound model quality faster than those that treat it as a retrofit project.

ANZ Banking

Every Major ANZ Compliance Framework Now Requires AI Infrastructure

Part 1

CPS 230 is in force. The Scams Prevention Framework is law. AUSTRAC's AML/CTF reform is enacted. ANZ banks are not facing a regulatory deadline — they are already operating inside one. The question is not whether to act. It is whether the AI infrastructure required to comply already exists.

Swiss Banking

FINMA Will Assess Board AI Governance in Every Review

Part 3

FINMA has stated it will assess board-level AI governance as part of regular supervisory reviews. It has published two guidance documents in five months documenting specific governance gaps across Swiss banking. The institutions that build governance-grade AI before the next examination cycle are making an investment. The ones that do not are accumulating a finding.

ANZ Banking

Five ANZ Frameworks Require the Same AI Governance Architecture

Part 3

APRA has positioned itself as AI-warm but governance-demanding. The board documentation, operational resilience testing, and human oversight requirements of CPS 230 are not future obligations — they are current compliance requirements. Institutions that built governance-first AI before July 2025 are now demonstrably ahead. Those that did not are remediating under examination conditions.

APAC Banking

Four Criteria That Reveal Which AI Decisions to Fix First

Part 2

Most AP banks are targeting AI at the wrong decisions. The ones that identify the right decisions are still failing for three consistent reasons. And the sequence in which to act is not obvious — which is why a structured prioritisation framework matters more in APAC than in any other region.

Swiss Banking

SIC5 Created the Fraud Problem Swiss AI Must Now Solve

Part 1

Switzerland launched SIC5 instant payments in August 2024. Cyber fraud cases rose 40% the same year. CEO fraud nearly doubled. FINMA's April 2026 survey found 42% of Swiss banks have no dedicated digital fraud policy. The infrastructure is live. The governance is not.

South African Banking

South Africa's Digital Fraud Storm Requires an AI Response

Part 1

Overall financial crime losses in South Africa fell 18% in 2024. Digital banking fraud rose 86% in the same year. The divergence is not a contradiction — it is the shape of the threat transition. Traditional fraud vectors are being controlled. AI-enabled fraud is accelerating. The detection systems built for the former cannot defend against the latter.

LATAM Banking

Start With AI Fraud Then Unlock the Thin-File Market

Part 2

Most LATAM banks are targeting AI at the wrong decisions. The ones that identify the right decisions are failing for three consistent reasons. In Brazil, where MED 2.0 enforcement is live and Pix volumes compound the fraud surface monthly, the cost of a failed or misdirected AI programme has never been higher.

North American Banking

Targeting AI at High-Consequence Decisions Delivers the Largest Returns

Part 2

High volume does not mean high value. Most AI roadmaps in financial services are sequenced backwards — and even institutions that identify the right decisions to target are repeating the same three failure patterns that have killed the majority of AI programmes in the industry.

North American Banking

The $61 Billion Compliance Problem That AI Can Solve

Part 1

Compliance cost crisis, enforcement escalation, and fair lending exposure are converging on North American banks at the same time. Each one alone would justify a fundamental redesign of how institutions make decisions. Acting together, they compress the timeline for action in ways that are now measurable in months, not years.

LATAM Banking

The AI Architecture That Stops Pix Fraud Before Settlement

Part 3

MED 2.0 enforcement is active. LGPD requires locally trained models. BCB supervisory expectations are explicit. In LATAM, AI governance is no longer a design preference — it is the mechanism that determines whether a Pix fraud prevention system prevents anything, whether a thin-file credit model can be deployed, and whether the institution compounds its advantage or watches digital-first competitors widen theirs.

North American Banking

The Banks Building AI Governance First Compound the Advantage

Part 3

Most banks treat AI governance as a regulatory obligation to be satisfied after the model is built. The institutions that will compound advantage over the next several years treat it as the architecture that makes deployment fast, model quality self-improving, and regulatory relationships cooperative. The difference is not philosophical. It is financial.

European Banking

The EU AI Act Applies to Models Already in Production

Part 1

Regulatory acceleration driven by the EU AI Act, compounding AML compliance costs, and competitive disruption from digital-first banks are converging on EMEA institutions at the same time. The August 2026 deadline for high-risk AI compliance is twelve weeks away. Most banks are not ready.

South African Banking

What the SARB and FSCA AI Report Means for Boards

Part 3

The SARB and FSCA have published their supervisory direction on AI governance in South Africa's financial sector. The report is not legally binding. It is the clearest signal yet of where binding requirements are heading — and the institutions that build governance-grade AI now will satisfy those requirements before they are enforced, not under examination pressure.

European Banking

Where EU Compliance Requirements and AI Value Creation Converge

Part 2

High volume does not mean high value. Most AI roadmaps in EMEA financial services are sequenced backwards — and even institutions that identify the right decisions to target are repeating the same three failure patterns that have killed the majority of AI programmes across every market.

Swiss Banking

Which AI Decisions Swiss Banks Need to Prioritise First

Part 2

Most Swiss banks are deploying AI but have not built the governance frameworks to make it defensible. FINMA's findings are specific: 42% have no digital fraud policy, model risk is systematically under-addressed, and board-level governance is absent in most institutions. The failure patterns are identical to every other market in this series.

ANZ Banking

Which AI Decisions to Prioritise When Three Regulators Are Watching

Part 2

Most ANZ banks are targeting AI at the wrong decisions. The ones that identify the right decisions are still failing for three consistent reasons. With CPS 230 in force and AUSTRAC reform enacted, the cost of a failed or ungoverned AI programme in ANZ is higher than it has ever been.

LATAM Banking

Why Pix Made AI Fraud Prevention an Infrastructure Requirement

Part 1

Brazil built the world's most advanced instant payments infrastructure. In doing so, it created the world's most advanced bank fraud problem. Pix processed 63.4 billion transactions in 2024 on an irrevocable settlement rail — a fraud surface that no rule-based system can adequately cover, and a financial inclusion opportunity that no traditional credit model can serve.