BROKER RISK MANAGEMENT
WEEKLY PRACTICE TIP
Artificial Intelligence Agents Explained
Broker Risk Management has previously addressed the legal considerations surrounding the use of generative large language models (“LLM”) artificial intelligence systems but has not yet addressed the emerging Agent and Agentic AI systems (“AI Agent”) that have recently been introduced into business operations and public life. Currently, many people’s understanding of AI is limited to large language model (“LLM”) systems such as ChatGPT, in which a user provides a prompt and the system generates a predictive response to the inquiry. AI Agents differ in that they can act autonomously to perform tasks without complete human control. Many of us have already interacted with such AI systems in the form of customer service agents before demanding to speak to a human operator.
AI Agents are evolving and are beginning to develop the ability to complete decision-making, problem-solving, and other performance-related tasks and capabilities. Gartner, a technology researcher estimates that by 2029, 80% of customer service requests will be completed by AI Agents. While this technology is progressing, the law has not developed at the same pace.
In 2024, in Mobley v. Workday in the U.S. District Court for the Northern District of California, the plaintiff filed an employment discrimination lawsuit against Workday alleging that the plaintiff was unfairly discriminated against by Workday’s AI agent after being rejected from over 80 job applications without an opportunity to interview. Workday’s motion to dismiss was granted with leave to amend. However, the court noted that the plaintiff could potentially plead claims of discrimination against Workday based on Workday’s AI platform. While the decision applies to employment discrimination claims, it could potentially open the door to applying to traditional agency law principles to hold businesses legally responsible for the actions of AI agents in the course of business. This reasoning could create liability for agents. For example, if an AI agent informs a potential client that Section 8 tenant applications are not accepted, the real estate brokerage could potentially be held responsible under principles of agency law.
Risk Management Recommendations for AI Agents
Onboarding an AI agent into your business should be treated more like onboarding a new employee rather than an autonomous tool. The agent should be provided with clear job descriptions and business role definitions, as well as continuous feedback and evaluations. Just like an employee, an AI agent must understand what you want it to accomplish, how to complete its tasks, and how to improve its performance. Without clear instructions and oversight, an AI agent may make assumptions about its perceived responsibilities that do not align with your intended goals or business objectives.
It is important to discuss with your AI agent provider whether they can offer the ability for you to instill “guardrails” as to what the agent can offer or discuss. In a real estate setting, this could be applied to an AI Agent for a real estate brokerage offering to represent a client for a severely discounted commission without the broker’s authorization. Setting discussion guardrails can help avoid such mistakes from occurring.
It is always important to remember that AI systems merely predict what they believe to be the highest probability of the proper response or action. As a human, you retain the final authority for judgment and critical reasoning.
Please remember that AI Agents cannot undertake any activities for which a real estate license is required. AI Agents do not replace the judgment, experience or expertise of a licensed agent. If the AI Agent makes a mistake, the Department of Real Estate (and Department of Civil Rights, if discrimination is involved) will hold the agent and potentially, the broker liable.
Practice Tips:
- Start With an AI Agent Business Plan: Create a highly detailed business plan outlining how you intend to use an AI agent in your business, and clearly define its limitations and responsibilities. This includes designating specific responsibilities and identifying company personnel to oversee its use.
- AI Agent Onboarding: Confer with a qualified AI professional (likely a representative from the AI Agent company you intend to employ) to discuss your intended use and objectives for an AI agent, as well as how to properly onboard and monitor the agent for your business. This discussion will also include how to properly define the AI Agent’s responsibilities and how to draft user prompts and feedback.
- Start Small and Specific: Limit the AI agent’s responsibilities during the initial phases of use to test its effectiveness and provide feedback, thereby reducing the potential risk of mistakes and liability. As the AI Agent is refined and proven to be effective, slowly increase its responsibilities.
- Avoid the “Set and Forget” Mindset: A designated AI professional or business employee should be responsible for monitoring the AI agent to ensure it is functioning effectively and appropriately. AI systems are designed to learn from user feedback and performance evaluations, which help them operate more efficiently and deliver stronger results. “Stress testing” is an effective form of testing AI systems by testing the AI system for vulnerabilities. Ask the AI Agent for a difficult or unusual question to test its response.
- Warnings: Do not allow the AI Agent to perform any activities for which a real estate license is required. AI Agents’ responsibilities must be limited to strictly administrative tasks.
WEEKLY PRACTICE TIP: DO NOT FORWARD TO CLIENTS. This Weekly Practice Tip is an attorney-client privileged communication for the exclusive use of clients of Broker Risk Management and their agents. It may not be reproduced or distributed without the express written consent of Broker Risk Management LLP. The advice and recommendations contained herein are not necessarily indicative of standards of care in the industry but rather are intended to suggest good risk management practice.