Skip to content
Concept site. Scenarios and walkthroughs are illustrative examples for higher education, not any campus's actual processes or policy.
Concept site. Illustrative examples only.
Campus Workflows Process-first improvement for higher education Browse scenarios

{{ searchStatus }}

Process first, then Apps Script

Better campus work starts with the process, not the tool.

Before you automate, rethink the work. This site helps higher-education offices decide whether a process should be retired, improved, redesigned or created, then use Google Apps Script to automate the repetitive steps that remain: intake, routing, reminders, records and reports. It includes {{ totalCount }} campus scenarios mapped to the Higher Education Reference Models (HERM), a shared map of what universities do.

New to Apps Script?

Google's built-in way to automate Sheets, Forms, Gmail, Calendar, and Drive

  • Already includedComes with your campus Google account. Nothing to install.
  • Runs on triggersA form submission, an edit, a schedule, or a menu click.
  • Small scriptsUsually a few dozen lines, often drafted with AI help.
Intro to Apps Script Google's docs ↗

The process at a glance

Seven steps, from problem to decision

Select a step to see what to do, what to ask, and who's involved. Apps Script doesn't come up until step 5.

  1. If Retire: stopNothing to build.
  1. If RetireStop here.

↻ Next review: back to step 1

{{ fs.n }}

Step {{ fs.n }} of 7 · {{ fs.kind }}

{{ fs.name }}

{{ fs.body }}

What you do

  • {{ t }}

Questions to ask

  • {{ t }}

You come away with

{{ fs.out }}

Who's involved

{{ fs.who }}

Why this matters

Most campus offices run on forms, spreadsheets, and inboxes. Requests arrive incomplete, nobody acknowledges them, and follow-up depends on who sees the email.

Generic Apps Script tutorials already exist. What's missing is the campus side: small fixes are rarely owned, measured, or retired. This site treats the code as the smallest part of the work.

  • Higher-ed specificEmergency aid, care reports, housing, conduct, and financial aid. Every scenario comes from a campus office.
  • Process before toolEach scenario maps to a HERM capability, so the conversation starts with what the institution needs.
  • Governed, not just builtFERPA, equity, ownership, and retention are part of every scenario.
  • Apps Script where it fitsSometimes the better answer is an existing campus system, or no automation at all.

Explore the site

Build

Apps Script, where it fits

Google Apps Script is built into Google Workspace. Small JavaScript programs read and update Sheets, send Gmail, create Calendar events, fill in Docs, and respond to Forms, with nothing to install. Use it after you've decided how the process should change, and only where it fits.

Runs in Google's cloudScripts run on Google's servers, not your computer. They keep working when your laptop is closed.
Written in JavaScriptThe same language as the web, with ready-made services for each Workspace app.
Lives next to your filesA script can be attached to a Sheet, Doc or Form, so the automation sits with the work it supports.
Included with WorkspaceNo separate license. Your campus Workspace admins control who can use it and what it can access.

What it can reach

Each app has a built-in service. Counts show how many catalog scenarios use each one.

How a script starts

Scripts run when something happens. These are called triggers.

  1. {{ t.name }} {{ t.what }} {{ t.count }} scenarios

Start from Google's samples

Google publishes working Apps Script projects. Adapting one beats starting from a blank file. Each is matched to the pattern it shows and the catalog scenarios it could start.

Quickstarts · working code in about five minutes Solutions · complete projects for a real business problem Codelabs · step-by-step tutorials
  1. {{ g.title }} {{ g.shows }} {{ g.apps }}

From the Apps Script samples library. More on GitHub. For a first try, use the automation quickstart.

Know this before you start

It runs as a person
A script acts with the permissions of whoever authorized it or installed the trigger. If that person leaves, it can stop working.
It asks for access
The first run shows an authorization screen listing what the script can touch. Read it; it's your data map.
It has limits
Google sets daily quotas, such as how many emails a script can send, and a time limit on each run. Plan for busy periods.
Your admins set the rules
Campus Workspace admins can allow, restrict, or review Apps Script. Check your local policy first.

Is Apps Script the right method?

