Topic hub
Applied AI and agentic AI in banking
Agentic AI describes AI systems that plan and act across payments, treasury and back-office workflows under delegated authority. It matters to banks now because adoption is broad, while rules on accountability, transparency and model risk are still settling. This hub gives verified regulatory positions, reading paths by role and direct answers on governance and operations.
At a glance
The facts that matter
- In a 2024 Bank of England and FCA survey, 75% of UK financial firms were already using AI and a further 10% planned to within three years. Source (checked 2026-11-10)
- The EU AI Act entered into force on 1 August 2024; Annex III high-risk rules now apply from 2 December 2027 after the AI Omnibus. Source (checked 2026-11-10)
- The European Commission lists credit scoring that can deny people a loan as an example of a high-risk AI use. Source (checked 2026-11-10)
- The FCA does not plan extra regulations for AI and relies on existing frameworks such as the Consumer Duty and SM&CR. Source (checked 2026-11-10)
- US model risk guidance SR 26-2 (17 April 2026) states that generative and agentic AI models are not within its scope. Source (checked 2026-11-10)
- The FSB consulted on 12 sound practices for responsible AI adoption on 10 June 2026, including questions on agentic AI. Source (checked 2026-11-10)
Start here
Three ways in
Boards and executives
Engineers and architects
- AI as the Operating System of Payments: Fraud, Routing, Resilience, and Compliance in 2026
- Agentic Payments in Banking: Consent, Liability, and the New Payment UX in 2026
- From Fragmentation to Fault Line: Corporate Banking's Missing API Standard in the Agentic Era
- Agentic Treasury 2026: From Autonomous Treasury Index to Production-Grade Co-Pilots
Risk, compliance and regulators
- DORA, the EU AI Act, and Data Sovereignty: The 2026 Compliance Stack for Banks
- Your High-Risk Deadline Moved to December 2027. Your Chatbot's Did Not.
- Out of Scope Is Not Out of Risk: The 2026 Model-Risk Rewrite Left Out the Models Banks Are Actually Deploying
- Why the FCA Is Regulating AI Without an AI Rulebook
Latest thinking
Latest articles
- Why the FCA Is Regulating AI Without an AI Rulebook
- The Mystery Seeds Behind NIST's Curves Were Never Hashes, and Claude Found Out in Under Half an Hour
- Agentic AI breaks the assumption behind every task-exposure forecast
- Modernisation Never Finishes, and the Exceptions Are What Pile Up
- An AI's attack on progression-free sets: a ledger of what stalled
- Physical AI and Workforce Governance: Trust, Liability, and the Redesign of Frontline Operations
All 91 articles
- Open-Source Physical AI: How Hugging Face LeRobot and End-to-End Imitation Learning Are Democratising Robotics
- Moving Past Moore's Law: In-Memory Compute Architecture and Modular Chiplets for Edge Robotics
- Humanoid vs Functional Robotics in Manufacturing Automation: Why Form Follows Function on the Factory Floor
- What a Machine Learning Textbook's Contents Page Says About the Field
- An internal OpenAI model pins down the exact decay rate of the Cohn-Elkies sphere-packing program
- A Detector That Reads Probabilities Instead of Writing Verdicts: What 44 Benchmarks Say About Jev
- When the Technology Lands Before the Institution Can Adapt: What MIT's AI Committee Says About Teaching and Learning
- Your High-Risk Deadline Moved to December 2027. Your Chatbot's Did Not.
