Peninsula School District
The Open Adaptive District • psd401.ai

The Playbook

A cycle uses three documents: a build plan in week 1, a check-in every Friday, and a wrap-up in week 6. This page has a template for each. If you only want the steps, the one-page Get Started guide has the whole cycle on a single sheet.

One rule: one build per team per cycle. If your idea won’t fit in six weeks, split it into two builds.

Good problems sit inside goals you already have: a smoother MTSS meeting, scaffolds built faster so every student gets grade-level work, family messages that people actually read.

The cycle at a glance

When What happens
Week 1: Plan Pick one problem and write your one-page build plan. Post it in your team’s chat space by Friday, along with your first check-in.
Weeks 2–5: Do Build it and use it, and adjust as you learn. Post a check-in every Friday.
Week 6: Study and share Your agent drafts the wrap-up from your check-ins. The team edits it (about twenty minutes), makes the call (keep it, stop it or run it again) and publishes it. Then read the other teams’ wrap-ups.
Weeks 7–8: Between cycles Teams take a break. Cabinet reads every wrap-up, and the superintendent sets the focus for the next cycle.

Your team

Three to six people: a team lead who owns the build, a scribe who posts the check-ins, and one to four others doing the work. Each team also has a sponsor, usually the leader who oversees its department, who clears blockers when the team is stuck. Willingness matters more than job title.

Set up a Google Chat space for the team and invite the agents by their agnt_ accounts: each member’s own PSD AI Agent, and Kris Hagel’s agent, agnt_hagelk@psd401.net. His agent has to be in every team’s space so it can see across teams and report back.

For other districts. The sponsor can be any senior leader with the authority to clear blockers. The chat space can be Google Chat, Microsoft Teams, Slack, or whatever your staff already use, as long as your AI agents can read it. Put one agent in every team’s space, so someone can see across all of them.

1. Build plan (week 1)

One page, written quickly. You will change it as you learn. Ask your agent to help you draft it.

Team: ___ Cycle: ___ Date: ___

Lead: ___ Scribe: ___ Sponsor: ___

The goal it serves. Which goal you already have does this support? For example, a school or department improvement goal, an MTSS commitment, or a priority in your strategic plan.

The problem. Two or three sentences. What is eating time, causing friction, or not working, specifically, in your building or department?

Our build. “We’re going to build ___, because we think it will ___.”

Tools we’ll try. Which AI tools will you use or test? Name them, so the wrap-up can say what each one was good for.

How we’ll know. Two or three signs of success, at least one of which you can count. Also name one thing that must not get worse, for example “families still feel personally communicated with.”

Week-one plan. Just the first week. You will plan the rest as you learn.

Not this cycle. What you are deliberately leaving out, so the build stays small enough to finish.

What it could become. If this works, what would everyone else get? Usually a skill added to the PSD AI Agent or a tool in AI Studio. One sentence is enough; you will know more by week 6.

Peninsula teams: once the plan is posted in your space, file it with the build plan form by the Friday of week 1. The form asks the same questions, plus a sign-off that agnt_hagelk@psd401.net is in your space.

2. Weekly check-in (every Friday)

Four questions, two or three sentences each, posted by the scribe in the team’s chat space every Friday. It should take five minutes. Your agent can remind you and post the questions. On Sunday afternoon, Kris Hagel’s agent reads every team’s space.

Tried. What did we actually do this week? Name specific actions.

Learned. What did we find out? Include the surprises.

Stuck. Where are we blocked?

Needs. What would unblock us, and who can help?

Our pilot teams taught us not to force every conversation into chat. Meet in person, in the hallway or on a call, then drop what came out of it into the space: a photo of the whiteboard, a transcript or recording, or three bullet points. The agent can only learn from what is in the space.

If a team misses two check-ins in a row, someone will check in to see what is blocking it.

3. Wrap-up (week 6)

Your agent drafts this from your check-ins. The team edits it in about twenty minutes and makes the call.

What we tried. From your build plan.

What happened. What you actually did and what resulted, in plain terms.

What we learned. Including the surprises and anything you still don’t understand.

The call. Pick one:

  • Keep it. It worked. We make it available to everyone, usually as a skill in the PSD AI Agent or a tool in AI Studio. Note what other teams should know before they use it.
  • Stop it. It didn’t work. Say why, so no other team spends a cycle repeating it. Stopped builds are published like any other.
  • Run it again. You’re not sure yet, but you have a sharper question. Write down the question the next cycle starts from. If a build has run twice without producing a sharper question, it is usually a stop.

What you don’t have to do