This is step 5 of the seven steps. Use the lightest method that works.

A good fit when

  • {{ t }}

Not a fit when

  • {{ t }}

If it's not Apps Script

  • {{ x.name }}{{ x.when }}

On Microsoft 365 or another platform, see the translation table. Building custom software with AI? Campus AI Development ↗ picks up from here.

To try it, open any Google Sheet and choose Extensions › Apps Script. Before building for real, read Build it safely; to draft the code, see Build with AI prompts.

Build

The same ideas on other platforms

This site is written for Google Apps Script, but the process thinking carries over. On Microsoft 365, a campus system, or another automation tool, use this page to translate. Rule of thumb: build where the data already lives.

Translation table

Swipe sideways on small screens.

Google Workspace Microsoft 365 Cross-platform tools Campus systems
{{ r.k }} {{ r.g }} {{ r.m }} {{ r.x }} {{ r.c }}

Strengths and watch-outs

{{ p.name }}{{ p.tools }}

Best for

{{ p.best }}

Watch for

{{ p.watch }}

On another platform, use this table to translate, and ask an AI assistant to adapt a scenario's prompt. The rules on Build it safely apply everywhere.

Build

Build Apps Script with AI prompts

You don't need to be a developer. An AI assistant can draft, explain, and fix Apps Script if you describe the process clearly. You stay responsible for what runs; this method keeps you in control.

Every catalog scenario has a starting prompt. Use this page to go further: review, fix, harden, and hand off.

The method

  1. {{ s.n }}
    {{ s.t }}{{ s.b }}

What a good prompt includes

Six parts. Most weak results come from a missing Data or Rules section.

{{ a.n }} · {{ a.t }} {{ a.b }} {{ a.ex }}

Prompt library

Copy, fill in the brackets, and paste into your assistant. Open any row. These are starting points: adjust them to your process, data, and campus policies, and expect to refine them as you go. Results vary by assistant.

{{ p.name }}{{ p.when }}
{{ p.text }}

Ground rules

  • {{ x.t }}{{ x.b }}

Before launch, read Build it safely. Building something bigger than a script? Campus AI Development ↗ covers review levels for AI-built software.

Step 1 · Map it

From institution to task

Every process sits in a chain, from the institution down to a single task. Apps Script works at the bottom, on tasks. Ownership and evaluation sit higher up, so map the chain before you build.

What is HERM?

A shared map of what a university does

The Higher Education Reference Models (HERM) are free, openly licensed models of how institutions work. CAUDIT, the Council of Australasian University Directors of IT, maintains them, and EDUCAUSE, UCISA, and campuses worldwide use them.

This site uses the Business Reference Model. Its capabilities name what the institution must be able to do, such as Student Support, without naming an office or a system. That makes them neutral ground when several offices share the work.

Why it helps

  • A common language. Offices that call the same work by different names can see it's the same work.
  • Fewer duplicates. Tag each tool with a capability, and five versions of one fix become visible.
  • Links to planning. IT and enterprise architects already plan against these capabilities.

Each capability has a code. In the catalog, you'll see pairs like BC052 Student Support or BC248 Financial Aid Management.

Explore the HERM Business Reference Model ↗
  1. Level {{ l.num }} {{ l.name }} {{ l.def }}
    Example {{ l.example }}
    What happens here {{ l.role }}

Explore the chain

Pick a domain, a capability, and a workflow to see its processes and tasks.

Domains

Capabilities

Workflows

Processes and tasks

  1. {{ p.num }}{{ p.name }}
    • {{ t.kind }}{{ t.text }}
{{ hNote }} {{ hWfLink }} →

Capability codes come from the scenario catalog. The grouping of capabilities into domains is illustrative; check it against the HERM version your campus uses. No scenarios fall in the Research family yet.

Step 6 · Pilot it

Scenario catalog

{{ totalCount }} campus processes where Apps Script can help, each mapped to a Higher Education Reference Models (HERM) capability. Filter for work like your office's, then select a scenario for its build steps, a ready-to-copy AI prompt, its risks, and who to involve.

Scenarios are illustrative. Effort estimates assume someone comfortable with Sheets and basic scripting.

