The Road to an AI-Native Enterprise
Highlights and Insights from EPAM’s Event at the New York Stock Exchange
In conjunction with our partners at First Derivative, we recently hosted an event at the iconic New York Stock Exchange, where a panel of experts discussed the current state and future of AI in the financial services industry. Guests from all swathes of the financial services industry were in attendance for a fireside chat delivered by some of the leading innovators within the AI space. Speaking on the panel were:
- Nir Kaldero, Chief AI Officer, EPAM Neoris
Nir Kaldero is a renowned AI expert, strategist and an acclaimed speaker specializing in the transformative power of artificial intelligence within businesses and society. - Julia Bardmesser, CEO, Data4Real, LLC
Julia Bardmesser is an adjunct professor at the NYU Stern School of Business, author of the recently published book, “From Data to Dollars: Turning Data Strategy into Business Value” and a board advisor to several startups. - Sherry Marcus, Ph.D., Head of AI at Tradeweb
Dr. Sherry Marcus is the Head of AI at Tradeweb where she will lead the next generation of Tradeweb's AI strategy serving more than 3,000 institutional, wholesale, retail advisory and corporate clients around the world.
While there’s no substitution for physically being in the room – akin to listening to a live album of a famous concert versus being present in the crowd – the insights from this discussion were too impactful to keep hidden from the wider financial services industry.
That being the case, what follows are some of the most pertinent portions of the panel’s conversation along with some key considerations for leaders within the financial services industry looking to reap the benefits of transforming into an AI-native organization.
“It’s the combination of humans and machines that makes automation successful, with humans moving from doers to supervisors.”
The discussion opened with participants reflecting on the evolution of AI from traditional Machine Learning to the modern evolution of generative AI (GenAI). As our speakers note, GenAI has made possible really powerful opportunities for automation, with the software development lifecycle (SDLC) being a prime example. However, these capabilities require human in the loop involvement to ensure accuracy. As AI moves from human-assisted SDLC to an agentic development lifecycle (ADLC) with employees playing more of a supervision role, enterprises will need to rethink how they prepare, train and upskill their workforce to unlock the full value of future AI opportunities as part of the shifting product development lifecycle (PDLC) paradigm.
“AI creates a kitchen window into how fractured your data management is.”
The discussion then turned to the importance of data management, which, as the speakers make note, is one of the key pillars of successful AI transformation. As it stands, far too many organizations – even those in the Fortune 500 – have large quantities of siloed and fragmented data along with vast amounts of hidden, unstructured data that can be used by AI. However, AI requires heavily cleaned and standardized data to work effectively. Similarly, GenAI can enable dynamic efficiencies if paired with proper metadata governance and master data management efforts. As such, organizations must prioritize data cleaning, metadata management and governance to prepare raw data for AI systems to leverage effectively and at scale.
“You can’t put AI on top of existing processes. It requires re-engineering the business to be AI-first.”
Next, the conversation turned to the topic of AI transformation and confronting ROI challenges. The speakers noted that widespread AI adoption faces barriers in business readiness, cultural change and process alignment. Our speakers discussed some of the common pitfalls enterprises face, such as executives investing in AI projects without understanding the strengths and limitations of the technology, thereby over-prioritizing hype over effectiveness. They then emphasized the need for effective change management programs to develop an AI transformation framework geared toward process re-engineering and cultural readiness.
“AI must serve business goals, not the other way around.”
From there, our speakers pivoted to the notion of ROI frameworks and the importance of tying ROI to broader business initiatives like customer retention or technology capability outputs as opposed to metrics tied to standalone AI projects. After an insightful exploration of appropriate ROI frameworks, our speakers settled on the idea that businesses need to pair traditional ROI predictions with process-specific data to increase investment credibility to stakeholders on the impact and performance of AI initiatives.
“Agentic AI isn’t just automation. It’s generating data that can transform your future enterprise.”
The conversation then moved to enterprise best practices in AI adoption. One of the use cases our speakers explored here was the use of GenAI by financial firms to automate CRM updates for enhanced sales strategies. Another use case was of a bank experimenting with portfolio optimization via agentic AI. In both cases, what made these projects so impactful was the use of AI to save time by automating tasks while simultaneously capturing clean, actionable data that can then be used by AI at scale.
“Data is liability beyond cost. It requires tight governance to mitigate risks.”
Of course, no conversation on AI could be considered comprehensive without touching on data management and privacy concerns. As our speakers make note, there are serious risks and liabilities that come with training AI models with personal or transactional data. Our speakers cautioned enterprises to consider strict data management protocols, and to limit which datasets are permissible for AI use. At the same time, organizations should deploy multiple data strategies and continue to iterate and evolve on the models that work best for their needs. Even then, organizations should use small data samples for experiments and take extra precautions to ensure clarity around data access governance and compliance.
“Businesses don’t need massive models. They need models fine-tuned to their work.”
Finally, the conversation turned to the outlook for AI adoption as our speakers reached their closing remarks. The point they all seemed to emphasize was that while the technology appears to be racing toward the adoption of larger models, business readiness for small, fine-tuned systems remains crucial. As such, enterprises should be looking to adopt AI solutions tailored to specific business goals rather than pursuing the blanket adoption of larger AI models. To that extent, small language models (SLMs) tailored to the industry or enterprise will likely be the most cost-effective and efficient approach to pursuing GenAI use cases.
Your Roadmap to AI-Native Transformation
At the end of the day, every enterprise is unique and their journey toward AI-native transformation will be equally unique. That said, our panel offered the following advice to enterprises looking to make the leap:
- Implement robust change management for AI transformation across the enterprise, even before you start building your technology strategy.
- Upskill talent for AI-driven development cycles, moving from SDLC with human intervention to ADLC with human supervision in the PDLC paradigm.
- Upskill talent to leverage AI for enterprise transformation through the automation of business functions.
- Invest heavily in your enterprise data (both structured and unstructured) to facilitate the deployment of AI at scale.
- Adopt ROI frameworks that combine business and technical priorities.
- Prioritize agentic AI for operational efficiencies and actionable data generation.
- Tighten data governance to mitigate privacy risks.
- Advocate for simple models and solutions tailored to business-specific goals.
For more of our exclusive insights on AI transformation within the financial services industry, read our report How Financial Services Organizations Can Unlock Real Value with AI.
