> ## Content Index
> Fetch the complete content index at: https://envoyfm.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# My AI Team Runs on Basecamp
- URL: https://envoyfm.com/my-ai-team-runs-on-basecamp/
- Published: 2026-11-11T14:00:00.000Z
- Updated: 2026-11-16T14:00:26.000Z
- Author: Scott Reyes

In the morning, while my family is still asleep, I go downstairs, make coffee, and plan my day. I start with Basecamp.

I open Grok Bot and ask Calloway, my AI agent, to scan Basecamp and let me know what needs our attention. He returns a list with links to the Basecamp cards. I click through them and decide what to do. For some, I message a person. For others, I ask Calloway to handle the work.

![A man on the phone reads a notebook and points to a stack of message cards, with blue envelopes on a green desk.](https://envoyfm.com/content/images/2026/11/2C-my-ai-team-runs-on-basecamp.png)

He has his own email address and Basecamp account, with tools that let him do the work I assign. If he has a question or needs me to review something, he puts it in Basecamp. Otherwise, he reports the result.

I've used Basecamp for years to manage work with people. Now, some of those conversations include an agent. As I give him more responsibility, I have to be clear about what starts a job, what he can decide, and what I need back.

![A man holding a pen looks at a laptop showing a step-by-step flow of answered questions that ends in a checked-off document.](https://envoyfm.com/content/images/2026/11/3B-my-ai-team-runs-on-basecamp.png)

## A job with a clear finish

The other day, we were getting an article by Justin Marino, our COO, ready for Envoy. Our editor, Antionette, had worked on the draft. I asked Justin to review her edits and let me know whether it was good to go.

He replied, "Already done."

I asked Jeric, our designer, to create the images. He prepared twelve options. Justin picked four, and Jeric uploaded them to Google Drive and put the link on the card.

I tagged Calloway: "Please add this post as a draft."

He returned a link to the draft in our publishing system and a screenshot of the preview. His update said he'd copied the title and text exactly from the working document, added Justin's byline, and included the four selected images with alt text. He left the post unpublished, as I'd asked.

Calloway's job was to prepare approved material for review. Publishing would require another decision.

All of it stayed on the same Basecamp card: the working document, image choices, request, and result. I didn't have to move the feedback into a separate AI chat or reconstruct which version we were discussing.

For a maintenance team, that means keeping the work order notes in its maintenance management software, where everyone can see what happened and what still needs attention.

Around the same time, another article came in from Jeric, about how AI had changed his work as a designer. He emailed it to the address we use for submissions.

Calloway passed it to Edie, the bot that handles our editorial intake. She checked for an existing card and created one in Basecamp with Jeric's name, the intended reader, a summary, and where to find the article. The card noted that it was email text, with no Google Doc yet.

I hadn't asked anyone to create that card. Receiving the article started a job we'd already defined. Edie organized it for me to review. I still had to decide whether to send it to an editor. When a submission arrives as a shared Google Doc, Edie also gives our editors access.

![A man slides a paused card along a blue track into a tray marked with an AI agent's face.](https://envoyfm.com/content/images/2026/11/4A-my-ai-team-runs-on-basecamp.png)

## Problems need a next step

Jeric noticed a problem with Justin's article preview: the author image looked misaligned. Calloway tagged me in a comment, but I wanted him to take it one step further.

I wrote, "In the future, when you see these, please add them as a card on the development project."

He created a bug card describing the problem and the requirements for a fix. It called for checks on phones, tablets, and desktop screens, with before-and-after screenshots for review.

The image still needed fixing. Now the development team had a card to act on and requirements for checking the result.

![A man points to a spread of image cards on a desk next to a card with an AI agent's face.](https://envoyfm.com/content/images/2026/11/5A-my-ai-team-runs-on-basecamp.png)

When an agent hits a problem, I need to define where it goes and who takes it next. Otherwise, a tag can leave me responsible for figuring out the next step.

It's tempting to keep adding jobs. I do it a lot. An agent handles one job well, and I immediately think of something else to give it. If I haven't defined that next job, I leave it room to guess. That's where the work can become unreliable.

I've just added a Ready to Schedule step for articles. Moving a card into that column gives Edie the job of preparing the approved article in the publishing system and scheduling it for the date on the card. If the date is missing, she needs to ask me. When we set it up, we'd checked the connection but hadn't yet run an article through the whole flow.

![A man holds up a finished document with an AI agent's face, beside a row of steps with cards on each one.](https://envoyfm.com/content/images/2026/11/6C-my-ai-team-runs-on-basecamp.png)

## Start with one maintenance handoff

Suppose a contractor reports a repair is complete, and you want the service report and required photos collected so a coordinator can review the work.

Before assigning that to an agent, your team should agree on what documentation is required for that kind of repair. An agent with access to the work order and permission to contact the contractor could check what's present and request what's missing. The result would be the documents attached to the right work order, with a note identifying anything still missing.

Attachments alone don't establish that the repair was successful. Your coordinator would still decide whether there's enough information to accept the repair. A photo could show the wrong equipment, or a service report could leave an important question unanswered. Define what the agent should check and what the coordinator needs to review.

Decide what happens when the contractor doesn't respond or says another visit is needed. The agent needs to know when to stop following up and whom to notify. That person should receive the work order, the responses so far, and the decision that needs their attention.

Try it on a small group of work orders and review what comes back. Were the documents for the right job? Did the agent flag what was missing? How much did the coordinator still have to chase down or correct? Use those results to decide whether to expand the assignment or improve the process first.

![A coordinator in a headset works at a laptop while a contractor in a hard hat talks on the phone, with an AI agent's status card between them.](https://envoyfm.com/content/images/2026/11/7A-my-ai-team-runs-on-basecamp.png)

## The AI-native department

I believe the future of work is AI-native. Facilities maintenance departments will have people and AI agents working as part of the same team. We'll need to give them clear responsibilities and build processes that let the work move between them.

My setup is one example of what that can look like. Your department has its own work to do and decisions to make. Think about the jobs an agent could take on, the responsibilities your people would keep, and how they could help each other get the work done.

What would that look like in your department?