Radar
Short, dated, sourced. What moved.
Two Deadlines and a Convergence: The Week Agentic AI Became Regulated Territory
OSFI E-23 (166 days), Canada's Real-Time Rail goes live, and three regulators converge on agentic AI governance.
ISO 20022 Hits November, BIAN Aligns, Canada Picks Its Standards
Swift ISO 20022 structured-address cutover looms, Canada's CDB framework moves to implementation, FDX launches agentic AI initiative, BIAN v14.0 aligns with ISO 20022.
The Intent Layer Gets a Specification
Canada's Real-Time Rail launches without Big Six banks, the EBA defines terms for agent-ready banking, and the intent layer gets its first formal specification.
What the regulatory radar watches
The scope and strategy behind a weekly Monday scan for regulatory deadlines, enforcement actions, and consultation windows in banking.
What the reference models radar watches
The scope and strategy behind a weekly Tuesday scan for banking reference models, data standards, and interoperability frameworks — who defines the rails the ecosystem runs on.
What the fintech radar watches
The scope and strategy behind a weekly Monday scan for fintech market signals — neobank launches, funding rounds, API economy, and innovation moves that reshape banking competition.
What the architecture radar watches
The scope and strategy behind a weekly Tuesday scan for software architecture, platform engineering, and system design signals at scale — where complexity accumulates and what breaks.
Agentic AI Reaches Production Banking as Stablecoin Infrastructure Goes Bank-Grade
Arva AI deployed at top-10 U.S. bank, Deutsche Bank launches crypto custody, Canada's RTR goes live, and open banking accreditation comments are due December 8.
Agent Infrastructure Matures, Governance Becomes the Work
KubeCon 2026 puts agent runtimes on Kubernetes. CIBC goes cloud-native core. The gap is no longer infrastructure.
From Building Agents to Building the Architecture That Makes Them Coherent
The market shifts from framework proliferation to operational specification — intent primitives, coordination fabrics, and minimum viable architectures.
Canada's Real-Time Rail Launches — and Wealthsimple, Not the Big Six, Is the Largest Participant
Canada's 24/7 instant payment rail goes live with no Big Six bank in the launch cohort. The EBA defines the Agent-Ready Bank. Wealthsimple stacks Penny, CADD, and the rail.
From Naming to Specifying: The Intent Layer Gets Architectural Teeth
Capgemini, Inteq, Gardner, and HKMA turn intent engineering from concept into enforceable specification and regulation.
Wealthsimple Launches Penny and a CAD Stablecoin — the Agent Integration Layer Goes Live
Wealthsimple ships Penny AI assistant, CADD stablecoin, and agentic payment cards. BMO deploys internal agent assist. Bank of Canada opens CDB accreditation.
Intent Engineering Crystallizes as the Layer Above Harness
Five independent sources formalize intent engineering as the discipline that prevents agents optimizing the wrong thing at scale.
Four of Six Canadian Banks Block AI Agents. National Bank Is the Only One Still Watching.
RBC, TD, BMO, CIBC block third-party AI agents; National Bank monitors. Ripple lands a US trust charter; Wealthsimple raises $425M at $13.8B.
Governance Planes Ship Faster Than Value Streams Can Deploy Them
Nine vendors shipped agent governance planes in six weeks. The reasoning gap is now the structural problem.
From Block to Build: OnePay Ships the First Consumer Agentic Banking Product
OnePay ships the first consumer agentic banking product, Lead Bank proposes an Authorized Financial Agent framework, and the block side hardens — scoped authorization becomes the production pattern.
Fail-Plausible: When Agents Narrate Their Own Failures as Success
BCG finds only 5% of companies have critical agent controls while a new failure class feeds humans counterfeit confidence.
JPMorgan Blocks Agents As The Market Builds The Bridge
Dimon draws the line on outside AI agents; two days later Self Labs ships the credential bypass and Crossmint commoditizes agent payments.
Specification Engineering: The Fourth Layer Above Intent
The agentic stack gets its first formal academic treatment with a new layer preventing fleet-wide agent fragmentation.
Visa Agent Payments Go Live As Stablecoin Settlement Scales
Five Canadian banks join Visa Agentic Ready, Bain pins the AI-native bank on a deterministic-probabilistic split, and stablecoin settlement reaches production scale.
Intent Becomes an Architectural Primitive; PocketOS Proves Prompts Are Not Security
The intent layer graduates from concept to formal spec while a Cursor agent deletes a production database in 9 seconds.
Open Banking Rules Land As Bank Stablecoins Go To Production
Canada operationalizes Consumer-Driven Banking, McKinsey maps agent disintermediation, and Fiserv ships the first production bank stablecoin on a public chain.
Runtime Governance Ships; the Knowledge Layer Crystallizes
Collibra, Neo4j, and Gartner converge on a shared governed substrate as governance moves from documentation to enforcement.
OSFI Governs Agentic AI As Charter Consolidation Accelerates
Canada makes agentic AI a regulated activity while Revolut and Questrade pair charters with proprietary foundation models.
What the enterprise harness radar watches
The scope and strategy behind a weekly scan for signals in agent harness engineering, intent governance, and the architecture that makes autonomous agents coherent at enterprise scale.
The AI Control Plane Eats the Model
Harness engineering, intent gaps, and value stream governance emerge as the real moat — not model capability.
Visa, Mastercard, Ant Align on Agent Identity as Chime Buys a Charter
Cross-network KYA interoperability arrives. Chime pays $590M for Stride Bank. The charter is the moat; the license is the defensible layer.
OpenAI Comes for Junior Bankers as Nubank Lands in the US
ChatGPT for Financial Services automates analyst work. Nubank brings 140M customers and stablecoin yields stateside. Two fronts, one week.
Bain Draws the Blueprint: Three Layers for Agentic Banking
Bain publishes the canonical agentic AI architecture. Gartner says 40% of projects will fail — infrastructure, not models, is why.
OSFI Draws the Line: Canada's Agentic AI Rules Are Now Concrete
OSFI bulletin plus Guideline E-23 give Canada the first explicit AI agent regulatory framework. Compliance clock is ticking.
ChatGPT Meets Plaid: The Battle for Banking's Intelligence Layer
ChatGPT Pro now reads 12K+ banks via Plaid. Banks still hold the moat — but for how long?
What the agentic banking radar watches
The scope and strategy behind a weekly scan for signals where AI agents start acting in banking — competitive moves, regulatory shifts, and the architecture that makes them coherent.