Latest news with #AgentOps
Forbes
24-07-2025
- Business
- Forbes
AgentOps: Operationalizing Agentic AI
Shailesh Manjrekar is the Chief AI and Marketing Officer at inventor of "The Agentic AI Operational Intelligence Platform." As AI systems evolve from simple chatbots to autonomous agents capable of complex reasoning and decision making, a new operational discipline is emerging: AgentOps (also known as AgenticOps). This discipline applies both to BizOps as well as ITOps. This represents the latest evolution in AIOps, building upon the foundation established by earlier disciplines—such as MLOps, DataOps and AIOps—that organizations have been adopting since the early 2020s. As organizations embarked on digital transformation journeys, new operational disciplines emerged to operationalize AI across different layers of the technology stack. MLOps and LLMOps focused on machine learning model lifecycle management, DataOps brought agility to data management and governance and AIOps applied AI to IT operations and monitoring. Each borrowed collaboration principles from DevOps, creating bridges between line-of-business and IT engineering teams. Now, as autonomous AI agents become more sophisticated, AgentOps represents the next frontier—managing not just models or data pipelines but entire autonomous systems that can perceive, reason and act independently in complex environments. What Is AgentOps? AgentOps is the end-to-end lifecycle management of autonomous AI agents—software entities that can perceive, reason, act and adapt in real time within complex environments. Unlike traditional software or even static machine learning models, these agents are dynamic, non-deterministic (stochastic) and capable of making independent decisions. Think of it as DevOps for autonomous AI systems. AgentOps extends the principles we know from AIOps and DevOps to address the unique challenges of managing AI agents that can: • Make autonomous decisions. • Interact with multiple external systems. • Collaborate with other agents. • Adapt their behavior in real time. • Self-heal and self-optimize. Key Capabilities Of AgentOps • Comprehensive Lifecycle Management: From initial design through deployment, monitoring and continuous refinement, AgentOps covers every stage of an AI agent's existence. • Advanced Observability: Unlike traditional monitoring, AgentOps provides detailed logging of agent decisions, action paths and interactions with external systems, enabling complete traceability and debugging. • Multi-Agent Coordination: Modern AI systems often involve multiple agents working together. AgentOps frameworks facilitate structured communication and coordination among agents to achieve collective goals. • Governance And Control: While agents operate autonomously, AgentOps ensures mechanisms exist for curated access, intervention, error-handling and alignment with organizational objectives. The Evolution Of AI Operations The journey to AgentOps began with the foundational disciplines that emerged during the early wave of AI adoption. MLOps established practices for model cataloging, version control and deployment, focusing on reliably integrating machine learning models from development into production. DataOps brought agility to data management, ensuring organizations could transform and operationalize data as their "new source code." AIOps applies artificial intelligence to IT operations, utilizing historical and real-time data for full-stack observability and automated incident response. Each of these disciplines addressed specific operational challenges, but they were primarily designed for more static, predictable systems. MLOps manages models that, once deployed, perform consistent functions. DataOps handles data pipelines with defined transformation rules. AIOps monitors and responds to infrastructure patterns that, while complex, follow observable patterns. AgentOps extends beyond these foundations to manage something fundamentally different: autonomous agents that don't just process data or execute predefined functions but make independent decisions, adapt their behavior in real time and coordinate with other agents to achieve complex goals. The infrastructure requirements reflect this evolution. Traditional disciplines rely on established platforms—GPUs and model registries for MLOps, data lakes and transformation tools for DataOps, monitoring systems for AIOps. AgentOps requires a new platform architecture: multi-agent frameworks, external API orchestration and sophisticated governance tools to manage autonomous behavior safely. The Complexity Challenge AgentOps introduces several layers of complexity beyond traditional MLOps/LLMOps: • Autonomous Decision Making: Agents don't just generate responses—they make decisions that can trigger real-world actions with significant consequences. • Multi-Agent Interactions: Managing communication, task delegation and conflict resolution between multiple autonomous agents. • Dynamic Adaptation: Agents that modify their behavior based on changing environments and new information. • Expanded Attack Surface: Autonomous agents interacting with multiple systems create new security and compliance considerations. Business Value Of AgentOps As organizations increasingly deploy autonomous AI agents for critical tasks, outcomes become essential to measure the ROI: • Business Agility: Increased digital customer value at marginal cost. • Quality And Resiliency: Operational effectiveness and efficiency. • Risk Mitigation: Preventing unpredictable or unsafe agent behavior before it impacts operations. • Transparency And Accountability: Understanding how and why agents make specific decisions. Maintaining trust and compliance as AI agents become more capable and independent. • Scalability: This is not about scaling compute or storage; this is about scaling intelligent (data-driven) decision making and/or executable actions at scale. The Blueprint For Success As you embark on this autonomous journey, follow a structured approach with well-defined KPIs: • Start with a business-driven use case. • Build agent-aware infrastructure using an AgentOps platform. • Develop agent-literate teams. • Scale through agent ecosystems. • Optimize continuously. Key Capabilities To Consider When Choosing An Agentic Platform Choosing the right AgentOps platform is one of the important steps in your agentic journey. Ensure the platform is able to support the agentic lifecycle, with access to curated datasets and with the right security, trust and governance framework. Some of the key capabilities should include: • Event-Driven Architecture: Real-time streaming and orchestration • Intelligent Agent Development: Visual tools and LLM-guided workflows • Advanced Quality And Risk Management: Agent-based QA and guardrails • Innovative User Experience: Generative UX and explainability • Context And Prompt Engineering: Innovative use of context management and prompt templates • Comprehensive Data Integration: Universal data integration, enrichment and orchestration using data fabric • Flexible Development: Low-code and dynamic tooling The future of AI operations isn't just about managing models; it's about orchestrating intelligent, autonomous systems that can think, decide and act on their own. AgentOps is how we get there safely. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Business Insider
