Adyen CEO Pieter van der Does was talking about machine learning for payment-fraud detection—not generative AI or automated lending—when he said he was surprised by how effective AI was. The remark came in a VentureBeat interview published December 5, 2019, around the Slush technology conference in Helsinki. Adyen’s early use case was practical: use transaction signals to catch fraud while avoiding needless declines of legitimate purchases.
What did Adyen’s CEO say about AI?
In a VentureBeat interview published December 5, 2019, Adyen CEO Pieter van der Does recalled that the company had approached AI cautiously. After internal trials of machine-learning algorithms for payment fraud, he said: “When we did our first trials with it, I was surprised how effective it was.” The full conversation was also available as a podcast associated with the interview.
As an Amazon Associate I earn from qualifying purchases.
The context matters: he was describing fraud and transaction-risk decisions within Adyen’s payment business. The interview was not about chatbots, generative AI, autonomous shopping agents, or deciding whether someone qualifies for credit. Nor should the headline be read as a new statement made in 2026.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Why fraud detection is also a sales problem
A payment-risk system has to distinguish harmful transactions from legitimate ones that happen to look unusual. Blocking a stolen-card purchase can prevent losses and chargebacks. Blocking a real customer, however, can cost a merchant the sale and potentially the customer’s future business.
#1 Best Overall
![Arduino® UNO™ Q 4GB [ABX00173]- Hybrid Board, Qualcomm Dragonwing QRB2210 microprocessor (MPU) & STM32U585 Microcontroller(MCU), AI Vision, Voice, IoT, Robotics, Linux Debian OS, Wi-Fi 5, USB-C](https://m.media-amazon.com/images/I/51cl4xygwGL._SL160_.jpg)
- Dual-Brain Hybrid Power: Combines the Qualcomm Dragonwing QRB2210 MPU (Quad-core Arm Cortex-A53 @ 2.0 GHz CPU, Adreno GPU, AI acceleration) and the real-time, low-power STM32U585 MCU for advanced applications like object recognition, voice commands, and motion detection.
- AI & Linux Capabilities: Unlocks AI-powered vision and sound solutions; runs Linux Debian OS for coding in Python and supports the Arduino ecosystem with libraries and Sketches; quick start with Arduino App Lab.
- Advanced Features: Equipped with 4 GB LPDDR4 RAM, 32 GB eMMC built-in storage, ideal for single-board computer (SBC) mode, running multiple simultaneous high-level processes, more complex AI or ML models, extensive logs. Dual-band Wi-Fi 5 (2.4/5 GHz), Bluetooth 5.1, and high-speed headers for vision, audio, and display peripherals.
- Seamless Expansion & Connectivity: Features the classic UNO form factor for shields compatibility, an 8x13 LED matrix, and a Qwiic connector for easy expansion with Modulino nodes; power and connect via the USB-C connector.
- Intended Use & Development: The perfect platform for prototyping robotics or IoT projects, empowering innovators with a unified development experience to mix Arduino Sketches, Python scripts, and containerized AI models in a single interface.
That makes the goal broader than reducing fraud. Merchants need to approve legitimate purchases while controlling fraud losses, chargebacks, manual-review costs, and customer friction. A model that catches more fraud but also rejects many good shoppers may make the overall economics worse.
The interview used an international flower purchase to illustrate the problem: someone in New York might use an Israeli-issued card to send flowers to Europe. A simplistic rule could flag that mismatch. A risk system with more context may be better able to assess whether it is a plausible purchase or a fraudulent one. The example is illustrative, not a guarantee that any model will correctly approve such a transaction.
How Adyen applied machine learning
Rather than treating AI as a standalone product label, Adyen’s early approach was to develop algorithms internally for an existing payments challenge. The company said it could use information from its payment platform to inform risk decisions and refine its models. That is applied machine learning embedded in payment infrastructure—not evidence that AI automatically improves every part of payments.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsThe 2019 account discussed Adyen’s Risk Engine and ShopperDNA. It described using more transaction data points to help assess risk and reduce unnecessary declines. A later Harvard Business School Digital Initiative case study about OLX identified signals such as location, email, average transaction size, card information, shopping-cart contents, and transaction history. Those examples do not establish that every merchant, integration, or market supplies or uses every signal.
Rank #2
- Dual-Brain Hybrid Power: Combines the Qualcomm Dragonwing QRB2210 MPU (Quad-core Arm Cortex-A53 @ 2.0 GHz CPU, Adreno GPU, AI acceleration) and the real-time, low-power STM32U585 MCU for advanced applications like object recognition, voice commands, and motion detection.
