Start here

Start Here — Setting Up Your Enrollment Layer

A free companion artifact from The Enrollment Director Field Guide to AI. The guide explains when and how to use it; this copy is yours to adapt for your school. More at haitani.org/field-guides/enrollment-director

Q3 2026 edition · last updated August 13, 2026

What this is. The setup page for a kit that runs the five-session strategic enrollment planning method from chapter 9 of the guide, Market Intelligence and the Strategic Enrollment Plan, inside your own AI account, using your own files. You build a small folder — your Enrollment Layer — and then run five working sessions against it. What you end up with is a strategic enrollment plan that traces back to evidence, and a folder that gets more useful every season you keep it.

You will need a paid account on a capable AI assistant with web search and projects: Claude (Pro or Team) or ChatGPT (Plus or Team) both work. If your school has given you Google Gemini instead, that runs this kit too, with a few adjustments covered in its own section below. Nothing in this kit runs on anyone else's system, and no part of your funnel data leaves your account.

Why the folder comes first

You already hold the context this work needs. The catchment area, the competitor set, the aid philosophy, the families who left and why, the parts of the funnel you would rebuild if you had a free month. That knowledge is what makes an AI session useful, and it is also the thing that has never had a good place to live outside your own head.

An Enrollment Layer is that place. It is a private folder of your own files that you point the AI at when you sit down to work, so every session starts with a partner that already knows your school instead of a blank window you re-explain yourself to.

The six documents below are the ones this particular job needs. A full Enrollment Layer holds more — a season calendar, messaging pillars, a voice guide, your season playbook — and it is built for more than planning. Chapter 5 of the guide, The Enrollment Layer and Workflow Redesign, develops it in full and then rebuilds the office's recurring work around it. It also makes an argument this kit does not: that the folder earns its keep even setting AI aside, because an Enrollment Layer is institutional memory that happens to be machine-readable — written for your successor, and for the associate covering while you are out, as much as for you.

What makes this different from asking questions

Asking a good question and getting a good answer is genuinely useful, and it is most of what AI gets used for. This kit does something different, and the difference is worth understanding before you start, because it is what makes the plan accumulate instead of starting over every time.

Two things are going on.

The AI reads your documents. Once you put your Enrollment Layer files into a project — that is a Project in Claude or ChatGPT, or a Gem in Gemini — the AI reads them at the start of every conversation you have there. You never re-explain your school. Your funnel history, your aid philosophy and the language you actually use are already in the room.

You save what it writes. At the end of each session you ask for a summary, and the AI writes that summary into the chat. Nothing lands in your Enrollment Layer on its own. You read what it wrote, correct what it got wrong, and put the corrected version into your Sessions folder yourself. Some tools offer to export a response straight into a document, which saves you a little typing — but the version that belongs in your folder is the one you have already fixed, and that one only you can make.

So the work divides cleanly: the AI does the reading and the drafting, and the saving is yours. It costs a couple of minutes of copy and paste at the end of a session, and it is what lets Session Two open by reading what you concluded in Session One rather than asking you to say it all again.

Build the folder

Your Enrollment Layer is one folder with six documents in it. Make the folder wherever you already keep your work — your own computer, Google Drive, your school's OneDrive — and call it Enrollment Layer. Inside it, make a second folder called Sessions and leave it empty for now; your session summaries land there as you go.

The six are ordinary documents. Google Docs, Word files, plain notes, whatever you already use every day. There is nothing to install and no particular file format to learn — the AI reads them the same either way.

Start with the School Profile. Everything else can wait, and three of these is already a real Enrollment Layer.

Have the AI write them with you

Six blank documents is a slow way to begin, and you do not have to start that way. Open a plain conversation with your AI before you set anything else up, paste the text below, and answer its questions. It drafts each document, you read it and fix what it got wrong, and you save it into the folder.

I am an enrollment director at a K-12 private school and I am setting up something called an Enrollment Layer: a small set of documents about my school that I will give you at the start of every planning session.

Help me write them, one at a time. There are six:

1. School Profile - school name, type (independent, Catholic, or other), city and state, the grade levels this plan focuses on, tuition, website, total enrollment, and the grades where most new families arrive.
2. Funnel History - three to five years of inquiries, applications, admits, enrolled, and retention by grade. Numbers only, no family names.
3. Aid Philosophy - what our financial aid is for, the discount rate we run, and the tradeoff we have chosen between access and net tuition revenue.
4. Voice and Messaging - our actual website copy, viewbook language, and a recent inquiry response, in the words we really use.
5. Last Season Post-Mortem - what worked last season, what did not, and what I would change if the calendar allowed it.
6. The Trust Map - the moments in our admission process that should always involve a person rather than a system.

For each one, ask me questions one at a time until you have what you need, then write the document in plain text with no formatting symbols, so I can paste it straight into a document and save it.

