Agentic AI: A deep dive into the future of automation

Beyond generative AI

The most transformative promise of AI has always been its potential for autonomy, to create systems that can act intelligently on their own without human supervision. However, this kind of “Agentic AI” has remained out of reach for most enterprise use cases, until now.

Across industries, two related trends will change our perception of what is possible over the next year and a half, according to Sam Witteveen, CEO of Red Dragon AI, an AI agent-focused consultancy:

• Agents in everything: AI agent-embedded alternatives to many familiar software tools and services will become available, allowing users to interact with them in natural language instead of using specialized interfaces or code.
• Building blocks for agents: A new generation of tools and frameworks for building custom AI agents is arriving, which will allow businesses to develop AI-driven strategies for different facets of their operations.

This article is part one of a multi-article deep dive into Agentic AI, which promises to be the next evolutionary phase of AI adoption for enterprises across industries. Over the coming weeks, this series will explore the full impact of Agentic AI on how organizations of the future will function, including cybersecurity, IT administration, business operations, sales, marketing and more. We’ll also explore the evolving ethical and regulatory landscape to help you stay oriented.

Since ChatGPT burst onto the scene, enterprises across the spectrum of industries have been swarming to integrate generative AI into their products, from image generation to enhanced customer service bots. Companies have adopted these products in areas ranging from content marketing to software development to threat detection, with a Google Cloud study showing 70% of companies had seen ROI on at least one use case. This impact will grow as solutions mature. According to a recent McKinsey report, generative AI technologies will add between $2.6 trillion to $4.4 trillion of value across business sectors, and reduce the total amount of work required by all employees by 50%-70%.

However, another wave of innovation is on the horizon—one that promises to do much more than produce captivating visuals or human-like text. Agentic AI is poised to revolutionize the very core of how enterprises function, as applications arrive that can autonomously monitor events, make decisions and take real actions, all on their own. It is now time to look beyond the chatbots and content generators that have dominated headlines so far. From embedded agents managing cybersecurity threats in real-time to marketing AIs autonomously generating hyper-personalized campaigns, Agentic AI is not only a technical advancement but a true paradigm shift that will have profound effects on enterprises and society.

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, Intuit and Asana. Seats are almost gone—apply for your spot today!

Defining Agentic AI: generative AI fused with classical automation

Agentic AI combines classical automation with the power of modern large language models (LLMs), using the latter to simulate human decision-making, analysis and creative content. The idea of automated systems that can act is not new, and even a classical thermostat that can turn the heat and AC on and off when it gets too cold or hot is a simple kind of  “smart” automation.

In the modern era, IT automation has been revolutionized by self-monitoring, self-healing and auto-scaling technologies like Docker, Kubernetes and Terraform which encapsulate the principles of cybernetic self-regulation, a kind of agentic intelligence. These systems vastly simplify the work of IT operations, allowing an operator to declare (in code) the desired end-state of a system and then automatically align reality with desire—rather than the operator having to perform a long sequence of commands to make changes and check results.

However powerful, this kind of classical automation still requires expert engineers to configure and operate the tools using code. Engineers must foresee possible situations and write scripts to capture logic and API calls that would be required. Agentic AI transcends these limitations in two radical ways: First, anyone who can use language can interact with the system, instead of access being limited to trained coders. Second, static scripts are replaced with LLM-generated code-on-demand to fit the unique situation.

In this new paradigm, intelligent AI agents can be assigned broad objectives or success criteria simply by describing them in language. These agents are then allowed to loop through cycles of assessing what needs to be done, validating what they’ve achieved so far, and deciding on the next steps toward the final objective–roughly what a human would do to solve the problem.

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 or submitting payment transactions–not just finding you a nearby pizza restaurant, but actually ordering for you, as shown in this demo.

In financial services, for instance, AI agents can continuously monitor markets, automatically execute trades or adjust investment strategies based on real-time analysis. These systems can process far more data than any human, potentially allowing businesses to operate with increased efficiency, reduced risk and improved decision-making.

