Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More
Agentic artificial intelligence (AI) represents the next frontier in AI, offering a bold leap beyond traditional IT automation by fusing it with the creative power of generative AI. In this paradigm, AI agents can evaluate data in real time, make decisions about novel scenarios and execute actions with real-world consequences.
The next level of automation
Generative AI has revolutionized how businesses approach text and image creation, allowing this work to be done orders of magnitude more quickly. Agentic AI applies its efficiency and speed to more complex work, such as task planning, decision-making and executing commands or function calls to external tools.
In this new paradigm, intelligent AI agents can be assigned broad objectives or success criteria. They are then allowed to loop through cycles of assessing what they’ve achieved so far and deciding on the next steps toward the final objective. AI agents can also interact with external tools or APIs, querying data from external sources and triggering real-world actions. This can include sending communications, making API calls, or submitting payment transactions–not just finding you a nearby pizza restaurant, but actually ordering for you, as shown in this demo.
The potential business value of AI that can make autonomous decisions and act as needed is immense. Consider the massive CrowdStrike outage earlier this year, illustrating the need for intelligent automation in enterprise systems.
On July 19, an automated configuration update fault caused widespread system crashes, affecting approximately 8.5 million Windows devices globally. The resulting chaos cost billions of dollars across industries such as finance, healthcare and transportation.
While security updates must be frequent and fast—making automation essential—they need not be blind. In the future, AI agents embedded in the installation process could monitor a rollout for early signs of failure and potentially halt the deployment. As enterprises push for greater automation, integrating AI agents capable of adapting to unexpected challenges will be crucial for mitigating risk.
Join us in San Francisco today (Sept. 10) for the AI Impact Tour, where we’ll dive into Agentic AI’s future. Don’t miss this opportunity to learn from industry leaders like Meta, Anthropic and Asana. Seats are almost gone—apply for your spot today!
Navigating safety and ethics in Agentic AI
AI systems are not infallible; they often struggle with edge cases, leading to misclassifications and unexpected, sometimes problematic, results. The prospect of trusting critical decisions to a system that can “hallucinate” is understandably concerning. As we stand on the brink of widespread enterprise adoption of agentic AI, the challenge lies in balancing its incredible potential with the inherent risks. Organizations must navigate this complex landscape carefully to stay competitive without compromising security or trust.
Key Considerations for Ethical AI Deployment:
Bias Mitigation: AI systems are only as reliable as the data they are trained on. Mitigating bias is crucial to preventing unfair or discriminatory outcomes and maintaining stakeholder trust.
Transparency: Enterprises must communicate clearly about what responsibilities are being delegated to AI, and about how risk is being assessed.
Accountability: Decision-makers must ensure robust governance frameworks are in place to oversee AI deployments and mitigate risks.
Regulatory Compliance: Staying compliant with emerging AI regulations is essential to avoid legal repercussions and maintain a competitive edge.
Frameworks such as the IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems and the EU’s AI Act provide a foundation for ethical AI deployment. Enterprises should adopt a cross-functional approach to AI ethics, involving stakeholders across the organization in decisions, and investing in training and discussion to foster a culture of responsible AI use.
Leading Agentic AI Solutions
Agentic AI solutions range from open-source experimental prototypes to early enterprise-ready offerings. Here are three options that decision-makers should consider exploring.
AutoGPT: The cutting-edge pioneer
AutoGPT is an extensible, open-source implementation of agentic AI, offering a glimpse into the future of autonomous IT systems. Launching in the summer of 2023, it pioneered using third-party LLMs to reason through the execution of real-world tasks. Already, a growing open-source community has contributed an impressive repertoire of plugins.
Key Features:
Memory: AutoGPT extends the capabilities of LLMs like GPT-4 with an embedded backend datastore, allowing agents to break down complex objectives into a prioritized sequence of sub-tasks without losing track of previous results or the overall flow.
Self-Prompting: AutoGPT generates its own prompts based on the outcomes of previous tasks, enabling it to navigate iterative workflows and evaluate its own results.
Tool and API Integration: The model interacts with external tools and APIs, allowing it to fetch data, manage databases, or trigger actions in other systems.
Iterative Learning: AutoGPT operates in a loop, continuously evaluating and refining its actions to better meet its goals, making it effective for long-term, complex projects.
Open-Source Flexibility: As an open-source project, it allows a community of developers to customize and expand its capabilities.
Draw-backs:
DIY Approach: As an open-source project, AutoGPT is not an enterprise-grade product off the shelf. Building on it requires taking full responsibility for the results and investing significant research and development time.
Reliability: AutoGPT is still experimental and relies on embedded third-party LLM models, making its outputs unpredictable, especially in edge cases or tasks requiring high precision.
Scalability: While AutoGPT includes tools for benchmarking agent performance, scaling it to enterprise levels would require significant investment in development and testing.
