The Rise of AI Agents in the Enterprise

AI Agents have become indispensable in modern enterprises, driving efficiency, innovation, and competitive advantage. These agents, which can perform tasks ranging from simple automation to complex decision-making, are transforming how businesses operate. The adoption of AI agents is widespread, with companies leveraging them to enhance customer service, streamline operations, and gain insights from vast amounts of data. According to a recent McKinsey survey, over 80% of enterprises are either using or planning to use AI agents in some capacity, underscoring their critical role in the digital transformation journey.

Types of AI Agents: Declarative vs. Autonomous

AI Agents can be broadly categorized into two types: declarative agents and autonomous agents. Understanding the differences between these types is crucial for businesses looking to implement AI solutions effectively.

Declarative Agents: These agents operate based on predefined rules and logic. They require explicit instructions from a human being to perform tasks and are typically used for specific, well-defined processes. Declarative agents excel in environments where the tasks are repetitive and predictable. For example, a customer service chatbot that answers frequently asked questions based on knowledge sources that it has access to, chat histories, and more, is a declarative agent. These agents are relatively easier to design and implement, making them suitable for businesses that need quick, reliable automation solutions.

Autonomous Agents: In contrast, autonomous agents are designed to operate with a higher degree of independence. They use other types of triggers, like APIs, changes in data, and other actions not necessarily prompted by humans to make decisions and adapt to new situations without human intervention. For instance, an AI-driven recommendation system that personalizes content for users based on their behavior is an autonomous agent. These agents require more sophisticated development and training but offer greater flexibility and potential for innovation.

What both AI Agent types have in common is that they:

1. Can learn from their interactions and improve over time, making them ideal for complex, dynamic environments, offloading manual tasks, and other critical business operations. 
2. Are connected to a variety of data sources, files, and are used for a huge variety of business use cases

Off-the-Shelf AI Agents vs. Custom-Built Solutions

When it comes to implementing AI agents, businesses have two primary options: using off-the-shelf solutions and/or enabling people to build custom agents.

Off-the-Shelf AI Agents: Solutions like Microsoft 365 Copilot and Salesforce Einstein are ready-made AI agents that businesses can integrate across their existing systems. These AI Agents come with pre-built functionalities and are designed to be user-friendly, allowing companies to quickly deploy AI capabilities without extensive development efforts. Microsoft 365 Copilot, for example, assists users with tasks like drafting emails, creating reports, and analyzing data within the M365 suite, and can be connected across a variety of Microsoft applications, systems, and datasets out of the box. Salesforce Einstein provides AI-driven insights and automation within the Salesforce platform, helping businesses enhance their customer relationship management (CRM) processes. Salesforce also unveiled a variety of specialized AI Agents like a customer service agent that can be used to answer customer queries automatically and dynamically. 

Custom-Built AI Agents: Custom-built AI Agents can also be built by business users of any technical background on low-code development platforms like Microsoft’s Copilot Studio and Salesforce’s Agentforce. For businesses with unique requirements, custom-built AI agents offer a tailored approach that allow business users to build purpose-built AI Agents and also extensions and plugins for the off-the-shelf AI Agents outlined above. Copilot Studio allows users to create and train AI models that can be integrated into various applications, providing a high degree of flexibility and control. They can also build extensions and other agents that can be integrated into 365 Copilot. Similarly, Salesforce Agentforce empowers business users to build custom AI agents within the Salesforce ecosystem, leveraging the platform’s robust capabilities to address specific challenges and opportunities. 

Security Considerations for AI Agents

Nearly every organization is looking for ways to enable the business with AI Agents in all their various forms. One of the things we’re hearing most frequently is that as AI agents become more integrated into business processes, security teams must address several critical considerations to ensure safe and effective use. 

Data Sensitivity: One of the primary concerns is the handling of sensitive data and that business users are in control of what data gets processed, who can access it, and how guardrails are enforced. By nature, AI Agents often process large volumes of information, some of which may be confidential or sensitive. Security teams need to implement robust data protection measures, including encryption, access controls, and regular audits, to prevent unauthorized access and data breaches. For Agents built on platforms like Copilot Studio, that are inherently built outside of the SDLC and CI/CD pipelines, further controls and granularity are needed.

Prompt Interpretation: Another important aspect is how AI agents interpret prompts and triggers. Misinterpretation can lead to unintended actions or responses, potentially causing operational disruptions or security vulnerabilities. It is essential to thoroughly test AI agents in various scenarios to ensure they respond appropriately and consistently. Additionally, implementing safeguards such as human-in-the-loop mechanisms can help mitigate risks by allowing human oversight and intervention when necessary.

Compliance and Governance: Ensuring compliance with relevant regulations and industry standards is also crucial. Security teams should work closely with legal and compliance departments to establish governance frameworks that address the ethical and legal implications of using AI agents. This includes maintaining transparency in AI operations, documenting decision-making processes, and providing clear guidelines for the responsible use of AI technologies.

Looking Ahead

AI agents are revolutionizing the enterprise landscape, offering significant benefits in terms of efficiency, innovation, and competitive advantage. By understanding the different types of AI agents and the options for implementing them, businesses can make informed decisions that align with their strategic goals. However, it is equally important to address the security and ethical considerations associated with AI agents to ensure their safe and responsible use. Understanding what risk exists is a good place to start; check out our Assessment Hub for a variety of open-source tools that you can use yourself to assess risk across both the Agentic AI and low-code ecosystems!