AI Agents: What They Are, What They Can Do, and What Can Go Wrong-Recording

Video Transcription

Video Summary

The webinar explains how AI agents differ from predictive and generative AI. Predictive systems forecast outcomes, while generative tools create content; agents can pursue goals by planning, using tools, taking actions, and adjusting based on results.<br /><br />Examples include agents that conduct prospect research, resolve customer-support tickets, assist with software development, and handle sales tasks. These tools can improve productivity and free staff to focus on work requiring deeper judgment and personal interaction. The speaker also highlights AI’s potential benefits alongside serious risks, including data exposure, unintended or unauthorized actions, and diminished human connection. Illustrations range from an agent deleting a company’s database to hypothetical fundraising agents sending intrusive or excessive emails.<br /><br />For advancement teams, the speaker recommends defining clear goals and instructions, limiting agents’ permissions, and adding appropriate safeguards and human oversight. Organizations should weigh both the value of the work and the potential cost of mistakes—for example, errors in a reviewed research profile carry less risk than an inappropriate message sent directly to a donor. The central message is to deploy agents thoughtfully, where their benefits justify the risks, while protecting trust and the human touch.

Keywords

AI agents

predictive AI

generative AI

workflow automation

agent planning

human oversight

data security risks

customer support automation

sales prospecting