Agentic AI: A new frontier for artificial intelligence.
Artificial intelligence is becoming a core part of the way businesses operate, and the organizations that succeed with its adoption know that AI is evolving constantly. One advancement in particular is transforming the way businesses use AI to drive value: agentic AI.
Learn about agentic AI.
Agentic AI is the next step in the evolution of artificial intelligence—a shift from systems that primarily analyze, predict, or generate to ones that can take action. While previous versions of AI assist with decision-making or content creation, AI agents can initiate decisions, plan action steps, and execute autonomously.
Gartner® predicts that by 2028, one-third of enterprise software solutions will include agentic AI, making up to 15% of day-to-day decisions autonomous. Across industries, AI agents are already making their mark, helping hospitals manage complex admissions, supporting individually tailored academic plans at universities, optimizing inventory management for retailers, and much more.
At Workday, we build agentic AI to transform processes and deliver better experiences for our internal and external stakeholders.
What is agentic AI?
Agentic AI combines multiple AI models in an orchestrated, integrated way to allow a program to act autonomously within a broader environment. It uses reasoning, learning, and iterative planning to handle dynamic and multistep challenges within an organization.
What sets AI agents apart is their ability to take direct action with little to no human involvement. They can monitor systems, interpret real-time conditions, and initiate tasks across connected applications. For example: an agentic system in a logistics platform may detect a shipping delay, reroute deliveries, notify customers, and automatically update inventory—all on its own.
The journey to agentic AI has been shaped by decades of innovation and breakthroughs in artificial intelligence. From its earliest foundations to its modern, transformative capabilities, AI has continuously evolved over time—and still is evolving—to become more adaptive, autonomous, and impactful.
AI’s evolution through the decades.
Early foundations.
In the 1950s, AI was rules-based—limited to strict if-then instructions and performing basic automation with no ability to adapt or learn. The 1960s and 1970s saw a major breakthrough with neural networks inspired by the human brain, which allowed AI to simulate learning for the first time. Though still in its infancy, this development laid the foundation for more dynamic and flexible AI systems.
The rise of expert systems.
By the 1980s, expert systems emerged, bringing AI closer to human-like decision-making. These systems used large rule sets to replicate expert reasoning, but their reliance on static logic made them inflexible and unable to handle new or evolving information.
Machine learning emerges.
The 1990s marked a turning point with machine learning (ML), shifting AI from a rules-based approach to one that could learn from data. This decade also saw the earliest exploration of AI agents and reinforcement learning, enabling AI to make decisions through trial and error rather than relying solely on predefined instructions.
Big data and cloud computing.
The 2000s accelerated AI’s progress with an explosion in computational power and the rise of cloud computing. AI systems were now able to process and access vast amounts of data in real time, making them more efficient and capable of handling increasingly complex tasks.
The Gen AI breakthrough.
Generative AI (Gen AI) introduced the ability to create original content—text, images, code, audio, and more—based on natural language prompts. It opened the door for interactive, conversational AI that could assist with writing, automate customer support, generate design assets, and more.
Foundational Gen AI research, such as transformer models and generative adversarial networks (GANs), began in the 2010s, but it wasn’t until the launch of models such as GPT-3, DALL·E, and ChatGPT in the early 2020s that Gen AI became widely accessible and practical for real-world use.
These systems brought a new level of fluency and contextual understanding but they still relied on human prompts to operate. Gen AI could suggest and synthesize but it wasn’t built to take action or make independent decisions.
Agentic AI.
Building on the foundation of machine learning and generative models, the latest evolution in AI is agentic AI—systems that not only understand and generate but that also take initiative and execute actions in real-world environments.
This shift marks a fundamental transformation from AI as a tool that responds to humans, to AI as a collaborator that can take action on its own. With this transition, we’re entering a new era of AI-enabled systems—ones that can optimize workflows, make decisions at scale, and dynamically respond to change.
From early rules-based systems to today’s autonomous agents, AI has been on a steady path of increasing capabilities—and that evolution is far from over. Agentic AI is just the latest step, and businesses must stay engaged to keep their AI strategies relevant and effective.
