How I Work With AI: General Guidance
Principles applied to daily practice
This document is a companion to How I Work With AI (v5). Where the main framework
describes values, philosophy, and governance, this guidance describes how those
principles show up in the day-to-day reality of working with AI. It is intended to be
practical without being prescriptive, giving enough structure to maintain consistency
while leaving room for the judgment the framework prioritizes.
Governing document: How I Work With AI, v5
Before You Start: The Front-Loading Principle
The single most impactful habit in my AI practice is front-loading context before asking
for output. This means providing the what, who, why, tone, constraints, and any relevant
source material before the first prompt. When I do this well, the first draft is usually
close. When I skip it, I’m looking at two to three extra revision rounds that cost more
time than the front-loading would have.
Before starting any AI-assisted task, I ask myself: Does the system have everything it
needs to do this well on the first pass? If not, I provide it before I ask for output.
Daily Practice Principles
Principle 1: AI drafts. I decide.
Every piece of AI-assisted output goes through my review before it reaches anyone else.
This isn’t a bottleneck. It’s the point. AI handles production. I handle judgment. The
review step is where my voice, my context, and my relationships with the audience get
applied. Skipping it is never acceptable, regardless of how routine the task feels.
Principle 2: Match the tool to the task.
Not every task needs AI, and not every AI tool is right for every task. Before reaching for
a tool, I consider whether the task genuinely benefits from AI assistance or whether I’m
using it out of habit. Some work is better done by hand. Some thinking is better done
alone. The goal is augmentation at the right moments, not automation of everything.
Principle 3: Iterate, don’t accept.
My best AI-assisted work comes from treating conversations as collaborative drafting
processes. The first output is a starting point, not a deliverable. I provide feedback,
redirect, push back on things that don’t land, and get specific about what’s working and
what isn’t. The quality of my output is directly proportional to the quality of my
engagement with the process.
Principle 4: Protect what’s sensitive.
Before entering any information into an AI system, I consider whether it contains
protected data, confidential information, or content that shouldn’t leave its institutional
context. FERPA-protected student information, personnel details, and sensitive
institutional data stay within approved systems. When in doubt, I leave it out.
Principle 5: Disclose when it matters.
I don’t hide my AI use, but I also don’t performatively announce it in contexts where it’s
irrelevant. My disclosure practice follows common sense: academic work gets cited,
professional contexts get transparency when asked or when it matters for trust, and
internal working documents don’t need a disclaimer on every draft. The test is whether
someone receiving my work would reasonably want to know that AI was involved.
Principle 6: Maintain the voice.
Every output that carries my name should sound like me. This means actively editing AI-
generated content for voice, not just accuracy. It means catching and removing AI tells
(the “Additionally” transitions, the em-dash habits, the over-structured prose). It means
reading output aloud when something feels off. Voice consistency is not optional. It is
the first priority.
Principle 7: Stay honest about quality.
If AI-assisted output isn’t good enough, I say so and either revise it or start over. I don’t
send mediocre work because it was efficient to produce. The standard for AI-assisted
output is the same as the standard for any output that carries my name: would I be
proud of this if someone asked me about it?
Principle 8: Keep learning.
AI capabilities change quickly. My practice should evolve with them. This means
periodically reassessing which tools I use, how I use them, and whether my habits still
serve me. It also means being honest about growth edges: the exploration-to-
implementation gap, the scope creep tendency, and the polishing question are all
patterns I continue to monitor.
Context-Specific Guidance
Professional communications
AI is most useful here for first drafts and structural thinking. I provide the audience, the
purpose, and the tone. I review for voice, accuracy, and whether the message actually
says what I need it to say. Sensitive communications (personnel issues, political
situations, crisis response) always get more careful human attention and are never sent
without significant personal revision.
Academic and scholarly work
AI supports my research through literature synthesis, concept organization, and
drafting assistance. I cite AI use in accordance with APA 7 guidelines and my
institution’s academic integrity policy. My scholarly voice, my analysis, and my
arguments are my own. AI helps me organize and articulate. It does not think for me.
Teaching and curriculum
AI helps me develop workshop designs, curriculum frameworks, and instructional
materials. When I teach students about AI use, I model the same practices I describe in
this document: transparency, integrity, and the understanding that AI is a tool that
requires ethical judgment to use well.
Student-facing materials
Materials that go to students and families carry particular responsibility. These
audiences are navigating significant life transitions and deserve accuracy, warmth, and
care. AI drafts are always reviewed with the audience’s experience in mind, not just the
content’s correctness.
The Daily Check
At the end of any significant AI-assisted work session, I ask:
Did I front-load enough context, or did I waste time on avoidable revisions? Does the
output sound like me? Is it accurate? Would I be comfortable if the full process were
visible? Am I using AI to be better at my work, or am I using it to avoid doing my work?
If those answers are honest, the practice stays healthy.
This guidance document is governed by and subordinate to How I Work With AI, v5. In
any conflict between this guidance and the main framework, the main framework takes
precedence.
See also: Agent Framework for governance of AI agent-specific operations.