26-06-2025
- Business
- Business Insider
Are AI Agents Doing Their Jobs? IBM's AgentOps Lets You Find Out
As generative AI moves beyond basic demos and starts transforming industries, one important question is becoming critical: How can you be sure that your AI agents are working the way they should? Unlike traditional software, these agents don't just follow step-by-step instructions. Instead, they take initiative, find answers, monitor systems, or even write code on their own. But as more companies begin to rely on these systems, they need clear ways to track and verify that their agents are doing their jobs correctly and consistently. To solve this challenge, IBM (IBM) researchers created something called AgentOps. Confident Investing Starts Here: Since AI agents often behave unpredictably, use flexible logic, and interact with other software and tools, it's difficult to monitor them with traditional methods. As a result, AgentOps acts like a 'dashboard under the hood' that helps developers and engineers understand how the agent makes decisions, what tools it uses, and whether it performs tasks as expected. Furthermore, it tracks changes in behavior, detects issues in real time, and compares current results to past performance. This not only helps improve reliability and accountability, but also allows agents to get better over time. AgentOps is built on OpenTelemetry (OTEL), which is a common software tracking standard. It works with platforms like LangChain, watsonx, and CrewAI to trace agent activities. Interestingly, IBM added an analytics platform on top of OTEL to give developers deep insights by showing exactly how agents behave. It even offers suggestions on how to make them faster or more accurate. These analytics are powered by AI and can recommend ways to improve workflows, as well as cut costs. It is worth noting that AgentOps is already being used in IBM products like Instana and Apptio. What Is the Target Price for IBM? Turning to Wall Street, analysts have a Moderate Buy consensus rating on IBM stock based on seven Buys, five Holds, and two Sells assigned in the past three months, as indicated by the graphic below. Furthermore, the average IBM price target of $269.46 per share implies 7.2% downside risk.
Business Wire
25-06-2025
- Business
- Business Wire
SUPERWISE® Launches First Open, Enterprise AgentOps Solution for Securely Running Third-Party AI Agents
NASHVILLE, Tenn.--(BUSINESS WIRE)--SUPERWISE®, the leading Enterprise AI Governance and Operations platform, today unveiled a bold advancement in the AI landscape. While much of the industry remains focused on building agents, SUPERWISE is tackling the much more complex - and often underestimated - challenge: how to operate AI agents at scale with full governance, observability, and control. This shift reflects a growing recognition that the true test of AI maturity lies not in development, but in dependable, compliant, and scalable operations. 'Building AI agents is only half the equation. The real challenge is managing them responsibly once they're live—and that's where SUPERWISE excels.' — Russ Blattner, CEO, SUPERWISE® With the launch of its open AgentOps platform, SUPERWISE enables companies to safely deploy agents developed in a variety of proprietary and open-source development platforms. This release emboldens teams to deploy, serve, and manage AI agents within the SUPERWISE platform, complete with built-in compliance, monitoring, and operational oversight. It marks a significant step forward in making responsible AI not just possible, but secure, scalable, and successful. 'Building AI agents is only half the equation,' said Russ Blattner, CEO of SUPERWISE. 'The real challenge, and where organizations often stumble, is in managing them responsibly once they are live. This is precisely where SUPERWISE's expertise and leadership have consistently distinguished us - at the operational layer. With this launch, SUPERWISE is enabling teams to use the best open-source tools to build agents, while relying on our enterprise-grade infrastructure to govern, observe, and scale them safely.' SUPERWISE's AgentOps platform will benefit a variety of stakeholders: AI Developers & Engineers: Use their preferred tools without sacrificing operational oversight Enterprise IT & AI Leaders: Centralize operations while enabling innovation and avoiding vendor lock-in C-Level Executives: Balance agility with governance, security, and scalability, and a lower total cost of ownership The platform enables developers to capitalize on a variety of common preferences, including open-source software, low-code interfaces, built-in integrations, strong community support, and ease-of-deployment. The platform today supports the deployment and operation of agents for its development framework Flowise, and a growing list of soon-to-be announced third-party frameworks, including Dify, CrewAI, Langflow, N8n and many others. 'Developers have their choices for open source frameworks. Rather than forcing them to switch in order to be governed, SUPERWISE allows developed agents to be run in our platform, which maximizes existing development investment without incurring the risks,' said Oren Razon, Senior Director of Product at SUPERWISE. About SUPERWISE SUPERWISE® is the Enterprise AI Governance and Operations Platform purpose-built for real-world AI. With a platform-first approach, it unifies artificial intelligence operations and governance, risk, and compliance into a single, scalable foundation, enabling complex industries to deploy, run, monitor, and manage AI with confidence and proper oversight. Recognized by Gartner as a Cool Vendor in Enterprise AI Governance and a pioneer in MLOps, SUPERWISE® delivers built-in guardrails, observability, explainability, and compliance, positioning AI as a trusted foundation for successful business transformation.
