Work and Productivity

Start With What Already Takes Too Long

This page is for staff work: the email, the reports, the meeting notes, the documents that have to exist. The Spring 2024 district survey found the most common uses were writing messages, summarizing information, and summarizing data — all forms of handling text that already exists rather than inventing something new. That is where these tools are most reliable.

The useful place to start is not "what can AI do" but "what in my week is repetitive, text-heavy, and not a judgment call." That is the work worth handing over first.

What you already have: Gemini, NotebookLM, and Workspace Studio are licensed for Maricopa faculty and staff with district data protections in place. See Tools for where to find them.

Tasks Worth Handing Over

The most reported use across the district. Drafting responses to common inquiries, writing the message you have been avoiding, adapting one announcement for three different audiences.

The workflow that works: get a draft, then make it yours. Readers increasingly recognize unedited AI prose, and in professional communication it reads as a lack of care. Paste in two or three things you have written and ask the tool to match how you write — far more effective than describing your style.

For recurring messages, write the instruction once, carefully, and save it. A reusable prompt is worth fifteen minutes of setup if you send that kind of message weekly.

Turning a long report, a stack of committee materials, or messy notes into something you can actually use. Converting notes into action items with owners. Producing a two-sentence version for a newsletter and a fuller one for the record.

Source-grounded tools are noticeably better here: NotebookLM answers only from documents you upload and cites which one each point came from, so verifying a summary takes seconds rather than a re-read.

Writing up a process that currently lives in someone head, drafting onboarding material, producing templates and standard forms, formatting existing documentation consistently.

This is high-value work that rarely gets done because it is tedious rather than difficult — which makes it a good fit.

Explaining what a spreadsheet formula does or building one you cannot quite construct. Summarizing trends in aggregated, de-identified operational data. Suggesting how to present figures more clearly.

The line matters here: aggregated data with no individual identifiers is one thing; anything that identifies a student or an employee is another, and does not go in. See Data Privacy and Security.

Breaking a project into phases and tasks, drafting agendas, building checklists, thinking through scheduling constraints. Expect timelines to be optimistic — the tool does not know your approval cycles or committee calendar.

Producing materials in another language, or in plain language, for students and community members who need them. One survey respondent described translating lessons into Spanish as an intermediate speaker — work that would otherwise have taken far longer.

Check the result for anything conditional. Simplification and translation both tend to smooth away exceptions, deadlines, and qualifying circumstances.

What Stays Off Limits

Five categories never go into an AI tool, regardless of which one:

  • Student records and anything identifying a student
  • Personnel records, evaluations, salary data, HR matters
  • Confidential institutional data — contracts, vendor pricing, grant details
  • Non-public institutional information
  • Intellectual property that is not yours to share

A test that works: would you send this to an outside contractor without additional approvals? Maricopa is also a public institution subject to records requests, so "would I be comfortable seeing this in a public forum" is a reasonable second check.

And the decisions that stay yours

AI does not make hiring decisions, personnel judgments, financial aid determinations, or anything requiring accountability for a specific person's outcome. It can help you draft the communication about a decision you made; it does not make the decision.

If you are moving beyond using a tool toward building something colleagues will depend on — an automated workflow, a configured assistant, something embedded in a service — that carries review requirements around data handling, procurement, accessibility, and records retention. Choosing the Right Tool covers what applies.