AI Agents - From Automation to Autonomous Operations

Definition of AI Agents

To start from a very basic definition, think of an AI agent as your assistant or as a Software Agent

When you ask your assistant to write an email, it works on it, taking it as a goal. When you ask it to write an email, it will write an email for you; it does not play music for you or do anything else. 

Hence, the first thing that can be said about these agents is that they are intelligent. They know what goals they are given and, correspondingly, what actions they need to take to get that work done. 

Purpose of AI Agents

Currently, AI agents can do a wide variety of tasks, too, extremely well. Some of those tasks are: 

• Browsing Web

• Write and Execute Code

• Book a meeting 

• Make a payment,

• Analyse and Query Data, etc. 

We see new advancements in Artificial Intelligence every day, so the current limitations on the list of things that an AI agent cannot do may become doable the very next day. 

Every new advancement in Artificial Intelligence, particularly agent technology, moves us closer to AGI (Artificial General Intelligence). 

Growing Preference for AI Interaction

51% of consumers favor bots over human agents for immediate service, reflecting a reliance on AI for quick support. Additionally, 68% believe chatbots should match the expertise of skilled human agents, emphasizing the need for enhanced AI capabilities

Investment and Future Expectations

64% of CX leaders intend to boost investments in advanced chatbots within the year, while 58% expect their chatbots to become more sophisticated by 2024, indicating confidence in AI agents to enhance customer experience.

Six Types of AI Agents 

1. Simple Reflex Agents

It is more like hardcoding the agent’s behaviour. It works on the condition-action rule, meaning it acts after perceiving the current condition. Agents can neither plan their next move nor learn and improve their reasoning by learning from past experiences.  

Although it is easy to implement and run, it is very inflexible to change. Also, since this category of agents is not equipped with memory do not store any state, they have very limited or no intelligence of their own. 

One example of a Simplex type agent is rule-based chatbots, which have a pre-planned set of responses to the user's queries. 

2. Model-based Reflex Agents

 AI Agents have autonomous reasoning capabilities for intelligent decision-making.  These agents work in a four-step process: 

• Sense: Here, the agents get to know the current state of the situation before taking action. 

• Model: This step makes a view for itself after sensing the current environment. 

• Reason: Based on the above-created model, it now decides how to act before taking action. 

• Act: Here, the agent acts. 

An example of Model Bases Reflex agents is AWS Bedrock, which uses various foundational models to make decisions based on user prompts. Although these models are quick and better at decision-making, they are computationally expensive. 

3. Goal-based Agents

Goal Based Agents differ from the above two as they perceive information from their environment to achieve specific goals.  

They have three parameters that they take into consideration when they work: 

• Current state  

• The end goal is to obtain 

• Set of actions needed to take to reach the goal 

These agents are very effective when deployed to attain a specific goal but may fail for a complex task. 

4. Utility-based Agents

Utility AI Agents are quite advanced as they can assign utility scores to different paths they need to take in scenarios when there is more than one possible path to complete a certain task. 

Consider a scenario when there is an agent designed to do research. However, a certain task has both options: search the web or go through the vector store to complete a sub-process. In this scenario, this agent can add utilities to these separate paths and then decide which one to take to complete that particular task. 

The main advantage of these agents is that they can perform well in various decision-making scenarios. It also learns from previous experiences and accordingly adjusts its decision-making strategy. 

5. Learning agents

Learning agents are types of AI Agents that can learn from past interactions and, with time, improve their performance. These AI agents learn from complex data patterns and may also receive feedback from humans in the loop to adapt accordingly. 

6. Hierarchical Agents

Hierarchical AI agents are similar to how things are hierarchically executed in an organization.  

• The agents in the lower-level hierarchy execute the tasks, and the agents higher above them supervise them. 

• This AI agent type is very good when prioritizing different tasks by assigning the right set of tasks to the right agent. 

Use Cases of AI Agents

1. Personal Virtual Assistants

It is a very popular use case involving AI Agents. They can assist us in various tasks, such as reminding us of important events, planning our day, writing emails, and planning meetings.  

2. AI agents for Healthcare

AI agents in the healthcare industry can make a significant impact. They can make everything possible, from basic tasks such as helping people with their medical queries to remarkable tasks such as helping in drug discoveries. 

Many pharmaceutical companies, like Gilead Sciences and others, have already witnessed the potential of AI Agents. Whereas research used to take years, they have accelerated the whole process, making it all possible within months or even days.  

3. AI agents for Finance as Finance Analyst

Financial institutions can leverage the power of AI agents to help them detect fraud by learning transaction patterns from previous data. They can also use agents to build customer-friendly chatbots that respond faster and more accurately to users’ questions.

Additionally, autonomous financial analysis agents can analyze market trends, assess risks, and provide insights for investment decisions, enhancing overall financial strategy.

4. AI tutor and Researchers in Education and Research

Here, agents can help research by making the entire World Wide Web accessible through Natural Language prompts. This has reduced the time required to manually go through research papers and made the best content accessible for any research. The workflow automation potential of these agents is transforming labour-intensive workflows in educational settings.

Autonomous Customer Agents enhance user experience by providing instant support, resolving queries efficiently, and personalizing interactions based on customer data.

Best Open Source Agents

1. AutoGPT

This is one of the first and highly capable open-source AI agents available. For a given goal, it first creates a set of sub-tasks. It then goes through each of those sub-tasks to get the work done. This process may even divide the sub-task into further sub-tasks based on the complexity of the task. 

It is the Agent dividing a complex task into sub-tasks and iterating through them to complete the work.

2. AutoGen

It is another of those remarkable releases in the field of Autonomous Agents. It allows you to build a Multi-Agent Conversational Framework inside your application for enhanced accuracy and better inference from the LLMs. 

For example, you must build an agent to query your structured database. Using AutoGen, you can pass on the initial query results through various other agents in the middle before finally producing the output to the user. Each agent has a defined task and a role assigned to it. If, at any step, an agent finds the responses unsatisfactory, it sends them back to the previous agent for re-evaluation. 

Levels of AI Agents 

1. Tools (Perception + Action)

2. Reasoning & Decision making

3. Memory + Reflection

4. Generalization & Autonomous Learning

5. Personality (Emotion + Character) and Collaborative Behavior (Multi-agents)

Figure -  Level of Personal LLM Agents

Figure -  Level of AI Agents

Reference:- https://arxiv.org/pdf/2405.06643

Challenges in Adoption for AI Agents

1. Data Dependent

The backbone of any AI agent is the large language model (LLM). Hence, the accuracy of the response and intelligent behaviour of the overall agents directly depend on the richness of the data on which the LLM was trained.  

The Agents may become highly biased if the LLM in the backend is not trained on the right data set. 

2. Limited Understanding of Context - Context Management and Intelligence 

Regarding the currently available open-source AI agents like AutoGPT and others, they have a short-term memory, making it hard for them to hold on to the context in a longer conversation. 

3. Security Concerns

AI Security and LLM software supply chain security is critical aspects for AI Quality and AI assurance and governance aspects 

4.  Trustworthy AI Aspects 

AI agents lack common sense and ethical perspectives; they can easily be made to work toward goals with malicious intent. Considerations regarding accountability, transparency, ethics, responsibility and bias in decision-making are critical in regulated industries, especially healthcare, finance, transportation