5 Questions to Ask Before an AI Agent Acts on Your Behalf
Quick answer: PRSA's Ethical Standard Advisory #22 states that AI agents are tools, not ethical actors, and the human deploying them is accountable. Before any AI agent acts on your behalf, answer five questions: disclosure, review, access, accountability and literacy.
I was recently at a local AI workshop for small business owners when the trainer asked the room a simple question: what is the difference between agentic AI and generative AI? Silence. A few people glanced at their neighbors. No one really knew, and these were smart, motivated people who came to a workshop on purpose.
That gap in knowledge is exactly what the Public Relations Society of America is trying to get ahead of. PRSA's Board of Ethics and Professional Standards just published Ethical Standard Advisory 22, "AI Agents in Public Relations," its guidance on AI systems that do not just draft your content but act on your behalf.
What Is the Difference Between Generative AI and an AI Agent?
Generative AI creates content when you ask it to. You prompt, it drafts, you decide what happens next. An AI agent is software that pursues goals and completes tasks on its own. It can plan, remember, use other tools and adapt as it goes. That is different from a chatbot you prompt and supervise sentence by sentence. Agents can monitor media, generate and send content, respond to stakeholders and coordinate with other agents, sometimes without a human approving each step.
Why This Matters More Than Previous AI Shifts
The advisory is blunt about why this matters more than earlier digital shifts: agents can act independently at scale, simulate human identity and adjust messaging in real time without anyone signing off. Adoption is not hypothetical. A 2025 Muck Rack survey found 59 percent of PR practitioners say AI and automation will grow in importance for them over the next five years, and Gartner predicts that by 2028, 33 percent of enterprise software applications will include agentic AI, up from less than one percent in 2024.
Who Is Responsible When an Agent Gets It Wrong?
The heart of the document is one question: when an AI agent initiates or shapes communication, who is responsible? PRSA's answer leaves no wiggle room. AI systems are tools, not ethical actors. The human practitioner is accountable for how those tools are designed, deployed and supervised. The algorithm did it is not a defense.
The advisory names practices to avoid, including deploying an agent to talk with journalists, customers or employees without disclosing it is an AI; using agents for hidden persuasion or microtargeting; coordinated agents posing as independent people; and publishing agent-generated statements, even in a crisis, without meaningful human review.
Two of the risks the advisory flags are not about deceiving the public at all. They are about what happens inside your own organization: invisible dependence, teams leaning so hard on agents for monitoring, drafting and recommendations that human judgment quietly atrophies, and reduced transparency, agents becoming so embedded in workflows that nobody in the building can explain why a message went out or what data drove a decision. Those are workplace culture problems before they are technology problems, and the advisory's answer is a training answer: build AI literacy across teams, establish human-in-the-loop governance and set clear policies about acceptable use.
A Small Business Scenario
Picture a small business with a team of five that sets up an AI agent to answer customer questions when the shop is closed. It is helpful, fast and available at 2 a.m. Then one night it confidently promises a customer a full refund on a custom order, a policy the shop has never had, and by morning three more customers have screenshots.
Now run PRSA's question: who is responsible? Not the agent. The advisory would ask whether anyone told customers they were chatting with software, whether a named human reviewed what the agent was telling people and whether the team ever assessed what files and systems the agent could touch, including customer contact information and order history. In my experience teaching PR students and working with organizations on AI adoption, almost nobody has answers to those questions written down. The tools arrive months before the governance does.
The Five-Question Governance Checklist
Before any AI agent speaks or acts for your organization, put these five questions in front of your team. They are distilled from Ethical Standard Advisory 22 and make a solid one-page governance starter.
Disclosure: Would the person on the other end know they are interacting with AI? If it would change how they feel about the exchange, disclose it.
Review: Who, by name, not by department, reviews what the agent says before or shortly after it says it?
Access: What data, files and systems can the agent touch, and has IT or security assessed that?
Accountability: When the agent gets something wrong, who owns the correction? Decide before it happens, not after.
Literacy: Has everyone working with the agent been trained on what it cannot do, not just what it can?
If your team cannot answer all five, that is not a reason for shame. It is an agenda for your next meeting.
Frequently Asked Questions
What is an AI agent, in plain terms?Software that completes multi-step tasks on its own toward a goal, rather than simply responding to a single prompt. It can plan, use tools and act without a person approving each step.
Who is responsible when an AI agent makes a mistake?According to PRSA's Ethical Standard Advisory 22, the human practitioner who designed, deployed or supervised the AI system is responsible. AI agents are classified as tools, not ethical actors, so accountability cannot be assigned to the software itself.
Does a small business need a formal AI agent policy?Yes, even a small one. At minimum, write down who reviews agent output, what systems the agent can access and who owns corrections when something goes wrong, before deploying any agent that interacts with customers or handles sensitive data.
Rolling Out an AI Agent and Need the Governance Conversation Started?
This is exactly the gap my AI workshops are built to close. If your team is deploying AI agents and the governance conversation has not started yet, reach out through Prompt & Proper Consulting and let's build the one-page policy before you need it.
Comments