Complexity
Data sensitivity
Pattern
What the patterns mean
{{ p.name }}
{{ p.what }}
Watch for: {{ p.watch }}
{{ resultLabel }} Pattern: {{ patternLabel }}

No scenario matches all of these filters. Clear one to see more.

  1. {{ s.idLabel }} {{ s.code }} · {{ s.cap }} {{ s.trigger }} → {{ s.action }} {{ s.complexity }} {{ s.effort }} {{ s.sensitivity }} sensitivity Starting point View steps and prompt →
← {{ topBackLabel }}

Scenario {{ d.idLabel }} · {{ d.code }} {{ d.cap }} · What's a HERM capability?

{{ d.trigger }}

→ {{ d.action }}

{{ d.sensitivity }} data sensitivity Recommended starting point
Full walkthrough available, with before and after, outcomes, and a decision rule. Read it →
{{ f.k }}
{{ f.v }}

Primary value

{{ d.value }}

Key risks

{{ d.risks }}

Build notes

  • {{ n }}

Institutional considerations

  • {{ n }}

Build it

From idea to a working pilot

The baseline from step 1 is what Evaluation measures against. Google's menus change, so check the current trigger documentation if a label doesn't match.

  1. {{ b.num }}
    {{ b.title }}{{ b.body }}
Prompt for an AI assistantA starting point, not a finished spec. Edit it to match your Sheet, your process, and your campus policies. Use only a campus-approved assistant, and never paste real student data.
{{ d.prompt }}

Treat the output as a draft: read it, test it with fake data, and have someone review it before it touches real data. For prompts that explain, fix, and review code, see Build with AI prompts. Before launch, read Build it safely.

Who to involve
Start from a Google sample
Patterns in this scenario

Select one to see other scenarios that use it.

Run

Build it safely

A useful automation and a liability can run the same code. The difference: visible settings, a log of every action, and an owner after the builder moves on. Start with the rules, then read what happens when they're skipped.

The rules

Open any row. Adapt them to your campus's IT, security, and data policies.

{{ n.num }} {{ n.title }}{{ n.sub }}
  • {{ x }}

Campus scenarios gone bad

{{ riskCount }} composite stories of small automations gone wrong. None needed bad intentions, only a skipped rule. Open one for warning signs, prevention, and recovery.

Before launch, run a pre-mortem: imagine it's a year from now and this went badly. Which story was it?

{{ g.name }}

{{ f.num }} {{ f.t }}{{ f.story }}

Warning signs

{{ f.signs }}

Prevent it

{{ f.prevent }}

If it already happened

{{ f.recover }}

More in {{ f.linkLabel }} →

The stories are composites, not reports about any specific institution. Have one to add? See About for how to share it.

Workshop toolkit

Work through one process in one hour

A ready-to-run workshop for a process designer and an office. You'll leave with a map of the real process, an agreed type of change, and a named owner. Tools come up only in the last ten minutes.

Agenda: 60 minutes

Block widths show how the hour is split.

  1. {{ a.n }} · {{ a.min }} min{{ a.range }}

Before the session

  • {{ t }}

What to bring

  • {{ t }}

Who's in the room

{{ r.t }}
{{ r.b }}

Facilitator script

Suggested wording for each block. Say it in your own words; the questions matter more than the phrasing.

  1. {{ c.n }} · {{ c.min }} min

    {{ c.name }}

    “{{ c.say }}”

    Ask

    • {{ q }}

    If the room gets stuck: {{ c.stuck }}

Exercises

The agenda uses six of these. Five whys and the jobs-to-be-done sentence are optional, for when the room gets stuck.

  1. {{ x.num }} · {{ x.min }} min

    {{ x.name }}

    {{ x.why }}

    Steps

    1. {{ t }}

    You come away with

    {{ x.out }}

    Facilitator tip

    {{ x.tip }}

Worked example

{{ example.title }}

{{ example.office }}. What each block of the hour produced. Details are illustrative.

  1. {{ x.n }}{{ x.s }} {{ x.o }}

See the full emergency aid walkthrough for the before-and-after.

