BNY has over 100 “digital employees” in the form of AI agents that have dramatically streamlined key functions including customer onboarding, a tech executive from the nation’s oldest bank said at this week’s Microsoft AI Tour in New York.
At the same time, another major Microsoft AI customer, Boston Scientific, laid out the principles and practices it uses to detect, protect, and govern sensitive data in the highly regulated medical products industry.
In total, the day’s sessions, including those with BNY and Boston Scientific Group, “Went deep on enterprise [considerations ] — real architecture, real implementation challenges,” said Paul Swider, CEO and Chief AI Officer of the healthcare industry consultancy RealActivity, who was in attendance, in a LinkedIn post, adding that real-world AI insights are “exactly what we need right now.”
Agents Proliferate — Under Human Supervision
The 240-year-old BNY handles 20% of all investable assets across the world. It transfers or makes payments of about $2.5 trillion every day. That level of business and operational scale dictates that BNY prioritize resiliency and innovation, said Yagnyesh Hariharan, Senior Vice President, Strategic Partnership Solutions for BNY.
When the company first began using AI roughly five years ago, it adopted the approach of democratizing AI to make it broadly available across its ecosystem, with the goal of delivering better experiences for employees, partners, and customers.
It promotes use of AI by employees for functions ranging from semantic searches for research into companies they’re interacting with, to agents that perform tasks such as repairing payments with issues, to customer service. “These are things that are provided to all of our employees,” Hariharan said.
In 2025, the bank began creating digital employees using agent technology to automate specific business functions. An agent in the firm’s multi-agent environment comes complete with an employee ID and reports to a human “so there’s a human in the loop for any decision that is being done by the digital employee.”
Hariharan gave a concrete example of how such an agent delivers value in the critical customer onboarding function that’s commonly referred to as Know Your Customer, or KYC. The process carries heavy regulatory requirements which previously required a significant amount of manual work.
“With the help of these digital employees what we’ve been able to do is, instead of humans looking at each and every document, validating the details, we’re now able to use the digital employee to do this,” Hariharan said. “This cut the entire timeframe from about 20 hours that it used to take across the entire manual spectrum to about less than 7 or 8 minutes.”
While there’s considerable debate – even angst – about the ROI of AI investments, the investment is clearly paying off at BNY. Consider that the process outlined above has seen a 99.4% reduction in time required to onboard a customer — and therefore enable revenue to start flowing from that customer. What could your company do with a single agent that can take 20 hours of work – important but manual, administrative work – off an employee’s plate?
Also critical to note: there are 130 “digital employees” or agents now available inside BNY, so there’s a massive potential multiplier effect. The company’s organizational approach to agents is one that I expect to be modeled broadly in the coming year and beyond.
BNY spoke at last year’s AI Tour as well; at that time the bank said it had 40 use cases in production.
Privacy Practices for Healthcare/Medical Products
Boston Scientific outlined mission-critical applications and considerations when it comes to AI usage, with primary focus on data protection. The company has a portfolio of medical tech products and therapies used by physicians for diagnosis and treatment in specialty areas including cardiology, urology, and neurology. It serves more than 40 million patients in 127 countries.
Because it must comply with a wide range of regulations around safety and privacy of data including patient data, Boston Scientific applies stringent controls around data privacy, data access, and governance, according to Saketh Mathur, Senior Information Technology Architect at Boston Scientific.
Mathur outlined practices that govern his firm’s use of AI:
- Use of large language models, which excel in contextual understanding, to precisely detect sensitive data (including private patient health data) through an AI-powered data detection and tagging process
- AI-powered de-identification of personal Protected Health Information (PHI) and Personally Identifiable Information (PII) so the data can be used for further research without being linked back to any individual person
- Ingrain, or embed, role-based data access controls within AI to ensure that data is being accessed legitimately by different users and groups
- Create a governance body with AI champions from different professional disciplines to surface new ideas, manage AI initiatives, and conduct audits. Governance also entails evaluating platforms against AI-specific certifications to ensure they provide required guardrails for compliance with HIPAA and GDPR privacy regulations
The governance and data protection measures practiced by Boston Scientific are critical from a foundational perspective to achieve the kind of broad deployment and impact that BNY is realizing, and those protections serve as a model for companies to derive the most value from AI in highly regulated industries.
— Maya Rock contributed to this analysis.
Related Microsoft and AI Insights:
- Microsoft Advances AI In Healthcare With Bristol Myers Squibb Partnership
- Copilot’s Advantage vs. Stand-Alone Chatbots
- Microsoft Advances ‘Zero-Wait’ Enterprise for Retail Industry
- Becoming Frontier at Microsoft AI Tour NYC: Inside the Agent-First Enterprise
