Artificial Intelligence at Royal Bank of Canada

Royal Bank of Canada (RBC), headquartered in Toronto, Ontario, is Canada’s largest financial institution and one of the world’s largest by market capitalization. Operating throughout 29 international locations and using upwards of 98,000 staff as of 2024, the corporate reported an annual whole income of $57 billion and a document internet earnings of $16.2 billion, pushed partially by the acquisition of HSBC Bank Canada and the sturdy efficiency of acquisitions within the UK and EU.

RBC’s dedication to inner AI deployment is evidenced by RBC Borealis, a analysis institute inside RBC that serves because the financial institution’s default AI center of excellence.

The institute describes itself as performing basic and utilized analysis in machine studying with the objective of advancing the state-of-the-art in AI for monetary providers and past. Launched in 2016 and now boasting 950+ staff, the institute’s progress and in depth analysis output affirm the extent to which RBC regards AI inside its enterprise technique.

Further proof of RBC’s dedication to AI is obvious in its vital capital allocation — a 2025 “Investor Day” presentation notes the financial institution is at the moment investing over $5 billion in know-how to speed up innovation. This funding is tied to a daring, measurable monetary final result: RBC’s strategic AI ambition targets producing a considerable $700 million to $1 billion in enterprise worth from AI by 2027.

The financial institution’s specific focus demonstrates that AI is extremely built-in into RBC’s company framework, the place basic analysis should yield a transparent, anticipated pathway to scale, monetization, and lowered danger.

This article examines two mature, internally deployed functions that illustrate AI’s central position in Royal Bank of Canada’s core operations:

• Leveraging machine studying to adapt to evolving danger: Implementing federated machine studying strategies for higher sample matching and the elimination of delicate knowledge alternate.
• Applying deep studying for optimised pricing methods: Utilising deep reinforcement studying to supply higher management to merchants whereas minimising slippage towards business benchmarks.

Leveraging AI and ML to Adapt to Evolving Risk

The monetary sector faces huge challenges as AI turns into extra widespread. A latest study by the Ontario Securities Commission (OSC) discovered that AI-enhanced scams pose considerably higher danger, with managed checks exhibiting contributors invested 22% extra in AI-enhanced scams than in typical scams. 

For context, knowledge from the Canadian Anti-Fraud Centre (CAFC) reveals that Canadians misplaced $638 million to fraud in 2024 alone. That quantity is anticipated to develop as AI know-how turns into extra broadly adopted, resulting in extra refined approaches. A blog post printed by RBC warns that social engineering assaults and ever-more complicated AI-powered scams have gotten much more plausible and profitable. 

To deal with these challenges, RBC launched a fraud modernization initiative that’s transitioning its protection methods from static guidelines to adaptive, real-time danger scoring engines powered by superior AI and Machine Learning (ML). 

According to an organization press release, the foundational infrastructure was custom-built, enabling complicated occasion processing and embedding habits analytics and fraud prediction capabilities. The system leverages the financial institution’s large knowledge stream, analyzing roughly 11 trillion safety occasions in 2024 alone, as reported within the Investor Day presentation.  

A notable instance of RBC’s strategy to tackling fraud is a joint experiment with Vector Institute to handle a brand new and rising risk to monetary providers: mule account fraud. 

Mule account fraud happens when criminals rip-off victims into permitting them to make use of their financial institution accounts, or when criminals open accounts utilizing stolen identification paperwork. Because they usually function throughout a number of establishments to obfuscate their actions, the approach has confirmed notably tough to counter. 

Fraud detection depends closely on analyzing delicate info, similar to transaction knowledge, system exercise, and consumer habits. The addition of AI will increase the danger of violating privateness legal guidelines, as AI methods can gather extreme knowledge, deal with it insecurely, and even misuse it. 

As reported by Borealis AI, the joint strategy to handle mule account fraud leveraged superior machine studying and federated studying. The federated approach permits a number of establishments to coach fashions collaboratively with out exchanging delicate buyer knowledge, whereas nonetheless benefiting from patterns noticed throughout a number of banks. 

According to the report, the taking part banks every prepare an area mannequin on their very own knowledge, with solely the discovered mannequin parameters being aggregated into a worldwide mannequin shared by all. The result’s the detection of fraud patterns which may be invisible to any single establishment. 

The outcomes confirmed “notable enhancements in efficiency metrics for all purchasers” and that shared studying throughout purchasers helped generalize higher to fraud patterns. 

Applying Deep Learning for Optimised Pricing Strategies

According to a World Economic Forum white paper, monetary providers corporations spent $35 billion on AI in 2023, with projected investments throughout banking, insurance coverage, capital markets, and funds companies anticipated to succeed in $97 billion by 2027.

An Ernst & Young analysis reveals the explanation for such strong funding: AI fashions persistently outperform most conventional fashions, yielding higher financial features.

The potential of fashions to course of each structured and unstructured datasets and analyze huge volumes from many numerous sources gives proof of features too quite a few for the sector to disregard.

The monetary providers sector can also be seeing tech corporations getting into their territory — the partnership between Apple and Goldman Sachs being one such instance.

EY reporting elsewhere notes that, in response to the business’s technological upheaval, banks are strategically reallocating their IT budgets towards improvements that may successfully counter threats from tech giants and rising enterprise fashions.

With the business primed for change, RBC and its RBC Borealis AI analysis institute launched Aiden in 2020, an AI-powered digital buying and selling platform that goals to supply higher insights and improved execution for the enterprise and its purchasers worldwide. 

According to the corporate’s landing page for Aiden, the platform’s volume-weighted common value (VWAP) technique seeks to minimise slippage relative to the market VWAP benchmark, even in periods of volatility. 

Using deep reinforcement learning to be taught from previous actions and execute buying and selling choices primarily based on dwell market knowledge, Aiden VWAP dynamically adapts to new info in actual time. Rather than following a set of guidelines, the Aiden algorithm learns repeatedly by way of trial and error and a strong reward system, finally discovering higher success with fewer inputs. 

Building on the success of the Aiden VWAP, RBC Capital Markets explains that Aiden Arrival seeks to handle the arrival value problem — the persistent problem of precisely calculating the arrival value of a monetary commerce — by way of a extra holistic benchmark analysis. Aidan Arrival additionally reveals additional optimisation, offering an expanded buying and selling motion set that gives higher flexibility and management for merchants.