Apexon, a digital technology services & platform solutions company, has launched AgentRise, a platform that streamlines the transition to AI-native operations by integrating Agentic AI into business functions and decision-making processes. It offers context-aware AI agents tailored for specific industrial processes and embedded within the organisation’s core operations.
AgentRise is a modular, enterprise-ready solution that integrates Apexon’s proprietary intelligence models, decision frameworks, and reusable assets with trusted third-party technologies.
It includes accelerators to speed up time-to-value and safeguards for security and responsible AI use, providing a solid foundation for organisations to implement agent-driven operations that boost growth and efficiency.
In an exclusive conversation with AIM, Mukund Kalmanker, the global head of data, analytics, and AI at Apexon, elaborated, “Agentic AI Brain integrates LLMs for inference and actioning, manages orchestration between agents, and serves as the foundation for this holistic platform. This Agentic Brain is supported by what we call the agent compliance, validation, and governance framework.”
He added, “Once you look at this at the core, you’ll see we’ve created interfaces referred to as agent interfaces and a supporting ecosystem called the agent ecosystem. This includes model context protocol (MCP) servers, agent registries, and we’ve been very quick in incorporating all major market releases into our platform at lightning speed.”
Designed for continuous development, AgentRise enables businesses to incorporate autonomous agents into their workflows, resulting in measurable improvements, such as up to a 40% reduction in processing time and a 30% decrease in workloads.
AgentRise supports the adoption of Agentic AI through two main pillars: Agentic Readiness, which focuses on secure AI foundations using governance frameworks and clean data, and Adoption at Scale, which involves identifying impactful use cases and redesigning workflows for effective collaboration with AI agents.
“Our approach ensures that agentic implementations are solid both at the model level and during execution. Since we work closely with regulated industries, we don’t expect fully autonomous environments yet. For example, with one client, we’re helping them understand how to use system thresholds to manage transactions. Some low-risk transactions can be processed autonomously, while others are routed to humans,” Kalmanker highlighted.
He added that goal alignment and hallucinations can be examined through two main aspects: the model level, which requires specific guardrails, and the agentic frameworks aimed at achieving goals.
A five-point guardrail framework has been created, covering domain adherence to keep agents within their scope, governance of learning for managing datasets, safety to prevent misuse, responsibility to tackle ethics and bias, and compliance with relevant regulations. This approach ensures effective management of hallucinations, responsible AI adoption, and output trustworthiness at the platform level, Kalmanker said.
“On top of that, because agentic systems can autonomously iterate toward goals, we’ve established a validation framework. It’s a ten-point strategy, including components like an agent jury,” he concluded.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.