IBM Watson Orchestrate: Enterprise-ready automation
IBM Watson Orchestrate offers an enterprise-friendly, low-code platform designed to help non-technical business users automate tasks and workflows with AI-driven agents (called assistants). Assistants can access dozens of data, communication, financial, and other service integrations via pre-designed skills, and users can define additional custom skills.
Strengths:
Low-Code Development: Watson Orchestrate’s low-code interface allows business users to build AI assistants without needing advanced programming knowledge, democratizing AI across departments.
Pre-Built Skills and Automation: The platform offers thousands of pre-built skills that can be customized and extended, enabling businesses to quickly automate common tasks and processes.
Enterprise-Grade Reliability: Built with enterprise needs in mind, Watson Orchestrate offers robust security, compliance, and integration with other enterprise applications like CRM, HRM, and procurement systems.
Contextual Learning and Memory: The platform applies contextual learning, enabling AI assistants to remember previous interactions and improve performance over time, making it valuable for tasks requiring ongoing adaptation and optimization.
Draw-back:
High Cost and inflexible pricing: As an enterprise-grade solution, Watson Orchestrate’s cost can be prohibitive and subscription options are limited. Currently, Orchestrate is only available as a service through AWS, with two 12-month subscription options, $18,000 and $54,000.
Microsoft’s Semantic Kernel: Modular, scalable and integrated
Microsoft’s Semantic Kernel is a comprehensive framework designed to integrate LLMs with traditional programming environments, offering a modular and extensible architecture supported by Microsoft’s enterprise ecosystem.
Strengths:
Modular Design: Semantic Kernel allows developers to integrate existing code and APIs as plugins, enabling AI agents to interact with a wide array of services and data sources.
Intelligent Planning and Execution: The framework’s planning capabilities enable AI agents to generate and execute multi-step plans.
Context and Persona Management: Developers can define an AI agent’s persona, shaping its interactions with users, while the kernel’s context management allows agents to recall and apply previous interactions, enhancing response relevance.
Microsoft Integration: Semantic Kernel is designed to work seamlessly with Microsoft Azure and the .NET ecosystem, including Azure AI studio, ensuring it can be deployed at scale within enterprise environments while maintaining high security and compliance standards.
Draw-backs:
Complex Setup: While offering extensive capabilities, Semantic Kernel requires significant development work to set up and optimize, particularly for businesses not deeply embedded in Microsoft’s ecosystem.
Reliance on Azure: Full integration and scaling are best achieved within Microsoft’s Azure environment, which may not suit every organization’s infrastructure.
Comparative Summary
Each platform offers distinct advantages depending on the specific needs and maturity of an enterprise’s AI strategy:
AutoGPT is ideal for organizations and researchers exploring the frontiers of autonomous AI. Its ability to self-prompt and integrate with various tools makes it powerful for automating complex, multi-step tasks, but it requires careful management and, without major investment in development, is best suited for smaller-scale or non-critical applications.
IBM Watson Orchestrate is a polished offering with a high price tag. Its focus on business process automation and ease of use makes it an ideal choice for enterprises looking to implement agentic AI quickly and effectively, especially with non-technical users.
Semantic Kernel is a versatile and scalable framework perfect for enterprises that need to integrate AI into existing Microsoft infrastructure. Its modular design and robust planning capabilities make it suitable for automating complex workflows while maintaining a high degree of control and customization.
Embracing the Future of Agentic AI
The rapid advancement of agentic AI presents both opportunities and challenges. Leaders must responsibly manage security and ethical challenges, but organizations can unlock the full potential of intelligent agents by understanding the landscape and adopting the right frameworks, tools and technologies. AutoGPT, Microsoft’s Semantic Kernel and IBM Watson Orchestrate offer early glimpses into the exciting future of this trend.
VB Daily
Stay in the know! Get the latest news in your inbox daily
By subscribing, you agree to VentureBeat’s Terms of Service.
Thanks for subscribing. Check out more VB newsletters here.
An error occured.
{Categories} _Category: Takes{/Categories}
{URL}https://venturebeat.com/ai/agentic-ai-a-deep-dive-into-the-future-of-automation/{/URL}
{Author}Aliyah Mohammed{/Author}
{Image}https://venturebeat.com/wp-content/uploads/2024/09/agentic-AI-Dalle.png?w=1024?w=1200&strip=all{/Image}
{Keywords}AI,Data Infrastructure,Agentic AI,ai,AI Impact Tour,AI, ML and Deep Learning,Anthropic,API,Asana,AutoGPT,autonomous,category-/Business & Industrial,category-/Computers & Electronics/Programming,category-/Computers & Electronics/Software,category-/Science/Computer Science,crowdstrike,data,Enterprise,enterprise ai,ethical AI deployment,EU AI Act,gen AI,Generative AI,GPT-4,IBM Watson Orchestrate,IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems,LLM-based AI agents,Meta,Microsoft Semantic Kernel,mitigating bias,VentureBeat AI Impact Tour{/Keywords}
{Source}POV{/Source}
{Thumb}{/Thumb}
