AI & Facilitation·Sep 30, 2026

What always-on AI agents could mean for Stormz

How background AI agents could use Stormz's MCP connector to create activities and prepare workshop reports, with human approval where it matters.

Alexandre Eisenchteter

A client changes the workshop brief while I'm working on something else. An AI agent picks up the email, reads the activities I've already prepared in Stormz and comes back with a suggestion. Once I've approved it, the agent makes the agreed changes.

That's one of the possibilities I have in mind when I talk about always-on AI agents. In my recent AI & Facilitation: The Next Frontier webinar, I described agents working in the background, with an ongoing mission and the ability to take initiative. OpenAI's new Dots announcement gives us a concrete example of that direction.

Unfortunately, I don't have access to Dots yet here in Europe. I haven't connected one to Stormz, and this article isn't announcing a Dots integration. It's about what this kind of agent could mean for workshop creation and reporting, given what Stormz's MCP connector already makes possible.

Stormz already gives AI assistants tools to work with workshops

The Stormz MCP connector lets a compatible assistant, such as Claude or ChatGPT, work with your workshops through your Stormz account. MCP is the connection that lets the assistant use tools in another application. Your existing access and facilitation permissions still apply.

You can ask an assistant to create a workshop, add activities and configure their questions and views. It can also retrieve participant contributions for analysis. The connector has tools for reading survey aggregates and card-voting results, so an assistant can use the workshop's actual results when drafting a report.

I've already described how I use these connections across my workshop tools, from session design to reporting. An always-on agent could continue following that work after I leave the conversation. The connector supplies the tools; the agent would need to manage the ongoing job.

A change in the client brief could become a proposed activity

Suppose an agent has a mission to help me prepare a team retrospective. It can follow the relevant email thread and read the preparation documents, assuming those connections are available.

The client writes to say that two teams will now attend together. Their main concern is the handover between them. The agent could compare that update with the existing Stormz activities and suggest a more specific question: “Where does work get stuck when it passes between our teams?”

I'd want to discuss the suggestion before it changes anything. Perhaps that question belongs in an existing activity. Perhaps it deserves a separate activity where participants first describe the difficulties, then suggest improvements.

Once we've agreed on the design, the agent could use Stormz's connector to create and configure the activity. Several authoring operations support previews without saving, which can help us review the proposed setup. I'd still check the result in Stormz before participants use it.

The useful initiative here is noticing that the original activity no longer quite fits the brief. I want to keep the decision about what to do with that observation.

Reporting could follow an async workshop as it develops

For an async workshop, participants might contribute over several days. I'd like to give an agent a mission such as:

Follow the contributions in this workshop until Friday. Keep a private draft summary up to date, with references to the cards behind each theme. Flag questions I may need to follow up on. Don't change participant contributions or share the report without my approval.

As new cards arrive, the agent could revise the draft and tell me what changed. A concern that appeared isolated on Monday might be echoed by several contributions on Wednesday. A disagreement might deserve a follow-up question rather than being smoothed into a tidy summary.

Stormz's connector can provide the content for that analysis. Checking for changes on a schedule, or reacting to new contributions, would need support from the agent and its connections. Adding an MCP connector alone doesn't make a workshop continuously monitored.

The same approach could help after a live session. An assistant could bring together the contributions and the results of the voting session I select, then prepare a report for review. I'd want it to keep the evidence separate from its interpretation: a highly ranked card isn't automatically a decision, and a proposed action doesn't have an owner unless someone agreed to take it on.

Human approval matters when I'm no longer watching every step

With background agents, human-in-the-loop principles become fundamental. I'd be comfortable letting an agent update a private report draft or add a question to my preparation board. Changing an activity participants are using, rewriting their contributions or sending a report to the client would need my approval.

I'd also want it to work within a clearly identified workshop and use only the information needed for that mission. One of my first things to explore would be how the agent's permissions and approval steps work in practice. Access to a tool doesn't tell me when an agent will decide to use it.

We can start tinkering with the connector today

You don't need to wait for Dots to explore activity creation and reporting through Stormz's MCP connector. With a compatible assistant and connector access, you can start with one workshop: discuss an activity, approve its creation, then ask the assistant to read the resulting contributions and draft a summary.

Connect Claude or ChatGPT to Stormz, and review what it produces before giving it a bigger job.

When I get access to Dots, I'd like to see how much of that workflow can carry on between conversations. For now, there's already plenty to tinker with.