Getting a return on the billions of dollars invested in artificial intelligence has become a defining theme of discussions over the past year and that was reflected in this year’s A-Team Group Data Management Summit New York City.
With the technology’s honeymoon over, now is the time for payback; over a day of expert panel discussions and keynote addresses, delegates heard just how capital markets participants are beginning to bring efficiencies and generate revenue from their AI deployments.In the opening keynote fireside chat between Samit Korgaonkar, Enterprise Data Platform and Governance at Google, and Peggy Tsai, Data Management and AI Leader, a central plank in that pursuit of ROI came into focus – scaling AI-ready data.
The discussion touched on the challenges organisations often encounter as they endeavour to bring the power of AI to bear across entire enterprises.
Agents of Change
Legacy tech infrastructure and data management strategies remain the most persistent obstacle, the summit heard, especially among organisations that have grown through acquisition, who may struggle to knit together the disparate systems of their procured businesses.
Often, it’s not only the data and hardware that’s siloed. Company culture can develop blinkered ways of thinking and this can be as difficult to overcome as a fragmented tech stack, the summit heard.
Unified data overlays and semantic layers are helping to alleviate many of those hurdles, as is the transformation of the data engineer’s role, from one of passive observation-and-response to embedded agent of change, harnessing AI to ingest, cleanse, manage and distribute data.
Agentic workflows are also beginning to take the burden of the data management processes necessary to make the most effective use of AI.
Buy More, Save More
Rodrigo Coelho, Chief Executive of Edge & Node, complimented the observations of the opening address in his keynote on how organisations can actually reduce costs by buying in more data. The summit heard how the deployment of specialised infrastructures for domain-specific agents would bring greater cost efficiencies than specialised agents working within a generalised tech stack. One of the biggest advantages of this strategy is that agents could be relied upon to set up the new architectures, eliminating establishment costs.
This approach would require some compromises, for instance in terms of latency, but careful balancing of the organisation’s operational needs against its objectives would bring cost savings, the summit heard.
Integrated AI
Sreelatha Purushothaman, Head of Enterprise AI and Data Governance at BNY, explained how the financial service giant has managed to achieve performance efficiencies by building its own AI platform, named Eliza.
The summit was told how Eliza had enabled the bank to increase software releases, accelerate onboarding and drastically reduce the time to complete sanctions reviews.
The bank has done this in part through the creation of specialised agents that possess domain knowledge, act as workflow assistants, perform some roles semi-autonomously and connect with vendors. Eliza, the summit heard, acts as the orchestrator of these agents in conjunction with the bank’s systems, its processes and its people.
Trusted Data
For some organisations, AI and agents have become the actuators of data management processes and Gille Halle, Financial Services Data Integration Leader at IBM, explained how that is being achieved through a foundation of trusted data governance and real-time event streams.
A “decision-driven organisation” connects trusted data directly to governed action so that employees or AI agents can respond in real time with the correct context, the summit heard. Those without a strong, trusted data foundation will struggle to get projects off the ground.
The summit heard how a foundation could be created with three core pillars: master data management to ensure data completeness; the capture of events such as payments and claims as reusable data streams for use across the enterprise; and reliable connectivity between mainframes and modern AI.
With strong governance policies, delegates heard, organisations can operate AI under defined boundaries while leaving high-impact cases to humans-in-the-loop. As well, legacy systems wouldn’t need a “rip and replace” programme, and organisations can incrementally expose mainframe transactions into governed event streams.
Knowledge Layer
Approaches to overcoming the restrictions of legacy systems were also a central plank of Rocket Software president for data modernisation Michael Curry’s keynote.
During this delivery, delegates heard how Enterprise Knowledge Layers can facilitate a controlled evolutionary change rather than a revolutionary overhaul.
Again, the connection of trusted data was highlighted as an essential ingredient that should go hand in hand with the establishment of a semantic layer, strong governance and human oversight.
By treating agents as co-workers, the Enterprise Knowledge Layer generates a complete audit trail, making decisions easier to defend from a regulatory standpoint than traditional manual spreadsheets, the summit heard.
Subscribe to our newsletter