Never invent a number, a school name, or a figure. If I do not know something, leave it blank and tell me what is missing.

Start with the School Profile and ask me your first question.

Your AI cannot reach into your computer or your Drive to make the folder for you. You make the folder and save what comes back; it does the writing.

What goes in each one

School Profile — the one to write first. If you would rather fill it in yourself, this is all it needs:

SCHOOL PROFILE
School:
Type: (independent / Catholic / other)
Location: city, state
Grade levels in focus for this plan:
Tuition:
Website:
Total enrollment:
Entry points: (the grades where most new families arrive)

Funnel History — three to five years if you have it: inquiries, applications, admits, enrolled, and retention by grade. Numbers only. No names.

Aid Philosophy — what your aid is for, in your own words. The discount rate you run, the one you would defend, and the tradeoff your school has actually chosen between access and net tuition revenue.

Voice and Messaging — paste your real website copy, viewbook language, and a recent inquiry response. The actual language, not a summary of it. Session One reviews this word for word.

Last Season Post-Mortem — what worked, what did not, and what you would change if the calendar allowed it. Write it honestly — nothing in this folder is for circulation.

The Trust Map — the moments in your funnel that will always be a person: the anxious-parent question, the aid conversation, the denial and waitlist messages, the family that has gone quiet. Naming them protects them. Chapter 6 of the guide, Top of the Funnel, is where the trust map is developed as a practice rather than a list: where it lives, who else on the team sees it, and why a trust map written in a calm month is worth more than a judgment call made in a busy one.

Set up the project

In Claude: create a Project, name it Enrollment Layer. Upload your folder files to the project knowledge. Paste the contents of 01 - Standing Instructions into the project's custom instructions.

In ChatGPT: create a Project, name it Enrollment Layer. Upload your files. Paste the standing instructions into the project instructions.

Either way, confirm web search is turned on. Sessions Two and Three depend on it.

If your school has given you Google Gemini

Gemini runs this kit. A Gem is the equivalent of a Project: standing instructions plus a Knowledge section it reads in every conversation, with Google Search built in for Sessions Two and Three. If your school is on Workspace Business or Enterprise you already have it, which is worth checking before you expense a second subscription.

Setting it up. Create a Gem and name it Enrollment Layer. Add your folder files under Knowledge, paste the contents of 01 - Standing Instructions into the Gem's instructions, and confirm search is on.

Point it at Drive rather than uploading copies. If your folder already lives in Drive, the Gem picks up your edits as you make them. This is the one thing Gemini does here that the others do not, and it is closer to what a Layer is meant to be: a folder you keep, rather than a snapshot you re-upload each season.

A Gem's Knowledge section holds ten files. Your Layer is six, and Session Five alone produces three. So keep one running Session Summaries document and append each corrected summary to it instead of saving a file per session. The method reads the same text either way.

Where the folder lives decides who can read it. A Gem built on your school's Drive files sits in your school's tenant, where your administrators can reach it. Google's commitment is that your content stays inside your organization and is never used to train models outside it, which is the right assurance for funnel data and a different question for a candid post-mortem about your own school. If the Post-Mortem is the file you need to be most honest in, keep that one on a personal account, or write it knowing the room it is in.

How the five sessions work

Each session is its own conversation inside that project. You open a new chat, paste the session brief from this kit, and work through it, closing with the summary you correct and save.

Expect the full sequence to take several sessions across a few weeks. That is faster than any honest version of this work has ever been, and slower than doing it badly.

The checkpoint discipline

This is the part that separates a plan from a pile of plausible text.

Every session ends the same way. You ask for the summary, you read it against what you know to be true, and you correct it before you save it. You are the expert on your school; the AI holds no institutional memory you did not give it. A figure it inferred, a competitor it mischaracterized, a sentiment theme it weighted too heavily — you catch those now, or they compound quietly through every session that follows.

The corrected summary is what goes in the folder. The raw transcript does not. This keeps each session focused and keeps the plan accumulating rather than restarting.

The line you hold

This one is yours. The AI does not enforce it for you.

Personas and segments describe markets and guide messaging. You never apply them to an individual applicant, and you never use anything this plan produces to rank, score, or screen a child or a family. That evaluation stays human, in full.

The standing instructions in this kit tell the AI the same thing, and that helps. But an instruction is a line of text that can be edited, replaced, or quietly dropped when someone copies your project. What holds this boundary is you.

Write it down before you start, and say it out loud to anyone on your team who picks up this kit.

What you have at the end

A synthesis report that holds your research against your goals and names where they diverge. A strategic enrollment plan with personas, segments, and a marketing plan built from evidence. A 30-60-90 day action plan, OKRs, a twelve-month content calendar, and a short list of metrics that will tell you within one quarter whether the plan is moving.

And a folder. Next season the second pass is much faster, because the context persists and only the deltas need work.