Agentic AI marks a fundamental transformation from AI as a tool that responds to humans, to AI as a collaborator that can take action on its own.
Understanding agents in the broader AI landscape.
As agentic AI gains traction, it’s important to understand where it fits within the broader AI landscape. Rather than existing as a standalone category, agentic AI builds on the capabilities of other systems—drawing from both rules-based logic and probabilistic models to operate with greater autonomy. By looking at how different types of AI process information and drive outcomes, we can better understand how agentic AI extends what’s possible.
Deterministic AI.
Follows predefined logic or rules. Outputs are predictable and repeatable given the same input.
Limitations.
Cannot handle uncertainty or adapt to new scenarios.
Example.
A decision tree that approves or denies loans based on fixed thresholds.
Probabilistic AI.
Uses statistical models and data patterns to generate predictions or outputs.
Limitations.
Requires large high-quality datasets. Outputs may vary and are often not explainable.
Example.
A generative model such as GPT-4 that writes text based on a prompt.
Agentic AI.
Builds on probabilistic models (for example, LLMs and RL) and orchestrates actions across systems—often integrating deterministic components for execution.
Limitations.
Still developing in areas such as seamless multisystem coordination and long-horizon planning.
Example.
A supply chain agent that monitors weather, predicts disruptions, reroutes deliveries, and updates inventory autonomously.
In summary, agentic AI isn’t a standalone category—it’s an evolution that builds on the strengths of other AI systems. Deterministic approaches offer structure and reliability, while probabilistic models bring flexibility and insight. Agentic AI brings them together with the added ability to take action, enabling organizations to be more scalable, adaptive, and responsive.
Agentic AI systems stand out for their ability to reason, plan, and act with a high degree of autonomy.
Key characteristics of agentic AI.
Agentic AI systems stand out for their ability to reason, plan, and act with a high degree of autonomy. Unlike earlier AI systems that require strict logic or predefined workflows, agentic AI interprets intent, evaluates options, and executes decisions on its own—often in complex, unpredictable environments.
These capabilities make agentic AI well suited to business challenges that require both flexibility and initiative. Here’s how those traits show up in practice:
Autonomous reasoning.
An agent can take a business objective—for example, reducing delivery delays—and independently determine how to achieve it by analyzing constraints, evaluating trade-offs, and initiating corrective actions.
Real-time adaptability.
Instead of following a static script, an agent adjusts its behavior based on what’s happening in the moment, rerouting logistics when conditions shift or reallocating staff in response to unexpected demand.
Multistep task execution.
Agents don’t just trigger single actions—they manage workflows. A single agent might identify a problem, gather inputs from other systems or agents, decide on a solution, and follow through to resolution.
Collaborative orchestration.
Agentic systems are designed to work together. One agent may detect an issue, while others handle communications, inventory updates, or policy checks, each contributing to a shared outcome without direct human coordination.
AI agents exist on a spectrum of complexity—from simple rule followers to fully autonomous systems that reason, learn, and collaborate.
Types of AI agents.
AI agents exist on a spectrum of complexity—from simple rule followers to fully autonomous systems that reason, learn, and collaborate. As you move along this continuum, agents gain greater decision-making power, adaptability, and independence.
Understanding the distinctions between these types is key to choosing the right approach for your business needs. The following categories show how agent capabilities evolve and how more advanced agents lay the foundation for truly agentic systems.
Reactive agents.
Reactive agents are the simplest type of AI. Built on rules-based systems, they respond to environmental changes using predefined instructions but they cannot learn or adapt. For example, a virtual assistant that retrieves preprogrammed responses to specific commands, such as turning off smart lights when told, operates as a reactive agent, responding to input without further reasoning or learning.
Model-based agents.
Model-based agents process their surrounding environment through an internal model, enabling them to reason about their actions and make informed decisions. Example: an AI-powered irrigation system that monitors soil moisture, weather forecasts, and crop needs in order to optimize watering schedules.
Goal-based agents.