- Out of Scope Is Not Out of Risk: The 2026 Model-Risk Rewrite Left Out the Models Banks Are Actually Deploying
- Certified Blockchains: The Next Era of Banking Trust
- 2026 Payments: Operating Model, Risk and Revenue
- From Fragmentation to Fault Line: Corporate Banking's Missing API Standard in the Agentic Era
- Reading the Emerging-Technology Risk Horizon for Banks in 2026
- The Agentic AI Index for Banks in 2026: Measuring Autonomy
- Agentic Treasury 2026: From Autonomous Treasury Index to Production-Grade Co-Pilots
- From Pain.001 to Programmable Liquidity: ISO 20022 as the Autonomic Nervous System of Treasury in 2026
- From Bank Statements to Unified Transaction Intelligence: Building an Open-Source Parser for Treasury Teams
- CloudCDN: An Open-Source Blueprint for the AI-Native Edge in 2026
- The Autonomous Treasury Index in 2026: Agentic Treasury, Programmable Liquidity, Tokenised Deposits, and Real-Time Cash Control
- The Wholesale Payments Index in 2026: ISO 20022, Tokenised Deposits, Real-Time Rails, and Cross-Border Settlement
- The Cloud Native Banking Index in 2026: DORA, Platform Engineering, Sovereign Cloud, and Operational Resilience
- The Quantum-Safe Banking Index in 2026: Post-Quantum Cryptography, QKD, Crypto-Agility, and Harvest-Now-Decrypt-Later Risk
- The Agentic AI Index for Banks in 2026: Measuring Autonomy, Governance, Auditability, and Business Impact
- The 2026 Banking Infrastructure Index: Measuring Readiness for Agentic AI, Quantum-Safe Security, Cloud Native Resilience, and Wholesale Payments
- The Multi-Rail Bank in 2026: Cards, A2A, Stablecoins, RTP, FedNow, and Open Banking in One Strategy
- Post-Quantum Payments Infrastructure: Why Banks May Replace Rather Than Retrofit Legacy Rails
- Wholesale Digital Markets in the UK: Tokenised Gilts, Settlement, and the New Champion Role
- ISO 20022 After Migration: Turning Payment Data into Banking Products in 2026
- DORA, the EU AI Act, and Data Sovereignty: The 2026 Compliance Stack for Banks
- Digital Assets in 2026: The Infrastructure Transition Banks Cannot Delay
- AI as the Operating System of Payments: Fraud, Routing, Resilience, and Compliance in 2026
- Stablecoins vs Tokenised Deposits in 2026: What Banks Actually Need to Defend
- Magnifica Humanitas: An AI-Industry Reading of the First Papal Encyclical on Artificial Intelligence
- Programmable Liquidity in 2026: AI, Tokenised Deposits, and Real-Time Treasury Orchestration
- The UK Payments Forward Plan in 2026: Stablecoins, Open Banking, Tokenised Deposits, and Agentic Payments
- Agentic Payments in Banking: Consent, Liability, and the New Payment UX in 2026
- The Renewal of Acid Jazz in the UK: 2026 Artists, Concerts, Albums, and the Return of Groove-Led Live Culture
- Tokenised Deposits in 2026: Banking Services, Stablecoin Competition, and the Status of Programmable Commercial Bank Money
- Cloud Native Banking in 2026: Kubernetes, DORA, Sovereignty, and the End of the VM vs Container Divide
- Global Wholesale Payments in 2026: ISO 20022, RTGS Renewal, and the Economics of Interoperability
- The Quantum Cryptography Reset in 2026: PQC Standards, QKD Assurance, and the Migration Work Banks Cannot Defer
- Agentic Engineering for Banks: A 2026 Blueprint for the C-Suite and the Engineers Who Will Build It
- The Best Cloud Infrastructure Architecture in 2026: An AI-Native, Multi-Cloud, Quantum-Aware Blueprint for Financial Services
- Stablecoin Yield by Another Name: BlackRock's BRSRV and BSTBL Filings Decoded
- Securing the Ledger: A Board-Level Guide to Post-Quantum Migration for Corporate Finance
- The November 2026 pacs.008 Structured-Address Deadline: A Six-Month View
- Lucy's Flash Drive Revisited: AI, Quantum and Knowledge
- Quantum Thresholds Are Moving: 10,000-Qubit Shor Risk
- Quantum-Safe Payments: Why the Industry Must Act Now
- Quantum Lattice Crypto: Bug in Chen's LWE Attack
- Quantum Algorithm Challenges Lattice-Based Cryptography
- OpenVoice: Leading Innovation in Voice Cloning Technology
- Fully Homomorphic Encryption (FHE) in a Banking Quantum Era
- Advancing AI with Multimodal LLMs: Insights from MM1
- Fast Real-Time Speech Recognition on macOS: OpenAI Whisper
- RustLogs (RLG): Structured Logging Library for Rust