Techday NZ
25-06-2025
- Business
- Techday NZ
Superwise launches AgentOps for secure & compliant AI agent management
SUPERWISE has announced the introduction of its open AgentOps platform designed to provide real-time observability, control, and compliance for companies deploying third-party AI agents. The new solution is intended to address what the company describes as a significant gap in the industry, as businesses ramp up their deployment of AI agents without adequate measures for risk mitigation and operational oversight. The AgentOps platform seeks to centralise and secure the management of AI agents, serving companies that increasingly rely on varied and decentralised agent architectures. Operational oversight The AgentOps release enables enterprises to deploy, serve, and manage AI agents created using a range of proprietary and open-source development platforms. Through this initiative, SUPERWISE provides built-in capabilities for compliance, monitoring, and operational management, positioning its service as a component for responsible and scalable AI deployment. Russ Blattner, Chief Executive Officer at SUPERWISE, highlighted the current challenges in the AI landscape. "Building AI agents is only half the equation," he said. "The real challenge, and where organizations often stumble, is in managing them responsibly once they are live. This is precisely where SUPERWISE's expertise and leadership have consistently distinguished us - at the operational layer. With this launch, SUPERWISE is enabling teams to use the best open-source tools to build agents, while relying on our enterprise-grade infrastructure to govern, observe, and scale them safely." Supporting diverse needs The AgentOps platform is aimed at a broad spectrum of stakeholders within the enterprise. AI developers and engineers are able to continue using their preferred frameworks and tools while maintaining operational visibility and controlled workflows. Enterprise IT and AI leaders are provided with centralised management, allowing them to encourage innovation while avoiding dependency on single vendors. C-level executives are presented with tools to balance agility, governance, security, scalability, and cost. The development philosophy behind AgentOps includes support for open-source software and low-code solutions, as well as built-in integrations and community-driven tooling. According to the company, the platform currently supports the deployment and management of agents developed through its Flowise framework, with planned compatibility for additional third-party frameworks such as Dify, CrewAI, Langflow, and N8n. Framework flexibility Oren Razon, Senior Director of Product at SUPERWISE, commented on the platform's role in letting developers maximise the investment in their tool choices. "Developers have their choices for open source frameworks. Rather than forcing them to switch in order to be governed, SUPERWISE allows developed agents to be run in our platform, which maximizes existing development investment without incurring the risks," he said. SUPERWISE claims that as the deployment of AI agents becomes more widespread, the need for integrated governance, risk management, and operational transparency will increase across industries. The AgentOps platform is positioned to offer enterprises a cohesive and extensible approach to agent oversight, aimed at supporting ongoing compliance requirements and auditability as regulatory frameworks evolve. The company points toward its experience in governance and operations as being central to this new release, making reference to the rising recognition within the industry that maintaining secure and auditable AI systems is becoming as critical as developing the agents themselves. SUPERWISE is recognised by analyst firms for its contributions to enterprise AI governance and MLOps, framing its platform as offering integrated guardrails and compliance functionality for enterprises seeking to embed responsible AI within their operational processes.