Goal-based agents work toward achieving specific objectives by evaluating different strategies and adapting their actions to reach a desired outcome. For instance, a navigation app that can plan the fastest route to a destination based on real-time traffic conditions is a goal-based agent.
Utility-based agents.
Utility-based agents focus on maximizing outcomes within a specific domain by weighing multiple factors to determine the best possible result. Consider, for example, a fleet management system that not only plans delivery routes but also factors in fuel efficiency, delivery-time windows, and vehicle maintenance schedules to help optimize operations.
Learning agents.
Learning agents adapt and improve over time by processing new data and experiences. These agents are designed to dynamically adjust their actions and decision-making processes. An automated financial portfolio manager that refines investment strategies based on evolving market trends is an example of a learning agent in action.
Collaborative agents.
Collaborative, or multiagent, systems work together to solve interconnected problems too complex for a single agent to manage. A logistics network that coordinates delivery drones, warehouse robots, and autonomous trucks to optimize the entire supply chain is a collaborative agent.
Business opportunities for agentic AI.
Agentic AI’s dynamic capabilities make it powerful for addressing real business needs in real time. Across diverse industries, it’s proving valuable in many ways—optimizing operations, enhancing employee and customer experiences, and opening up new avenues for growth. Let’s look at some of the most important agentic AI applications already happening in different sectors.
Real-world examples: AI use cases by industry.
Higher education.
In higher education, agentic AI supports students and faculty by autonomously managing personalized academic planning. For example, an agent can detect when a student is at risk of falling behind; evaluate course availability, program requirements, and the student’s long-term goals; and then propose a revised schedule. It can notify the student and advisor, recommend support services, and adjust the degree path—all without manual coordination or intervention.
Healthcare.
In healthcare, agentic AI acts as a real-time collaborator within clinical environments. An agent might detect a surge in emergency room admissions, assess patient needs, check resource availability (such as ICU beds or staff), and initiate resource reallocation or call additional staff. It may also coordinate with other agents overseeing patient discharge or inventory management to help maintain continuity of care.
Retail.
In retail, agentic AI enables dynamic, end-to-end management of both operations and customer experience. An agent might notice a spike in demand for a product, forecast stock shortages, and reroute existing shipments or trigger new purchase orders. At the same time, it can adjust promotional strategies, update website content, and notify support teams as needed.
Manufacturing.
Agentic AI enhances manufacturing by responding in real time to disruptions or inefficiencies. For instance, an agent might detect performance drops from machinery, initiate diagnostics through another agent, and schedule maintenance during a downtime window. It can also adjust production schedules or reorder materials to avoid delays—keeping processes running efficiently and proactively.
Finance.
In finance, agentic AI supports autonomous, adaptive planning and risk management. A finance agent might recognize early signs of market volatility, adjust portfolio strategies, or flag areas of exposure. It can also align its actions with compliance agents to ensure all activity meets regulatory requirements—minimizing risk without sacrificing speed or responsiveness.
Communications.
In the communications sector, agentic AI helps manage engagement in real time. An agent might detect negative sentiment trending on social media, launch targeted response campaigns, and shift ad spend or messaging strategy accordingly. It can coordinate with agents managing customer support, CRM, or email marketing tools to ensure a consistent and timely response.
Enterprise solutions: Workday examples.
At Workday, we are exploring AI agents to help streamline expenses, optimize succession planning, and transform recruiting.
Our expense agent automatically itemizes receipts and creates expense reports. When an employee checks out of a hotel, for example, they can take a photo of their receipt with their phone, and our agent will automatically extract relevant information from the receipt, create a new expense line item, and add it to the correct expense report.
For succession planning, our AI agent helps us maintain a strong pipeline of successors across career stages by analyzing factors such as current business needs, required team skills, predicted attrition rates, and suggest potential successors across the country. The agent can proactively identify high-potential employees and even generate personalized growth plans to help them prepare for future roles.
Our recruiting agent incorporates HiredScore* capabilities to source passive candidates who may have expressed interest in the past. By understanding the current needs of the business and analyzing candidate profiles, the agent can automate outreach, recommend top candidates, and even schedule interviews.