The following set of properties generally define Agentic AI systems:

• Generation: Modern Agentic AI systems harness the analytic and creative capacity of LLMs. Unlike simple gen AI apps, however, they don’t simply output a generated text back to the user as a result. Instead, they can use generated outputs as intermediate steps within a complex workflow, mimicking the role of human thought.
• Tool Calling: In agentic systems, AI can call upon specific tools or APIs, querying data and triggering events according to the reasoning generated by the LLM.
• Discovery: Agentic systems can access real-world data from a variety of tools and data streams, escaping the limitations of their training data. Further, they can harness LLM generation to decide what data they need and to ask for it, rather than being limited to human-provided input, as in retrieval-augmented generation (RAG). For example, an AI agent tasked with maintaining supply chain logistics might write its own queries to weather data APIs and supplier inventory databases, to predict shortfalls and determine possible solutions.
• Execution: Agents can take real-world actions, such as interacting with external systems or triggering processes, without human intervention. An AI agent might send emails or other communications to humans, send purchase orders or fund transfers, grant or revoke access to secure systems, or take any action that can be connected to an API.
• Autonomy (Self-prompting): Agentic systems are “always on;” they do not need to be triggered to do a specific thing at a specific time, the way a simple chatbot can only respond to a prompt. Instead, once active they can monitor for the right moment to act, relieving humans from this kind of “watch and wait” labor. They can loop through cycles of acting, evaluating and planning, continually ‘self-prompting’ to proceed toward a desired end-state.
• Planning: Agentic systems can generate, prioritize and manage sets of subordinate tasks to pursue an overall goal.
• Composition: Agentic systems can assemble multiple components—such as queries, scripts or subroutines, calls to APIs or remote functions, into a cohesive action or response. Unlike a script in traditional automation, an AI agent composes a unique solution to a specific problem, using an LLM to reason out how to combine the available resources. This can include delegating work to other AI agents, either by creating them on demand or by communicating across a service boundary.
• Memory: Agentic systems can build and maintain their own internal knowledge representations, allowing them to accumulate and utilize information extracted through discovery, and the output of previous actions. This capacity enables agents to function more autonomously, as they can index, store, and retrieve information about the world for use in further tasks. For example, a personal shopper agent for a retail website might maintain an idiosyncratic list of themes and facts about a user extracted from their chat interactions and purchase behavior, and use it to customize both conversation and recommendations.
• Reflection: Agentic systems can evaluate the solutions they generate and try again if necessary, rather than delivering low-quality results. For instance, a marketing agent that generates user-customized campaign copy through a multi-step, retrieval-assisted process, might submit all documents to an evaluator AI that predicts the user’s ratings and critical feedback, ensuring that customers only encounter the best possible results.

Diagram: Agentic systems can access tools for discovery and execution,and can plan goals to achieve real-world events.

Transforming enterprises

The implications of agentic AI are enormous, complex and dynamic. Organizations in every sector must prepare to adapt.

AI agents are still under development, and the technology faces challenges as it matures. It depends at its core on LLMs, which are still prone to hallucination. If an agent does a web search for specific links, for example, it might bring slightly wrong backlinks. And that LLM might not know what to do with it, and find itself in an endless loop, running up costs for the agent’s human creator as it consumes more and more tokens. But at the same time, developers have flocked to experiment with, and improve, these agents. Over time, smart design will prevail as engineers learn to combine the agentic components into robust systems.

Three main agent frameworks have emerged as particularly popular: Langraph, Autogen and CrewAI. One review found them roughly equal, though each has its advantages and disadvantages. Over the next few weeks, this series of articles will consider use cases in a variety of industries, reviewing leading product offerings for off-the-shelf AI agents, as well as considering the kind of projects that companies are building now with these DIY tools and frameworks.

Here are just a few examples of how agentic AI is already having an impact:

1. Sales: Next-Generation Lead Management

Agentic AI is revolutionizing the sales process by automating entire pipelines, allowing businesses to scale lead management like never before. Tools like Conversica and Relevance AI are already offering AI-powered assistants that autonomously engage with potential leads, qualify them and nurture prospects through the sales funnel. Conversica, for instance, uses AI-driven Revenue Digital Assistants to initiate conversations, answer inquiries and schedule follow-ups across email and SMS. These assistants ensure no lead is neglected, helping businesses achieve up to a 5x increase in qualified sales opportunities by ensuring timely, personalized interactions.

Similarly, Relevance AI provides AI agents like their AI Sales Development Representatives (SDRs), which automate repetitive tasks like lead qualification and follow-up. These AI agents analyze lead behavior in real time, scoring and prioritizing them for human sales reps to focus on high-value opportunities​.

