While the world is attached to AI applications oriented towards consumers, the really transformational work occurs below the surface.
“Agentic AI represents a significant discrepancy in observability, going beyond traditional surveillance to autonomous systems and focused on AI that detect, diagnose and even solve problems in real time,” explains Kishore Gopalakrishna, co-founder and chief executive Startree.
Under these AI applications are a complex network of systems that monitor, analyze and act more and more on the constant flow of operational data. This is where real transformation can take place.
“The greatest transformation will be in the detection of anomalies in real time and the analysis of the deep causes, where the agents of the AI will not only surface anomalies faster, but will also provide contextual information,” explains Gopalakrishna.
Passive observers to active participants
Imagine all your infrastructure as an organism living with countless interconnected systems, each generating millions of data points per second. Traditional observability has been the equivalent of placing sensors in all this organization and waiting for alarms to be triggered, how human operators rush to diagnose and solve the problem.
In the agentic AI paradigm, these sensors do not only detect – they send swarms of autonomous digital workers to investigate and repair, often before users notice anything.
“Imagine an e-commerce platform undergoing sudden payment failure,” said Gopalakrishna. “Traditionally, a small team of engineers would manually diminish newspapers and measurements, looking for anomalies – a process that could take hours and often lack deeper models on the scale of the system. With an agentic AI, a swarm of autonomous agents can instantly surpass itself through the whole infrastructure, scanning newspapers, measurements and traces simultaneously. »»
It is not just faster surveillance; It is an entirely new operational model where detection is only the opening movement in a sophisticated sequence of automated responses.
“Unlike traditional observability tools that stop at ideas, agency AI can act,” said Gopalakrishna. “If the problem is a faulty API, agents can automatically reach traffic, provide an additional calculation to mitigate the bottlenecks or temporarily deactivate a defective microservice – without waiting for human intervention.”
Technical foundations
For AI engineers working on data pipelines, the technical challenges of this change are substantial. Traditional observability platforms have not been designed for reading / writing models of AI agents, which could generate thousands of complex requests per second while simultaneously update system configurations according to their results.
The transition requires rethinking everything from data storage to the optimization of requests. Apache Pinot, an open source real-time analysis database which feeds the Startree platform, was designed by thinking of these workloads.
“What really distinguishes us is how the agents of the talkative AI are – they ask for a massive volume of requests when they iterate, refine and synthesize ideas,” notes Gopalakrishna. “This plays directly in the original design philosophy of Pinot: support extreme levels of requests per second (QPS) with low latency.”
The technical battery which feeds agentic observability includes several critical layers:
- Data ingestion and evolution of the diagram: As Gopalakrishna points out, “one of the biggest challenges is that real -time streaming data are constantly changing – a Kafka subject scheme can evolve without notice.” Systems must adapt transparently without breaking the requests or requiring manual intervention.
- Vectorization and search for similarity: Modern observability is not only a question of exact correspondence – it is a question of finding models and similarities between large sets of data in real time. “When we introduced RAG to Pinot by providing support for vector data types, Pinot was already capable of ingestion, indexing and provision of integration of vectors available for real -time interrogation,” explains Gopalakrishna.
- Autonomous sanitation systems: The last part is the action layer, where AI agents can make changes authorized to infrastructure based on real -time analyzes. This requires sophisticated authorization models and guarantees.
Real -scale real world applications
Potential applications cover all industries with complex digital infrastructure.
“Surveillance of the electrical network where public services can constantly analyze the data of IoT sensors to detect abnormal tension models by comparing real -time readings with historical events, by identifying and preventing breakdowns before they occur,” explains Gopalakrishna.
In financial services, traders can “identify movement behavior of actions similar to thousands of titles in real time, which allows them to respond instantly to emerging trends”.
But perhaps the most promising application lies in the observability of AI systems themselves. As linguistic models and other AI systems become more anchored in critical trade processes, monitoring and maintaining their performance, precision and safety are becoming more and more vital.
The economy of intelligent observability
The transition to the observability of the agency AI is not only a technical improvement – it fundamentally modifies the structure of the costs of maintaining digital infrastructure.
“Observability costs are already decreasing as companies move away from monolithic platforms and owners to open source solutions such as Openedlem, Kafka, Apache Pinot and Grafana,” notes Gopalakrishna. “The open source economy, associated with managed cloud services and more foreseeable pricing models, broke the cost cycle on a linear scale with data volume.”
However, the impact of the most significant costs comes from the considerable reduction in stopping and performance problems. “When the agent AI adds value, it is not only the reduction in costs, but more quickly to detection and resolution – which is ultimately where companies lose money,” said Gopalakrishna.
The future: from observability to orchestration
As agentic AI matures, the line between observibility and orchestration will blur. The watching systems and the action that will become increasingly integrated, creating digital environments that are continuously campaigning.
For engineers working at this border, the challenge is not simply technical. It reconcepts their role in a world where the infrastructure is more and more maintained. The most precious skills will drop from reactive troubleshooting to the proactive design of autonomous systems.
Gopalakrishna considers this transition to be inevitable. “The market happens to us – because Startree already offers the performance of the subsecond, ingestion in real time and the infrastructure ready for the AI required to operate AI on large -scale structured data.”
In this new landscape, the silent sentries who monitor our digital universe gain both intelligence and the agency. The question is not whether automated systems will take control of our management of our infrastructure – this is the speed with which engineers can adapt to their new role as architects of autonomous digital ecosystems.
Image credit: istockphoto /Artemediana