Sophisticated agents that take over multiple steps or complete a process, like the alert adjudication agent, described above, will find intelligent automation platforms like Pega or UiPath with embedded genAI support and orchestration strengths to be best.
Expect proliferation of domain-specific agents for financial services. While large language models from Google, Microsoft and Amazon offer a powerful foundation, a gap exists for training data specific to niche applications.
This includes proprietary and non-published data critical for tasks like fraud detection or investment allocation.
On top of these considerations, the underlying models have important differences, such as transparency, governance, IP filters, context window adjustments and multilingual attributes.
With updates at a near insane pace, confusion on how to build agents will be constant.
The reality of AI agents in finance
All articles like this get you excited by the potential of AI – but they must take you down a few notches, and this one will do the same.
We have all seen the headlines: 'AI robots to take over your finance job' and 'The future of finance is fully automated'.
But before you envision AI agents relieving the global financial workload, let us take a deep breath.
I recently interviewed senior financial leaders across investment management, banking and insurance, and let me tell you, caution was pervasive.
AI agents in financial services are coming. They will pose a challenge: balancing and managing a proliferation of agents within the business, brought in by employees and adopted independently.
Enterprises will need a comprehensive plan for orchestrating and coordinating various AI agents and rooting out redundancies and conflicts.
Craig Le Clair is VP principal analyst at Forrester