Forbes
11-04-2025
- Business
- Forbes
XOps For Enterprise AI: The Convergence Of MLOps, LLMOps And AgentOps
Chiranjiv Roy spearheads globally products, solution and consulting across industries at getty Today's enterprise AI landscape faces exponential growth in model complexity and data volumes, posing significant challenges. As organizations rapidly scale their AI ambitions, they inevitably encounter bottlenecks related to operational efficiency, compliance and scalability. To address these challenges comprehensively, businesses require an integrated approach—XOps, which combines MLOps, LLMOps and AgentOps. This unified framework isn't merely about execution; it's about strategically leveraging AI operations to deliver sustainable business value. AI operations today require more than discrete practices. MLOps helps streamline traditional machine learning workflows, LLMOps enables the efficient deployment of sophisticated language models and AgentOps coordinates complex autonomous agent systems. However, implementing these components in isolation misses significant opportunities for holistic efficiency and strategic value. XOps solves this by bringing these distinct yet complementary operational disciplines under one strategic umbrella, ensuring smoother, more scalable adoption of AI capabilities. For example, in one impactful experience with a global consumer electronics company, supply chains suffered from manual, resource-heavy processes, slowing insights and innovation. By developing a no-code, intuitive ML platform with automated data pipelines and AutoML capabilities, business and data analysts independently designed and deployed models without extensive IT involvement. The results were transformative: • Dramatically faster project cycles • Significantly reduced dependence on engineering teams • Enhanced strategic agility, empowering quicker, informed supply chain decisions As another example, in healthcare, inconsistent and siloed workflows complicate and delay AI adoption due to compliance risks. Establishing a standardized, end-to-end MLOps pipeline ensures consistent, compliant model deployment across diverse teams. In our experience, automating data preprocessing, model validation and real-time monitoring can significantly shorten deployment timelines, improve regulatory compliance and strengthen collaboration between technical and business stakeholders. A digital analytics agency we worked with faced slow insights generation and scalability issues from fragmented NLP processes. Integrating CI/CD pipelines for NLP models on cloud infrastructure accelerated insights and improved model accuracy. Automated data preprocessing and robust governance mechanisms ensured reliable and trustworthy analytics. Business outcomes included: • Analysis time reduced from weeks to near real time • Increased accuracy and reliability of marketing insights • Improved scalability and responsiveness to changing market demands Implementing XOps successfully at an enterprise level requires more than technology and talent—it demands a structured, strategic approach that aligns clearly with business objectives and operational realities. Please keep in mind that both engineering and machine learning/data science teams need to get aligned so that both learn the ways of working. Begin by combining deep data science expertise, domain-specific experience and MLOps proficiency. Assemble cross-functional teams, including data scientists, domain experts and solution architects who champion end-to-end ML lifecycle management and are familiar with industry-specific use cases, particularly in areas like consumer packaged goods (CPG). Move from basic DevOps to full-scale automated MLOps by setting up structured automation stages: • Automated data gathering and version control • Automated training with robust monitoring and model evaluation • CI/CD-driven automated deployment with infrastructure as code (IaC) • Automated retraining to sustain model performance and interpretability Address the critical gaps between model development and operational deployment by ensuring continuous governance, standardized metrics and integrated training processes. Focus not just on sophisticated models but on building reliable, repeatable processes that enable smooth transitions from development to production. Establish comprehensive ML operational excellence by implementing the following: • Version control for traceability • CI/CD pipelines for streamlined deployment • Infrastructure-as-code for reproducible infrastructure • Model monitoring to proactively address degradation • Automated model deployment to minimize manual intervention • Data operations to ensure data traceability and integrity To execute a planned production approach effectively, it is essential to begin with a thorough understanding of the data and models involved. Next, refactor the code to ensure scalability in a production environment. Develop automated pipelines to standardize workflows and maintain consistency. Finally, implement deployment strategies that incorporate seamless monitoring, allowing models to adapt dynamically to real-world conditions. The transition should be smooth from concept to engineering, as this process involves serious change management. To effectively scale an AI application from concept to production, we follow a structured, iterative process encompassing multiple clearly defined stages: • Define: Begin by collaborating closely with business consultants and SMEs to articulate the business questions, objectives and requirements. • Design: Proceed with comprehensive data acquisition and preparation. This stage ensures the data quality is robust and suitable for further modeling. • Describe: Implement feature engineering, model training and experimentation. Evaluate and compare models meticulously to select the optimal approach for deployment. • Deploy: Integrate models into user-centric applications via intuitive UI/UX designs, dashboards or web and mobile applications. The deployment also includes ensuring data schema alignment and leveraging granular-level model optimization using approaches like GraphRAG. • Drive: Continuously monitor and track model performance in production. Incorporate consumer feedback for ongoing model refinement and improvisations, fostering a responsive and adaptive model lifecycle that aligns with real-world performance and consumer expectations. Looking ahead, success in 2025 and beyond hinges on effectively integrating predictive, generative and autonomous agent capabilities. The XOps approach, rooted in structured operational excellence and proactive governance, positions businesses for sustained leadership. Organizations must move beyond isolated AI initiatives toward scalable, governed ecosystems that continuously evolve, shaping their industries and setting new standards for operational excellence and innovation. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