- AI & Linux Capabilities: Unlocks AI-powered vision and sound solutions; runs Linux Debian OS for coding in Python and supports the Arduino ecosystem with libraries and Sketches; quick start with Arduino App Lab.
- Advanced Features: Equipped with 2 GB LPDDR4 RAM, 16 GB eMMC built-in storage, ideal to develop in PC-connected mode, running the OS, Python scripts, and basic network services (SSH) without a demanding GUI or heavy multitasking; great for lightweight AI and memory-optimized TinyML applications, needing local storage for basic OS and core libraries. Dual-band Wi-Fi 5 (2.4/5 GHz), Bluetooth 5.1, and high-speed headers for vision, audio, and display peripherals.
- Seamless Expansion & Connectivity: Features the classic UNO form factor for shields compatibility, an 8x13 LED matrix, and a Qwiic connector for easy expansion with Modulino nodes; power and connect via the USB-C connector.
- Intended Use & Development: The perfect platform for prototyping robotics or IoT projects, empowering innovators with a unified development experience to mix Arduino Sketches, Python scripts, and containerized AI models in a single interface.
Signals that can inform a risk decision
- Identity and shopper context: email, account history, and available device or behavior information.
- Payment context: card or bank details, payment method, issuer, and authorization response.
- Commerce context: basket contents, order value, delivery location, and merchant category.
- Behavior and outcomes: transaction velocity, repeated attempts, bot activity, chargebacks, disputes, confirmed fraud, and fulfillment results.
Which data can be collected and used depends on the integration, jurisdiction, merchant relationship, and applicable privacy and data-protection requirements. Merchants should establish what information is transmitted, why it is needed, how it is handled, and what they are permitted to share.
What the reported results do—and do not—show
VentureBeat reported that Adyen’s Risk Engine reduced transaction-review times by 30%. That is a company-attributed figure reported in 2019; it is not a current independently verified benchmark or a performance guarantee for a merchant considering the product today.
The OLX case study reported a 2.6% increase in authorized transactions after eight weeks. That is a merchant-specific case-study result, not a prediction for other businesses. Results depend on such factors as the merchant’s baseline rules, transaction mix, data quality, geography, and definition of an authorized transaction.
How Adyen’s risk products have changed
The 2019 interview referred to RevenueProtect. Adyen’s current documentation presents Protect as its risk-management system, and its transition guidance recommends Protect instead of RevenueProtect. Protect is the current successor terminology and tooling; merchants should not assume the older product and current system are identical. Some documentation and integrations may still contain legacy RevenueProtect references.
Rank #3
- Single core ARM Cortex-A7 32-bit core, integrated with NEON and FPU
- Built in Micro's self-developed 4th generation NPU, with high computational accuracy and support for mixed quantization of int4, int8, and int16. Among them, int8 has a computing power of 0.5 TOPS and int4 has a computing power of up to 1.0 TOPS
- Built in self-developed 3rd generation ISP3.2, supports 4 million pixels, and supports various image enhancement and correction algorithms such as HDR, WDR, and multi-level denoisin
- It has powerful encoding performance, supports intelligent encoding, adapts to save bit rates according to the scene, and saves more than 50% of the bit rate compared to conventional CBR mode, making the captured images high-definition, smaller in size, and doubling the storage space
- The design with built-in RISC-V MCU supports low-power fast startup, 250ms fast capture, and simultaneous loading of AI model library, enabling facial recognition to be completed within 1 second
Protect is part of the broader Adyen Uplift optimization layer, whose modules are Tokenize, Protect, Authenticate, Optimize, and Personalize. Adyen describes Uplift as combining fraud prevention, authentication, checkout and payment-method decisions, and cost optimization. The modern product story is therefore broader than the early fraud-detection use case.
Capabilities and tier differences
Adyen’s documentation describes a configurable risk flow that can allow, block, review, or send a transaction for 3D Secure authentication. The system is not simply a single automatic “fraud or not fraud” classifier. Current Protect features include bot and attack protection, risk profiles, custom rules, analytics, experiments, case management, and dynamic 3D Secure controls. Feature availability differs between Basic and Premium; Adyen identifies machine-learning fraud detection as a Premium capability. See its Protect tier information and description of how Protect works.