*HiredScore is a Workday company.
At Workday, agentic AI is transforming areas such as expense tracking, succession planning, and smart recruiting.
Strategies for agentic AI deployment.
The use cases and examples covered in the last section are just some of the countless ways companies are using agentic AI to drive value and achieve goals. Deploying agentic AI successfully means identifying specific ways your business should be using the technology, defining your goals, putting a strong execution plan in place, and monitoring your progress over time.
Let’s walk through a step-by-step framework for effectively implementing agentic AI into your business.
1. Discover and define the problem.
Start by clearly articulating the business challenge you want to address. Go beyond listing features or tasks—focus on the underlying problem. For example, rather than just automating expense approvals, aim to ensure compliance while reducing processing time and errors.
2. Understand user needs and concerns.
Gather continuous feedback from users to identify pain points and opportunities for improvement. Test prototypes with end users and observe their interactions with the technology. Note moments of joy and challenges and adjust accordingly. Continue to do this at every stage of deployment to ensure your AI aligns with actual workflows and creates moments of ease or joy.
3. Develop a data strategy.
Effective agentic AI systems depend on quality data. Develop a strategy for collecting, cleaning, and labeling data while ensuring compliance with privacy and security standards. Thoughtful preparation prevents bottlenecks down the road during training and deployment.
4. Choose the right tools and models.
Next, select the tools and technologies that align with the needs you’ve identified in step 1. Consider the different types of agentic AI we covered earlier, and keep in mind that agentic AI may not always be the right answer; for example, rules-based systems might work better for strictly automation-focused initiatives.
5. Prioritize privacy and security.
Understanding data privacy in the context of new AI systems is paramount. Protect sensitive data by building robust privacy and security mechanisms into your systems, and ensure compliance with relevant regulations such as GDPR or CCPA. Regularly audit data access controls, encrypt sensitive information, and implement automated alerts for unusual activity. Integrate privacy-by-design principles throughout the development process to safeguard both individual and business information at every stage.
6. Acknowledge, assess, and mitigate risks.
Agentic AI introduces new types of risks alongside new capabilities. Regularly evaluate model performance, data quality, and unintended consequences—not just during deployment but also throughout the AI lifecycle. Just as important, develop clear mitigation strategies, such as escalation protocols, human oversight triggers, or feedback loops, to reduce harm and ensure alignment with your organization’s goals and values.
7. Enable agents to work across systems.
Agentic AI draws strength from its ability to act on data that lives in different parts of the business. Each agent uses the sources it has access to—whether that’s internal platforms, third-party tools, or real-time inputs—to reason, make decisions, and take action.
When agents coordinate, they bring their individual context into a shared effort. One agent might assess employee engagement data while another brings in market trends. Working together, they can inform a workforce plan that reflects both internal needs and external realities—without requiring tightly coupled systems behind the scenes.
8. Test and monitor.
Rigorous testing is essential at every stage of development. Ensure your AI system delivers the intended outcomes and adjust as necessary. Post-deployment, continuously monitor performance and user feedback to identify areas for optimization. Regular updates and retraining ensure the AI remains aligned with evolving needs and challenges.
Agentic AI is an advancement toward distributed, adaptive systems that work in tandem with people to drive progress.
Looking ahead.
Agentic AI is becoming a core capability for organizations navigating complexity and change. But its real promise lies in what comes next: a future where agents collaborate—not just within systems but also across teams, businesses, and industries.
By enabling secure, goal-oriented coordination, agentic AI can help organizations share knowledge, respond faster to disruptions, and solve problems too complex for any single system to manage alone.
As these systems mature, agents will be able to reason across broader contexts, connect previously siloed datasets, and take more informed actions to create entirely new possibilities for enterprise intelligence.
Agentic AI is an advancement toward distributed, adaptive systems that work in tandem with people to drive progress. Businesses that embrace this shift will be better equipped to innovate, scale with confidence, and lead in a fast-moving, interconnected world.
Learn how Workday can help you transform with AI.