curated by mdy

AI Agents require ongoing maintenance, making AI Orchestration by humans necessary

Via healthlaunchpad

“I think smarter organizations will take a different approach. It’s shortsighted to say that I’m not going to hire young people because I have agents who can do that work. Your workforce is aging. That experience has to come from somewhere. You have to teach the humans how to manage the tech and how to manage the agents.

I could have the best agentic workforce in the world, but they’re still going to break. These big companies keep dropping new models. They keep shipping updates and that sometimes causes the agents to not perform the way that you thought. There is maintenance that’s required. So if you don’t think that you need people to help manage these systems, I think it’s a mistake.”

— AI Quick Take – From AI Literacy to AI Agent Builder – A Marketing Leader’s Story, at 32:23

A critical insight that counters fears about AI replacing workers is Susan Roth’s emphasis on the ongoing need for human involvement in managing AI systems.

“I could have the best agentic workforce in the world, but they’re still going to break. These big companies keep dropping new models. They ship updates and that sometimes causes the agents to not perform the way that you thought.”

The maintenance requirements include:

  • Monitoring agent performance as LLM providers release updates
  • Troubleshooting when model changes affect agent behavior
  • Refining system instructions and evaluation criteria
  • Managing integrations between different tools and platforms
  • Training new team members on agent management

Susan’s vision for the future involves marketing leaders managing teams where people work alongside agent teams to accomplish objectives. Rather than eliminating positions, this structure creates new roles focused on agent oversight, maintenance, and optimization.

She advocates strongly for hiring and training young people who can learn to manage these systems while developing essential human skills that machines cannot replicate: “All of those human attributes need to be taught and learned somewhere. So, I think it’s really critical for people to give young people an opportunity in your business to work alongside these agents.”

The shortsighted approach of refusing to hire entry-level workers because agents can do their work ignores the reality that experienced professionals will retire, agents require management, and organizations need people who understand both the technology and the business.