Big AI companies are moving beyond chats
The recent announcements give comms teams a reason to reassess what they delegate, even as access remains uneven.
Stephanie Nivinskus is principal, Ragan’s Center for AI Strategy.
Since ChatGPT launched, many communications professionals have come to understand that using AI means opening a chat, writing a prompt and applying their judgment to refine the output. That is accurate. It is also the equivalent of describing a smartphone as something you use to make calls.
The month of September made the chasm between how many use AI and how they could use it glaringly obvious. Within three short weeks, Meta, Microsoft and OpenAI introduced persistent AI agents that can work on cloud computers long after you walk away from yours.
What makes persistent agents different
In a basic chat, you ask and the AI answers. An AI agent can take an assignment and use tools to carry out the steps. The excitement around persistent agents revolves around their ability to retain the assignment’s context and resume work without another prompt. Depending on the product and permissions, these agents can check for updates, follow up, or pick up an assignment days later.
What that looks like in comms
Imagine giving a persistent agent an assignment hours before you leave for vacation. Then coming back a week later to find that a project you have been working on for months kept moving while you were sipping a piña colada on the beach in Mexico.
Consider a product announcement awaiting input from legal, an executive and a subject matter expert. With approved access and instructions, a persistent agent could track incoming feedback, flag conflicting edits and follow up with reviewers you have authorized it to contact.
If the launch date changes, it could propose a revised review schedule. Your team would still decide what gets published, but the assignment could keep moving without someone manually chasing every update.
Explore a pilot with IT
OpenAI’s persistent agents, called “dots,” are in beta for eligible enterprise customers, but administrators must enable access. Microsoft’s Autopilot is available only to selected Copilot customers for testing. Meta launched Muse as a consumer app and its release does not establish enterprise availability.
Whichever is your platform of choice, comms teams can start identifying work worth delegating and explore these steps with IT:
- Fix the workflow first. Identify where a recurring assignment stalls. Clarify who reviews it and who approves it before exploring automation. Work with IT to assess whether an approved tool supports the assignment and its required information.
- Agree on authority and accountability. Comms should define which editorial decisions require human approval. IT should establish technical access and controls. Tracking approvals does not automatically authorize accepting edits, changing deadlines or publishing an announcement. Assign a comms owner to supervise the work.
- Evaluate a small pilot together. Test an approved assignment for accuracy and follow-through. Measure the effort required to supervise it, including correcting mistakes. Use those results to decide whether to expand its role.
