GBST has built and deployed agentic artificial intelligence into its Composer Software as a Service (SaaS) platform to automate complex, high-volume wealth administration processes in line with regulatory requirements while removing manual intervention.
Traditional Artificial Intelligence (AI), like ChatGPT, is typically used to analyse data, make predictions, or support individual decisions within a process. In contrast, agentic AI is designed to execute a sequence of steps end to end. By designing agentic AI to operate within defined limits as part of Composer, GBST’s clients can automate complex, high-volume wealth administration processes that currently require significant manual intervention.
Currently these workflows vary by client, product and risk appetite and the lack of standardised pathways makes traditional rules-based automation difficult. GBST’s agentic AI can carry out client-specific operating procedures consistently and at speed without bespoke development, incorporating human oversight where required to approve, intervene or review outcomes. This ensures automation can scale without compromising explainability, auditability or regulatory compliance.
Rob DeDominicis (pictured), CEO of GBST, said, “There’s a lot of noise around AI disrupting software. In reality, in regulated, high-volume environments, it provides an opportunity to improve how work gets done. Composer is already highly automated, with around of 94% of transactions processed automatically. With the introduction of agentic AI, we can move automated processing closer to 99%, targeting the most complex, client‑specific workflows that have historically resisted automation.
“We’ve developed a scalable framework that can be applied across multiple processes and clients over time. By applying AI inside Composer, we can automate complex operational processes in a controlled and auditable way while reducing cost, delay and operational risk. That means working with each client’s own procedures, while maintaining the governance and oversight our clients require.”
GBST has completed its first controlled release of agentic AI capabilities, developed in collaboration with long-term partner, AWS. This release demonstrates how agentic AI can carry out multi-step operational workflows within Composer, with full auditability, human oversight and control. AI agents are configured to operate like experienced administrators, progressing work across the platform, handling exceptions, and knowing when to wait, retry or pass items to the appropriate queue for approval.
The initial application, being run in the UK, has been to automate pension transfer-out requests, which require multiple validations, checks and actions across different parts of the platform. Manual processes that can take hours or days to complete end to end, due to queues, handoffs and waiting for dependent activities, can be executed in seconds once initiated, with the AI agent progressing each step immediately. Rather than forcing standardisation, the approach allows each client’s own processes to be executed consistently and at scale, while remaining aligned to existing policies and controls.
GBST is now engaging clients to apply these capabilities to live operational processes, focusing on areas where manual activity creates cost, delay or risk, ahead of broader rollout.