Handout

One-page commitment

{{ f }}
Process owner signature Date

Next: bring the commitment to the seven steps, or find a matching scenario in the catalog.

Summary email

Send within 24 hours, while the room still remembers. It mirrors the commitment, so you can copy straight from it.

Template
{{ summaryText }}

Fast track

Just want to build? Start here.

No mapping, no workshops. Pick a simple, low-risk starter, follow five rules, and copy the prompt. If a warning sign at the bottom shows up, switch to the full process.

1

Pick a starter

{{ fastCount }} scenarios are simple to build and use low-sensitivity data. Each one opens with build steps and a ready-made prompt.

2

Five things you can't skip

  1. {{ m.t }}{{ m.b }}
3

Copy the prompt and build

Open your starter. Its Build it section has setup steps and a starter prompt for a campus-approved AI assistant; adjust it to your setup. New to Apps Script? Read the intro. New to AI prompts? See Build with AI prompts.

Know when to stop

Switch to the full process if

Fast-track tools grow. When one of these happens, it's no longer a quick fix. See what goes wrong when people skip this.

  • {{ x.t }}{{ x.b }}

Step 4 · Involve people

Every change has more than one owner

Change succeeds when the right people weigh in at the right time. A low-risk change may need only the builder and the process owner. Sensitive data or several offices need more voices.

Who's involved, and when

● leads or signs off · ○ consulted · Select a role for details.

Role {{ p }}
{{ r.name }} {{ m.sym }}
  1. {{ r.num }}

    {{ r.name }}

    {{ r.who }}

    Most involved: {{ r.most }}

    Questions to ask

    • {{ q }}

    Responsible for

    {{ r.owns }}

    Warning sign

    {{ r.flag }}

Each scenario page lists the roles to involve, based on its data sensitivity and who it reaches. Scale review to the risk; a reminder email shouldn't need a committee. Process designers: the Workshop toolkit is built for you.

Ways to change a process

Decide the change before you pick the tool

Change comes in six depths. Teams that jump straight to automation usually automate the process they already have, problems included. Choose the depth first; the method comes later.

Which change is this?

Answer the questions from the top. Select an outcome for details.

  1. Question {{ t.n }}{{ t.q }}

    {{ t.side }}
    {{ t.down }}

Augment can follow any of the others. After you improve, redesign, or create a process, ask whether tools would help.

{{ ms.num }} · {{ ms.depth }}

{{ ms.name }}

Also called {{ ms.aka }}

{{ ms.idea }}
Choose it when
{{ ms.when }}
On campus
{{ ms.campus }}
How to start
{{ ms.start }}
Role for Apps Script
{{ ms.script }}
Read the full entry ↓
  1. {{ m.num }} · {{ m.depth }}

    {{ m.name }}

    Also called {{ m.aka }}

    {{ m.idea }}

    Choose it when
    {{ m.when }}
    On campus
    {{ m.campus }}
    How to start
    {{ m.start }}
    Role for Apps Script
    {{ m.script }}

Step 3 · Fix it first

Fix the process, then automate it

Automating a broken process makes it break faster. These disciplines show where work gets stuck. Most need only a whiteboard and an hour with the people who do the work.

Ask in this order

Before automating any step, ask these in order. They come from industrial engineering, and they still work.

  1. {{ e.num }}
    {{ e.name }} {{ e.q }} Example: {{ e.ex }}

Start from the symptom

Know the problem but not its name? Start here.

The disciplines

  1. {{ p.num }} · Best at the {{ p.level }} level

    {{ p.name }}

    {{ p.idea }}

    On campus
    {{ p.campus }}
    Try this
    {{ p.tryit }}
    Where automation fits
    {{ p.auto }}

Levels match the institution-to-task chain. Fixes at the workflow level usually beat fixes to a single task. To run these with an office, use the Workshop toolkit.

← {{ topBackLabel }}

Walkthrough {{ demo.num }} · HERM {{ demo.caps }}

{{ demo.headline }}

Process: {{ demo.process }}

The idea in one picture

Six stages, one register. People make every access decision; the script records it and prompts the next step.
  1. {{ c.n }} {{ c.name }} {{ c.who }} Script: {{ c.script }} Register: {{ c.rec }}

↺ A role change at step 05 sends the person's access back to review at step 04.