- Le Chat by Mistral AI: Multilingual Conversational AI
- Google Gemma AI: Transforming Open-Source AI Development
- Unlocking Gemini 1.5: Google's AI Revolution Explained
- EU's AI Act: Pioneering Ethical AI Regulation Worldwide
- Àkàndé: GPT-Powered Voice Assistant for Executives
- Generative AI in Advertising: Amazon Rufus and Meta
- Audio Analyser: Azure Speech, NLP, and Translation Pipeline
- AI Prompt Engineering 2024: Techniques That Work
- Alien Studio: My Tech-to Art Journey in Photography
- Qiskit and Quantum Fourier Transform for Credit Ratio Analysis
- AI Trends 2024: Insights and Predictions for the Future
- Revolutionising Finance with AI-Enhanced Quantum Algorithms
- State of AI and Quantum Computing in Banking: A 2023 Review
- Quantum Key Distribution Revolutionising Security in Banking
- Efficient Date and Time Management with DateTime (DTT)
- KyberLib: Rust CRYSTALS-Kyber for Post-Quantum
- CRYSTALS-Kyber: The Safeguarding Algorithm in a Quantum Age
- Generative AI in 2023: How It Works, Where It Lands
- Mathematical and Cryptographic Constants for Rust Security
- LibMake: Rust Library Scaffold Generator
- Protecting Data in the Quantum Age: The Hash Library (HSH)
- Static Site Generator: Fastest Rust-Based SSG
- Automating ISO 20022 Payment Files Creation with pain001
- The making of the Express Transaction Credits Platform
- Unveiling a new Cryptocurrency and Faster Payment Solution
- ERC-20: The Ethereum Token Interface That Changed the World
- Understanding the Technology behind Blockchain
- Blockchain explained, the technology that matters the most
- Bitcoin: A Year In Review of the First Cryptocurrency
FAQ
Questions and answers
What is agentic AI in banking?
Agentic AI means AI agents that plan and carry out multi-step tasks, such as payments, treasury or back-office work, under delegated authority rather than only answering questions. The banking questions it raises are consent, liability, audit trails and interoperability.
Where is AI used in payments operations?
In fraud prevention, payment repair, routing, liquidity forecasting and compliance screening, and increasingly in controls for agent-initiated payments. The deciding factor is design discipline: which data, rails, controls and liabilities sit together.
How widely do UK financial firms use AI?
A 2024 Bank of England and FCA survey found 75% of firms already using AI, with a further 10% planning to within three years. Foundation models accounted for 17% of use cases.
When do EU AI Act high-risk rules apply to banks?
After the AI Omnibus, which entered into force on 27 July 2026, high-risk rules for Annex III systems apply from 2 December 2027. Credit scoring is listed as a high-risk use.
Did the AI Omnibus delay everything?
No. The AI Act's transparency rules took effect in August 2026, so customer-facing chatbots and generated content carry obligations now, even though the high-risk deadline moved.
Does the FCA plan AI-specific rules?
No. The FCA says it does not plan extra regulations for AI and relies on existing frameworks, including the Consumer Duty and the Senior Managers and Certification Regime.
Does US model risk guidance cover generative and agentic AI?
Not directly. SR 26-2, issued on 17 April 2026, says generative and agentic AI models are not within its scope and that a bank's wider risk management and governance practices should guide their controls.
What is the difference between human-in-the-loop and human-on-the-loop?
Human-in-the-loop means the agent cannot act without explicit human approval. Human-on-the-loop means the agent acts autonomously while a human monitors and can intervene; every high-risk agent needs a stated position on which applies.
Who is liable when an AI agent makes a payment?
That question is not yet settled, because payment law was written for transactions decided by humans. Emerging protocols try to capture the user's instructions and final approval as auditable evidence of intent before money moves.
Are there international principles for AI in finance?
The Financial Stability Board consulted in June 2026 on 12 sound practices for responsible AI adoption, covering governance and the AI lifecycle. NIST's AI Risk Management Framework, released in January 2023, is another widely used reference.