From Hype to Hard Truths: HSBC’s AI Development Journey
Building scalable solutions, navigating risk, and creating an AI-ready culture remain critical tasks
HSBC is navigating a complex and rapidly evolving artificial intelligence landscape. It must leverage cutting-edge technology while managing significant operational, regulatory, and strategic risks.
As one of the world’s largest financial institutions, the bank’s AI journey isn’t just about adopting the latest tools—it’s about deploying them at scale, responsibly, and in ways that bring tangible value.
“Our ambition is probably similar to the ambitions of a lot of other people using AI,” said Dara Sosulski, Head of AI and Analytics Business Enablement and Securities Services at HSBC. “We want to…transform our business and operations into something that’s agile, efficient, resilient, scalable, and allows us to reinvest into areas where we think we’re going to grow.”
That transformation, however, is easier said than done. Moving AI tools from proof-of-concept to production in the highly regulated world of finance requires not only technical expertise but also governance, infrastructure, and cultural alignment. “It’s not a panacea, and it’s really hard to do well, and it’s really, really hard to do at scale responsibly,” Sosulski warned.
Speaking at Tech Show London on March 12, 2025 at Excel, London, Sosulski presented an in-depth look at HSBC’s AI strategy under the theme “Lessons Learned from the Frontlines of Responsible AI.”
Drawing from over a decade at HSBC and with a background in neuroscience, she explored both the promises and pitfalls of integrating AI into large-scale financial operations.
HSBC, headquartered in London, is a global banking and financial services company known for its mission of “opening up a world of opportunity.” The bank leverages its vast global reach and expertise to support clients, communities, and markets in over 60 countries. Its AI efforts reflect a commitment to both innovation and responsibility in a highly competitive sector.
The Gartner Hype Cycle
Central to Sosulski’s talk was a reference to the Gartner Hype Cycle, a widely respected framework for tracking the adoption and maturity of emerging technologies. The cycle begins with the “innovation trigger,” rapidly ascends to the “peak of inflated expectations,” and then plunges into the “trough of disillusionment.”
Surviving technologies move on to the “slope of enlightenment” and finally reach the “plateau of productivity.”
“I love the Gartner Hype Cycle. Like every year, I wait with bated breath to see the new figure come out,” Sosulski shared.
Reflecting on the hype around generative AI, she remarked, “By and large, we passed that peak of inflated expectations for generative AI. At least, we’re past the honeymoon phase, and we’re all getting a much better sense about how to do this in practice.”
Despite reaching that point of “enlightenment” with GenAI, the rise of agentic AI systems has once again sparked a wave of inflated expectations. “Everybody’s really excited about this stuff, and we have to go through that whole cycle again,” Sosulski noted.
HSBC’s AI strategy revolves around three core goals: enhancing client service, improving operational efficiency, and strengthening risk management. Practical applications already include the Dynamic Risk Assessment Model, developed with Google to rate customer risk, and the AI Markets Chat Bot, which supports sales and trading teams using advanced natural language processing and now, generative AI.
But the focus is not on chasing the newest model; it’s on effectively operationalising AI.
“It’s not about taking the latest model and plugging it into something. It’s about pairing the right model with your proprietary data and control frameworks,” Sosulski explained.
Build or Buy? What to Consider?
A pivotal decision for companies on the AI journey is whether to build solutions internally or buy from external vendors. Sosulski advised businesses to assess several factors:
In-house skills: “Do you have the skills in-house—IT developers, data scientists, or data engineers—to do this?”
Infrastructure: “Do you have the infrastructure that you will require on-premises? Can you access it easily? Do you have the knowledge to figure out what you need?”
Vendor dependency: “Do you want to exclusively leverage the infrastructure of one vendor, or do you want to give yourself a couple of different options that you can plug and play with?”
Use-case complexity: “Not all problems require all of that firepower. There are some you can get pretty far with prompt engineering, hard coding guardrails, and using RAG.”
“There is no right answer. It’s all about what suits your business the best, essentially,” Sosulski emphasized.
Scaling Responsibly
As AI integration deepens, risk governance has taken center stage at HSBC. Sosulski stressed the importance of policies that are “simple, risk proportionate, and business conscious,” cautioning that overly complex or rigid frameworks can stifle innovation. “Most people don’t do that… really, just pay a lot of attention to how obviously proportionate and easy these things are to understand and enact.”
She highlighted unexpected risks that emerged in practice, including data privacy issues, systemic vendor reliance, and even the rise of AI washing—misleading clients about AI use. “The SEC has started fining people for misleading clients about how they’re using AI and machine learning,” she revealed.
Another concern is over-reliance on a few AI vendors. “There’s still an imbalance… and that does introduce systemic risk because of the server reliance,” Sosulski said. “If somebody like Sam Altman decides to change his pricing model tomorrow, lots of us are screwed.”
HSBC emphasizes scalable, well-controlled IT environments to mitigate risk while fostering innovation. These environments include sandbox platforms where employees can experiment safely. “It allows lots of people to get hands-on experience… and it’s very useful for your institution if you want to continue developing talent from within,” Sosulski explained.
Within this structure, successful ideas can be transformed into modular AI tools accessible across HSBC’s platforms and adaptable for varied internal and external users. The goal is interoperability, allowing HSBC to remain flexible amid rapid technological shifts.
Creating an AI-Ready Workforce
Perhaps the most human element in HSBC’s strategy is the drive to create an AI-ready culture. Sosulski believes strongly in inclusivity and continuous learning. “Everybody’s got a role to play, and everybody should feel like they can and should learn more and figure out how they best plug into this AI pipeline of work.”
She emphasized collaboration over competition. “We take them, we identify them, and then we make them work together to develop a solution… so we don’t have a bunch of people wasting time.”
To support this, HSBC has developed an AI Academy, a hub of curated courses, books, and resources for employees to explore AI at their own pace.
“It’s amazing what people can do if you give them a sandbox and a little bit of a steer,” Sosulski said.
HSBC’s AI strategy is not about chasing trends but thoughtful, responsible, and scalable innovation. Through careful investment, robust governance, and a culture of learning, HSBC is positioning itself to ride the AI wave and shape it in a way that delivers lasting value. The challenges are real, but so is the bank’s commitment to transforming opportunity into action, with people—and purpose—at the core.