The world of financial services is a high-stakes arena where IT disruptions can have far-reaching consequences – from regulatory exposure and reputational damage to significant financial losses. While many institutions have invested in monitoring tools, a common frustration persists: IT teams often find themselves in a reactive cycle, constantly chasing “red alerts” and symptoms rather than proactively addressing the root causes of issues.
But what if there was a way to predict failures before they happen? What if IT systems could move beyond simply flagging problems to autonomously resolving them? This is precisely the shift we are seeing with the integration of Artificial Intelligence (AI) and Machine Learning (ML) into financial services operations, specifically through the evolution of AIOps.
The Scale and Complexity Challenge
The sheer scale and complexity of modern IT operations in financial services are outpacing the human ability to keep up. Relying on “eyes on glass” monitoring and reactive problem-solving is no longer sustainable. As Dale Skeen noted, “The only way that we can achieve the scale, whether you’re provisioning the system or whether you’re monitoring the system, is really to start bringing AI and agentic AI into the loop.”
The advent of Large Language Models (LLMs) and generative AI (GenAI) is proving to be a game-changer. These technologies are enabling the development of AI agents that are mostly autonomous that can assess the situation, evaluate, decide, and act on their own.
Shifting Goalposts: Towards Self-Handling Systems
The traditional view of AIOps is expanding. Infrastructure, once primarily defined by physical data centers and hardware, is now taking on a more existential context. The new goal? Self-healing systems. This means systems that can:
- Monitor and detect problems.
- Correlate disparate alerts.
- Analyze the situation.
- Decide on the best course of action.
- Remediate problems.
We are still at the beginning of this journey, but the next generation of AIOps platforms, powered by GenAI, are providing the foundation for this transformation. These advancements offer incredible potential, but also introduce new challenges, particularly around establishing trust in autonomous AI decisions.
Building Trust: The Power of Knowledge Graphs
How do we build trust in these AI-powered systems, especially at a technical level? The key lies in structured knowledge, often referred to as knowledge graphs. These graphs capture knowledge about all aspects of IT systems and operations, including:
- System components, applications, and microservices
- Topologies (device, service dependencies)
- Diagnostic knowledge (fault patterns, symptoms, root causes)
- Environmental knowledge (server location, data center temperature, weather conditions)
This knowledge is captured in a format that is both human and machine-readable, making it accessible to AI and GenAI systems for reasoning. By providing this rich, structured knowledge, several benefits emerge:
- Improved Accuracy and Precision: More data and better knowledge lead to more accurate answers from the AI
- Better Reasoning: Structured knowledge allows AI to infer stronger logical connections and improve reasoning capabilities.
- Explainability: With externalized knowledge, AI can explain its answers based on its chain of reasoning through the knowledge graphs, enabling humans or other AIs to review and verify the decisions.
- Guardrails: Knowledge can be used to set boundaries for AI reasoning, ensuring decisions are based on provided data and highlighting when the AI is making conjectures or guesses
- This comprehensive approach to knowledge not only enhances the accuracy of AI but also builds essential trust and confidence in these increasingly autonomous systems.
Realizing Efficiencies: Faster Resolution and Beyond
The benefits extend beyond building trust. AI-powered AIOps platforms are driving significant efficiencies:
- Reduced Time to Resolution: The time it takes to resolve incidents is drastically decreasing.
- Autonomous Remediation of Simple Failures: For low-impact failures, AI can now manage detection, correlation, analysis, decision-making, and remediation entirely autonomously. (Some organizations see a high rate of autonomous resolution for these simple issues)
- Human-in-the-Loop for Medium Problems: For more complex, medium-level problems, humans can remain in the loop, collaborating with AI to resolve issues efficiently.
- AI plays an elevated role for very complex problems. Unlike a “bot” that just responds to questions, AI acts as a co-pilot that actively “brainstorms” with the Operations Engineers in a collaborative fashion. (This is radically different from what was expected just a few years ago.)
The shift towards AI and ML in financial services IT operations represents a move from a reactive, symptom-chasing approach to a proactive, predictive, and increasingly autonomous model. By embracing self-handling systems powered by GenAI and built on robust knowledge graphs, financial institutions can enhance efficiency, reduce risks, and build more resilient and trustworthy IT environments.
