Most people know of, or regularly use, generative AI such as ChatGPT or Microsoft’s Copilot to do everything from writing report summaries at work to seeking travel recommendations for their next trip.
But an emerging form of artificial intelligence called “agentic AI” takes these capabilities to the next level, with the ability to make decisions and proactively perform autonomous tasks without human guidance.
Agentic AI, which has come to the forefront over the past year, is the “next big thing” and has incredible potential for financial services and businesses at large, said Yannick Lallement, Scotiabank’s Vice President, AI Engineering & Labs.
“Agentic AI systems are capable of taking actions in the real world to pursue and achieve specific goals,” he said, adding that Scotiabank has begun to deploy a form of it in its Commercial Banking division.
Unlike generative AI, which is prompted to provide an output such as content or a photo, agentic AI uses real-time analysis from large data sets to autonomously make complex decisions.
If 2024 was the year of generative AI adoption, 2025 will be the year of agentic AI, according to a Scotiabank Global Banking and Markets report called Focus on 2025 – Technology, published late last year.
“We think agentic AI will be the next big wave in software, translating into smarter and more automated workflows,” said the report’s authors.
Agentic AI leverages a mix of machine learning, natural language processing and automation technologies, according to Harvard Business Review. It doesn’t rely on prompts by a user, but instead is calibrated to achieve a certain goal, and can execute a sequence of activities to achieve that goal.
Agentic AI reduces client wait times
With these added capabilities, this form of AI has the potential to transform many areas of work that previously were too complex for automation.
That’s why it was a good fit for Commercial Banking, where client requests and inquiries can vary widely, said Lawrence Engel, Scotiabank’s Vice President of Business Banking Operations.
On a typical day, this department would receive more than 1,500 emails to a centralized email address, or roughly 60,000 to 70,000 requests per month, he added.
Previously, each emailed client request would be opened individually by a Scotiabank team member, who would read it and determine what was needed, what department to send the request to, and then log it in the system.
The client asks in Commercial Banking are complex and irregular, ranging from withdrawing on a credit line to investigating a transaction or a request from the client’s lawyer on their behalf.
Human-reviewed emails were sent to the correct department the vast majority of the time, but it could take as long as one to two hours.
AIDox developed in house and patented
Meanwhile, Scotiabank had been developing AIDox — which uses artificial intelligence to analyze and compare documents — since 2018.
It had already deployed the AI tool to other departments where requests are more straightforward. For example, AIDox was being used in retail banking to confirm insurance letters for mortgages.
In 2023, it began using AIDox in Commercial Banking and training it — feeding in data and teaching it how to respond.
But this year, agentic AI capabilities were added to AIDox to enable it to process client emails in Commercial Banking on its own, said Lallement.
Not only can AIDox understand the complex requests e-mailed by Commercial Banking clients, but it also forwards the email to the correct team to handle it and creates a case in Scotiabank’s system for processing and fulfillment.
The result is faster turnaround times and greater accuracy, said Engel. Rather than hours, client requests are directed to the correct teams within minutes.
“The benefit here was the speed at which the work was being done. That was the huge win,” said Engel.
Improved client and employee experience
Now, of those 1,500 or so daily emails, AIDox handles roughly 90% of them. More complex cases are flagged to be reviewed by human eyes.
“We have a very good process set up to deal with the anomalies that come in. And then we've got a SWAT team that focuses on the more challenging components,” Engel said. “This also allows those requests to reach the fulfillment queues faster as well.”
For clients, not only is the response time quicker, it’s not limited to business hours as AIDox can be implemented and responsive continuously.
“We have seen client pulse scores consistently go up over the course of the last year,” Engel said, referring to a score used to measure client satisfaction. “It’s hard to pinpoint it on just one thing, but it’s a sign that we’re doing things better in the eyes of the client.”
Agentic AI has been an improvement for the workforce as well, allowing them to dive into more complex cases, he added.
“This allows our employees to focus on engaging, value-added work,” Engel said.
The use of AIDox, which Scotiabank has patented, has now allowed Commercial Banking to redeploy 70% of the team responsible for directing those emails into other areas.
There is much potential to deploy Agentic AI with similar benefits in many other parts of the Bank, said Beth Einwechter, Scotiabank’s Director of Strategy & Operations in Canadian Banking.
“We probably picked one of the hardest use cases first, it was very difficult,” she said. “But, because of that, we have had a lot of success now and we’re well-positioned to add other use cases across the Bank.”
“AI is going to allow Scotiabank to offer better products that make banking easier for clients.”
AI at work in many areas
Scotiabank is leveraging AI to serve clients with greater speed and accuracy in several other domains.
It is also using AIDox to confirm over-the-counter derivative trades, shaving off up to 1.5 hours in response time. A person will still review the confirmations, but AI does the bulk of the work, expediting the process and reducing risk of human error, said Lallement.
“That gives us a competitive advantage because we are a good counterpart to trade with because we confirm faster, while also improving controls,” he said.
The Bank is also exploring the use of artificial intelligence to review documents related to mortgage applications and extract the pertinent information more quickly and accurately than a manual review.
As well, Scotiabank has launched an AI-powered chatbot to help contact centre agents at the Bank’s Client Experience Centre and branch employees. AskAI was developed using Google’s Gemini large language model. Using generative AI, the chatbot can understand and respond to queries without needing to use specific keywords. AskAI, which was launched earlier this year, can answer questions about banking and services quicker than contact centre agents and branch staff can look up the answers themselves. For example, it can answer complex questions such as “How do I move a business account from sole proprietorship to corporation?” in seconds. It would take on average five minutes to find the answer using search tools and scanning documentation. With AskAI, clients are served more efficiently with shorter call lengths and wait times, and feedback from employees has been positive as well.
Ethics and protecting privacy key considerations
While clients may have some concerns about risks related to the use of AI, Scotiabank applies rigour to the use of AI, Lallement said.
“Every single AI model has to go through an ethics review first,” Lallement adds.
He adds that there is a team of senior leaders tasked with examining data and AI risk.
This review involves asking the business line proposing the application of AI to answer a detailed list of questions to determine the risk level. That includes questions about the kind of data used and how the AI model comes to a particular decision. The higher the risk level, the higher the level of approval needed, he added.
With those considerations at the forefront, the potential for AI to improve the financial services sector and other industries is vast.
“I see a similar path to the mobile revolution and the internet revolution before that. AI is going to allow Scotiabank to offer better products that make banking easier for clients.”
