A Simple Definition of What Is, and What Is Not, an Agentic Application
Significant advances in generative AI have driven lots of organizations to experiment with the technology. This mostly takes the form of companies either building their own agentic applications, or adopting agentic applications that have been built by a third-party.
A simple definition of agentic application
But what exactly is an “agentic application”? In simple terms, I would argue that an agentic application is any application that makes use of a Large Language Model (LLM) under the hood.
What these applications do, how they work with machine learning, and how they leverage LLMs for business process optimization are all things that we, collectively, are figuring out. Most people agree that GenAI is going to change the world, but the technology is so early, and developing so fast, that we can only speculate and experiment to figure out what that future looks like.
A controversial definition of agentic application
That simple definition is not uncontroversial. Most commentators have instead defined agentic applications in terms of their “autonomy” or their “agency”.
Research on agentic systems existed before LLMs. In 1995, Wooldridge and Jennings described two types of AI agents¹. Basic agents that can work independently, interact with other agents, respond to changes, and work toward goals. And advanced agents, which have the same abilities plus human-like traits such as beliefs, desires, and knowledge.
Critics of the LLM-based definition also point out that it is possible to use LLMs in applications in ways that do not appear to be “agentic”. (You can, if you wish, use ChatGPT as a calculator - if you do, this does not look like an “agentic application” - it looks like an old fashioned calculator, even if there is an LLM under the hood.)
But a useful definition
In spite of those objections, I would argue that defining agentic applications in terms of systems that use LLMs is still a useful definition.
Whilst work on agentic systems pre-dates LLMs, LLMs have transformed what agentic applications are able to do. Previous techniques in AI, including both supervised and unsupervised learning, meant that applications could be written to solve very tightly defined problems like classification - is this user a human or a robot? Is this transaction fraudulent or not? Which customer segment does this person belong to? What customers does this customer look most similar to?
Now with LLMs it is possible to write agentic applications that can:
- Develop hypotheses about the user, or about her behavior, that are worthy of further investigation, using sources such as customer service interactions
- Create new content and/or experiences for the user based on decisions from feedback loops
- Experiment with delivering that content, measuring the result
- Learn and iterate
This is an enormous expansion in capabilities - well beyond what agentic systems were capable of previously.
Another observation is that the architectures of applications that use LLMs are being modeled, more and more, on how human cognition works. Modern agentic AI systems are often architected as multi-agent systems, with different agents using LLMs to perform specific functions, which are coordinated (often by one of the agents in a “conductor” role) so that they’re working to a particular end goal.
Each agent is conceptualized as an individual with goals, memory, some form of perception, and ability to make decisions and to adapt and learn. The high level architectures for these agents are even referred to as “cognitive designs”.
Why agentic applications will be fueled by real-time customer context
Our customers are pushing the boundaries of innovation in AI. They’re building agentic applications that proactively assist users with complex tasks, from managing finances and health to planning trips and conducting research.
But when we speak with product and AI leaders tasked with accelerating their agentic roadmap, we hear a common blocker. It’s not in model training, but in building the data infrastructure needed to reliably feed those models at scale. Without that foundation, agentic applications hit the “cold start problem” because they lack sufficient user context to act in real time.
So we’ve extended our expertise in real-time event data collection, processing, and delivery to design a new real-time customer context solution for product, engineering, and data science teams building customer-facing agentic applications.
Snowplow Signals computes real-time customer context from your behavioral data and serves it to your applications and agents over an API. Its Agentic Context feature gives an agent a live record of what the user has done in the current session, as JSON or a plain-language narrative. It’s designed to be developer-first, not an opinionated black box. So you stay in control of your data, logic, and models while Signals handles the heavy lifting of computation and delivery in real time.
Start Signals for free: add your own events, define an attribute, and read it from your app or agent, with no credit card or sales call. Or book a demo to see how Snowplow Signals can help you accelerate your agentic applications roadmap.
¹ Wooldridge and Jennings, Intelligent Agents: Theory and Practice (1995)
Agentic application FAQs
What does agentic mean?
Agentic describes something that acts with agency: it pursues a goal, makes decisions, and takes actions instead of only responding to instructions. The word comes from psychology, where agency means the capacity to act independently, and it is a real, if technical, word. In AI, agentic usually describes systems that use a large language model to plan steps, call tools, and adjust based on the results.
Is an agentic application the same as an AI agent?
No: an AI agent is a component, while an agentic application is the product people actually use, built on one or more LLMs or agents. An agent is an LLM-driven unit with a goal, memory, tools, and the ability to choose its next step. The application wraps agents in a user interface, data, and business logic, and often has one agent coordinating several others toward a shared goal.
What are some examples of agentic applications?
Common examples include support assistants that resolve a customer's issue instead of only answering questions, shopping and travel assistants that compare options and build a cart or itinerary, in-app copilots that suggest a user's next best action, and research agents that gather and summarize sources. What they share is an LLM deciding what to do next based on the goal and on what it knows about the user.
Can an agentic application run on batch data, or does it need real-time data?
An agentic application can run on batch data, but customer-facing agents work best with real-time data, because batch profiles miss what the user is doing in the current session. A daily summary tells an agent who a customer is; only live events tell it the customer is stuck at checkout right now. Snowplow Signals handles both, computing attributes from the event stream as events arrive and syncing modeled data, such as CRM attributes, from your warehouse.
How do you give an AI agent context about the customer?
Give the agent a current, structured view of who the user is and what they have just done, either in its prompt or through a tool it can call. In Snowplow Signals, an Agentic Context captures the events you choose from the user's current session and returns them as JSON or a plain-language narrative. Your application can pass it to the agent with other attributes, or the agent can read both through the Signals MCP Server.
How do you track what an AI agent does in your application?
Track an agentic application in three layers: client-side events for what the user did, server-side events for what the agent did, such as LLM calls and tool use, and self-tracking events where the agent records its own reasoning and decisions. Snowplow trackers can capture all three, so agent activity lands in your warehouse alongside the rest of your behavioral data and you can analyze both together.