The ability to personalize at scale is a game-changer for sales teams, allowing human representatives to focus their time on high-value prospects while AI agents handle routine customer engagement. In fact, a Gartner report suggests that by 2025, 75% of B2B sales organizations will augment their teams with AI-driven agents to automate routine tasks and improve overall productivity.

2. Marketing: Hyper-Personalized Shopping at Scale

Agentic AI is transforming how businesses personalize customer interactions, with tools like Netcore’s Co-Marketer AI and Salesforce’s Agentforce leading the charge. Co-Marketer AI empowers brands to engage users across multiple channels, such as email, WhatsApp and SMS, by offering dynamic, personalized content based on real-time data. This AI-driven platform continuously learns from user behavior, allowing brands to deliver highly relevant recommendations and offers that adapt to individual customer journeys, significantly boosting engagement and conversions.

Salesforce’s Agentforce uses AI agents to autonomously craft and optimize personalized marketing campaigns. These agents analyze customer data, such as past purchases and browsing history, to generate tailored campaigns and offers at scale. By automating these processes, businesses can focus on higher-level strategy while ensuring customers receive highly personalized, relevant content across every touchpoint, driving deeper customer relationships and increased revenue growth.

Both platforms showcase the power of agentic AI to deliver hyper-personalized, scalable marketing solutions that elevate customer engagement to new heights.

3. Cybersecurity: Real-Time Defense

Cybersecurity is one of the most obvious applications of agentic AI, where speed and accuracy are paramount. In this space, companies like Darktrace and Vectra AI have developed AI-driven agents that continuously monitor network traffic, identify threats and autonomously initiate responses.

Vectra AI uses AI-driven agents to autonomously detect and respond to security incidents across cloud, data center and enterprise networks. Vectra’s agents continuously monitor network traffic, learning the patterns of legitimate behavior to better identify anomalies that could signal an attack. Once a potential threat is detected, the AI agents autonomously initiate the response—whether it’s isolating the compromised segment of the network, blocking malicious traffic or quarantining affected systems.

The shift to agentic AI will allow security teams to operate more effectively, handling threats in real-time without human intervention. This always-on, autonomous defense could be the key to preventing breaches and minimizing damage from cyberattacks, allowing businesses to operate securely in an increasingly digital world.

4. Infrastructure and IT Operations: Proactive Management

Managing IT infrastructure has traditionally involved a significant amount of manual oversight, configuration, and constant monitoring. However, with the rise of platforms like Qovery, the future of IT operations is becoming increasingly autonomous, leveraging agentic AI to transform how businesses manage their infrastructure.

Qovery’s platform offers a glimpse into how agentic AI can reshape IT operations. Designed to automate the deployment of applications in the cloud, Qovery’s agents perform tasks such as setting up environments, managing scaling and ensuring uptime through self-healing systems.

This is not just an extension of traditional IT automation tools like Kubernetes or Terraform—Qovery’s AI agents act with higher-level decision-making capabilities. For instance, they can anticipate application needs, dynamically adjust environments, and even optimize costs by reallocating resources, all while requiring minimal human input.

AI agents interpret user commands in natural language, reducing the need for companies to maintain expertise in IT management. Qovery claims its platform “eliminates your DevOps hiring needs.”

What’s Next?

AI agents can empower businesses to operate with greater efficiency, agility and speed. This technology is in its early days, but as more robust offerings become available–and this is expected to happen very quickly–the business case for its adoption will grow.

However, the implementation of agentic AI requires thoughtful design, as these systems will not be one-size-fits-all. Specialized AI agents will need to be created for some jobs, and the right AI-enabled tool chosen for others. Whether developing their own or deploying third-party agentic AI, enterprises will need to understand the hype and reality, the promise and peril, of this new technology.

Throughout this series, we will explore how enterprises can build these systems, the tools and platforms they can use and the industries that are poised to benefit most from the rise of agentic AI.  We will take a closer look at how agentic AI is reshaping marketing, sales, cybersecurity, customer service and business operations. We’ll also explore the emerging regulatory landscape and how using sound principles of AI governance can help you maintain the trust of your users and partners while forging your path ahead. Stay tuned for the future of AI-driven business.