Agentic AI in Fund Administration – the missing piece of the puzzle for client service?
The AI revolution is accelerating as we move through 2025. Near-daily advancements in AI technology are being announced, with a seemingly limitless range of fields being overhauled.
From Chat GPT and Claude, to Firefly and more, tasks which used to be manual are now being completed by AI platforms at the touch of a few buttons. As the world undergoes this profound transformation and our daily lives - both professionally and personally – see significant structural changes, the impact on asset servicing is also dramatic.
One of the most cutting-edge developments within the asset servicing world is around Agentic AI. Representing a leap forward in the application of AI, Agentic AI is characterized by autonomous decision-making and content creating capabilities, interjected into new or existing automation processes.
It goes beyond traditional automation by allowing users to further scale or automate tasks which would have been impossible just a few years ago, improving accuracy, helping to streamline compliance, and transforming in-house processes for alternatives managers.
Agentic AI: A new digital colleague
Agentic AI uses a complex network of autonomous software components known as "agents" that draw from massive amounts of data and learn from user behaviour to improve over time.
These agents are capable of independent decision making and adaptive problem-solving. When used in the context of asset servicing, this means they can ask questions independently to solve problems; for example, an agent could ask “What transactions on this report are the root cause of this exception?” and then email the employee who booked the exceptions, asking for support.
The agents function by following simple, short instructions known as tools. In the asset servicing world these tools currently cover actions such as:
- Open outlook
- Run accounting report
- Populate general ledger template
To power their decisions, agents are given the context for decision-making or taking actions. Known as its knowledge base, this could be sections of a Standard Operating Procedure relevant to the agent’s goal, for example, or client-specific exception handling. To function effectively, agents also need more detailed instructions, called prompts. These are carefully curated directions or queries that guide the agent's actions by stating their role and goal. A simple prompt could look like the following:
“You are a Fund Accountant reviewing a transaction report which contains each trade for the month. You need to determine which transactions sum to the exception total amount.”
Agentic AI can also be used to identify entirely new automation opportunities which would have previously had too many scenarios to be considered viable. For example, focusing on administrative tasks which have a very wide range of outcomes, Agentic AI could now be used to review legacy automations to see if there are enhancements to be made to upstream or downstream processes, or to human (known as human-in-the-loop) decision points.
Offering significant power-ups to traditional robotic process automation (RPA), Agentic AI enables increased straight-through processing and the ability to minimize human error thanks to its autonomous AI systems.
Client benefits
Fund administrators deploying Agentic AI can deliver significant benefits to their fund manager clients through enhanced operational efficiency and data-driven insights. The fund administration market has evolved to a stage where providers are proactively automating end-to-end processes, moving beyond basic automation to build systems that deliver bespoke capabilities and perform complex calculations on a regular basis.
AI technology enables fund administrators to create sophisticated tools that can scrape, ingest, normalize and synthesize vast amounts of data, forming the foundation for more complex and detailed automation. This capability allows administrators to deliver data and reporting that provides meaningful insights and informs critical decision-making for their clients.
Next steps and implementation
The successful integration of Agentic AI has the potential to revolutionize financial operations, offering enhanced efficiency and accuracy. While it is easy to be enthusiastic about the future, challenges remain. Maintaining a human-in-the-loop mindset for complex decisions, building a library of tools for the agents to use, and ensuring risk, security and compliance processes are in place, are some of the foremost hurdles.
Nonetheless, given its potential to learn and develop further, it is expected that agentic AI will transform many aspects of asset servicing operations.
At the Citco group of companies (Citco) we are committed to assessing and, where possible, implementing technological breakthroughs to deliver ever greater levels of service to clients.