In conversation with Jayakrishnan Saisdharan, Executive Director and Chief Information Officer at Geojit Financial Services, we explore how AI agents are moving from pilots to production in wealth management.
CIO&Leader: Where are you today on AI agents in financial services industry?
Jayakrishnan Saisdharan: We have a clear strategy in leveraging AI across the organization. AI agents are a key part of the strategy and we have them at various stage including POCs, pilots and in production. The key areas where we are focussing on in the first phase are in research, customer support and operations and onboarding.
CIO&Leader: What’s the most advanced AI agent you’ve deployed so far?
Jayakrishnan Saisdharan: We have few advanced AI agents in the works. The few we have in production are in the following areas.
We have deployed an Intelligent insights agent that performs a Statement-to-Sheet automation for our PMS business. This has brought in automation, speed and improved operational efficiencies. We also have an AI-driven Equity Research Assistant that generates structured summaries and draft research notes, improving research turnaround and analyst productivity.
CIO&Leader: Where do you draw the autonomy line in regulated workflows?
Jayakrishnan Saisdharan: We believe we have a lot more to leverage AI in non & semi-autonomous areas. As a strategy we want to be a retail wealth partner for our customers and towards that we are increasing our branch presence and customer support initiatives. We want AI to play an assistive role and further this strategy.
CIO&Leader: How much financial or regulatory risk can you delegate to AI?
Jayakrishnan Saisdharan: This is an area that is currently under evaluation as we want to be diligent on how AI can assist an already mature capability that has been tuned over many years of operation. While, financial exposure and compliance accountability remain with humans, we will have AI agents that will support and hasten this process to help our customers leverage the new capabilities our markets provide.
CIO&Leader: If regulators or auditors asked you to explain every autonomous decision made by an AI system last quarter, could you provide a clear audit trail of what happened and who approved the underlying rules?
Jayakrishnan Saisdharan: We strongly believe production deployment requires full auditability—data lineage, model/version tracking, decision logs, and human approval trails. This is a key factor in our AI governance council as we evaluate initiatives.
CIO&Leader: As AI agents connect to trading platforms, CRM, KYC/AML systems, core banking integrations, and market data platforms, what safeguards ensure access control, traceability, and auditability?
Jayakrishnan Saisdharan: Today, we have controls which include least-privilege access, maker-checker workflows, policy-based limits, immutable logging, real-time monitoring, and kill-switch capability. AI can strengthen and also leverage the very same mechanisms to ensure control.
CIO&Leader: Have you paused or rolled back any AI agent initiatives?
Jayakrishnan Saisdharan: The main challenges are not AI itself, but more around AI readiness. For us, data standardization, business rules clarity, and operational readiness are focus areas than model capability in order to get the best out of AI.
CIO&Leader: How are you redefining ‘control’ in financial services as machines act?
Jayakrishnan Saisdharan: Control is shifting to policy guardrails, strict execution limits, continuous monitoring, automated alerts and human override. Human orchestrators will audit, report and prioritize how agents and humans will work together and at the same time manage regulatory and compliance needs.
CIO&Leader: In the next three years, which financial services decisions do you realistically see AI agents taking on autonomously and which will always require human judgment?
Jayakrishnan Saisdharan: AI will support workflow routing, document processing, reconciliation, and fraud triage, while financial and regulatory decisions remain human-led. BFSI will invest and strengthen fundamental systems on data, quality and expiry so that AI can act on “real” data that is both consented and proprietary. Organizations will reimagine operations and back-office structures to leverage the agentic world and create more adaptive and cross functional value chains.
CIO&Leader: Five years from now, do you see AI agents becoming a core execution layer across brokerage and wealth platforms, or will scale be constrained by regulation, architecture limits, data quality, or cost pressures?
Jayakrishnan Saisdharan: AI agents will become a key operational layer across brokerage and wealth platforms, with adoption shaped by regulatory comfort, architecture readiness, data quality, and cost discipline. BFSI organizations will significantly accelerate product creation and launches by transitioning from manual, siloed workflows to agentic AI-first operating models.