Adyen also advertises an average 86% reduction in manual risk rules, savings of up to 5% on total payment cost, a 10% conversion-rate increase in a company-published customer-initiated transaction example, and models trained on “trillions of dollars” in payments data. These are Adyen’s marketing claims, not independently established results for a typical merchant. Outcomes depend on a merchant’s setup, participation, transaction mix, and comparison baseline.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →What merchants need for an effective implementation
Risk decisions are only as useful as the data and operating process behind them. Adyen’s current Uplift requirements, checked August 18, 2026, call for an online payments integration that supports Uplift. Recommended web integrations use Web Drop-in version 6 or Web Components version 6; relevant integrations require Checkout API version 71 or later. Adyen also lists enabled and accepted webhooks and useful shopper and transaction data as requirements or recommendations. These details can change, so implementation teams should verify the live documentation before deployment.
Rank #4
- 【POWERFUL ESP32‑S3 CONTROLLER】Built‑in Xtensa 32‑bit LX7 dual‑core processor, 512KB SRAM, 8MB PSRAM, 16MB Flash for stable AI voice computing and multitask processing.
- 【Preloaded Dual AI Platforms】Comespre-installed with complete Deepseek and OpenAI voice dialogue projects.Experience intelligent voice interaction instantly. (Note: OpenAI functionality requires your own API key.)
- 【STABLE WIRELESS & CLEAR AUDIO】Integrated 2.4GHz Wi‑Fi + Bluetooth 5 (LE); dedicated audio decoding module for natural, responsive voice interaction.
- 【USER‑FRIENDLY VISUAL & PLUG‑AND‑PLAY】2” TFT‑SPI color screen shows real‑time chat; modular design, no extra wiring, ready to use after setup.
- 【FULL LEARNING SUPPORT】45 programmable GPIOs, rich interfaces, online web tutorials, free technical support for beginners & developers.
The documentation highlights shopperEmail as important for shopper recognition and optimizing fraud and 3D Secure decisions. Raw-card-data API integrations can also create additional PCI-compliance obligations. Merchants should account for privacy notices, data-processing roles, retention, cross-border transfers, and any applicable rules for automated decisions; requirements depend on the integration and jurisdictions involved.
How to tell whether payment AI is working
Do not judge a risk product by its “AI” label or by fraud reduction alone. Compare results with the merchant’s actual baseline and assess the combined economics. Useful measures include:
- Authorization rate and incremental approval lift for legitimate transactions.
- False-decline rate, including effects on repeat customers and new shoppers.
- Fraud rate, chargeback rate, net fraud losses, and dispute costs.
- Manual-review rate, review time, and labor cost.
- Conversion rate, customer-support contacts, and cost per approved transaction.
- Performance by country, payment method, issuer, device, and customer segment.
Also check whether confirmed fraud, disputes, refunds, and fulfillment outcomes return to the system in a timely and accurate way. Chargeback labels can arrive late or be incomplete, and poorly labeled outcomes can undermine tuning. Test changes through backtesting or controlled experiments where possible; do not assume a global rule will work equally well across markets.
Recommended Free Tools
Trade-offs and failure modes to consider
- More data versus greater responsibility: richer context may help decisions, but brings privacy, governance, security, and integration obligations.
- Fraud control versus customer friction: aggressive declines or frequent 3D Secure challenges can reduce fraud while also rejecting good orders or increasing checkout abandonment.
- Adaptability versus explainability: machine learning can respond to patterns that explicit rules miss, but risk teams still need usable reasons, controls, and auditability.
- Historical patterns versus new behavior: models can overfit to past customers and penalize new shoppers, gifts, travel purchases, or legitimate cross-border orders.
- Broad payment data versus concentration: an integrated platform may simplify operations, but merchants should ask how their data is used and consider vendor dependence and switching costs.
Other common blind spots include focusing only on stolen cards while missing account takeover, card testing, bots, refund abuse, marketplace abuse, or friendly fraud; sending incomplete checkout data; and optimizing for a weak baseline. A specialist fraud provider may make sense if a processor’s native controls do not cover a merchant’s specific risks, but adding a vendor also adds integration and operational complexity.
What to compare before choosing a provider
Adyen is one option, not a universal answer. A merchant comparing payment-risk tools should evaluate net approved revenue, fraud and dispute costs, risk fees, engineering effort, geographic and payment-method coverage, reporting, case management, and contract terms. Adyen’s public pricing page states that payment processing has a fixed processing fee plus a payment-method fee, with no setup or monthly fee; pricing varies by method, country, contract, and product, and risk or premium features may cost extra. See Adyen’s pricing page for current terms.
Depending on the merchant’s existing stack and needs, comparisons might include Stripe Payments with Stripe Radar, enterprise-focused Checkout.com, or a specialist such as Forter alongside an existing processor. Their suitability depends on geography, payment methods, integration needs, and negotiated terms; this comparison does not establish that one provider is cheaper or better. A merchant should not switch processors solely because a vendor markets its system as AI-powered.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