Today: where it breaks

The process as it usually runs, with the gap at each step.
  1. {{ p.step }} · {{ p.who }} {{ p.text }} ✕ {{ p.gap }}

The ideal: how it should run

Dark steps stay with people; light steps are handled by tools.
  1. {{ f.step }} · {{ f.tool }} {{ f.text }}

What could be automated

Each stage, split into what a script can do and what stays with people.

{{ s.stage }}
Script
  • {{ x }}
Nothing at this stage
People
  • {{ x }}
Nothing at this stage

With more integration

Next steps once the basic version is trusted. Each needs the owner of the connected system to agree.

  • {{ x }}

Current friction

  • {{ x }}

The automation

{{ demo.augmentation }}

{{ demo.human }}

Before and after

{{ r.stage }}
Before{{ r.before }}
After{{ r.after }}

Cost, use, and outcomes

How this walkthrough is judged, using the evaluation approach.

Cost

{{ demo.cost }}

Use

  • {{ x }}

Outcomes

  • {{ x }}

Decision rule

{{ demo.decisionWhen }}

{{ x.verb }}{{ x.when }}

Institutional considerations

  • {{ x.k }} {{ x.v }}

Running your own pilot? Step 7, Measure and decide, shows how to judge it.

Reimagine with AI

Don't just speed up the work. Design it again.

Most of this site removes waste from processes that already exist. This page asks a different question: what would the work look like if you designed it knowing software can read, sort and draft at scale? Some processes exist only because nobody could read everything or reach everyone. When that limit goes, the process can change shape. One rule doesn't change: people make the decisions.

1. Three levels of change

One process, course feedback, changed three ways. Only the last changes what the office can do.

AutomateCopy survey results into a report each term.The same report, produced faster.
Improve with AIAI summarizes the comments for the report, and a person checks the summary.A better report, still read after the term ends.
ReimagineEvery comment is read as it arrives. Fixable issues go to the office that owns them while the course is still running, and students hear what changed.Problems fixed during the term, not noted after it.

2. What AI makes possible

Five kinds of work that were impractical for a small office. Each is a starting point for a new workflow, not just a faster one.

  1. 01Read everythingBefore: Staff read a sample because there's no time to read it all.Now possible: Every comment, email and open-text answer is sorted and summarized, and people read what matters.
  2. 02Reach people firstBefore: Offices wait for students to ask.Now possible: Students who match an agreed pattern, such as a missing form near a deadline, get a personal message before they fall behind.
  3. 03Check at the doorBefore: Missing items are found weeks later, during review.Now possible: Each submission is checked when it arrives, and the student learns what's missing that day.
  4. 04Prepare before people meetBefore: Staff open a case cold and read the history in the meeting.Now possible: A short case summary is ready before the meeting, with sources linked so staff can check it.
  5. 05Answer at any hourBefore: Simple questions wait for office hours.Now possible: Drafted answers come from approved policy pages, with a clear handoff to a person for anything else.

3. AI at every stage, people at the decision

Most workflows run through the same seven stages. AI can help at six of them. The decision stays with a person.

  1. 01IntakeAIChecks the submission for missing items and pulls out key detailsPersonDesigns the form and decides what counts as complete
  2. 02SortAISuggests a category, urgency and ownerPersonConfirms the route; handles anything unclear
  3. 03PrepareAIDrafts a case summary with sources linkedPersonChecks the summary against the record
  4. 04DecideAINothing. No AI at this stagePersonMakes the decision and records the reason
  5. 05ActAIDrafts the message or documentPersonEdits and sends it
  6. 06Follow upAISpots cases with no reply and drafts remindersPersonChooses when to step in personally
  7. 07LearnAIFinds patterns across cases and commentsPersonDecides what to change in the process

Decisions about eligibility, awards, holds, grades or conduct always stay with a person. That matches the limits on For leaders.

4. Example: transfer credit

Today the process is limited by how fast evaluators can look up courses. Reimagined, they spend their time on the hard matches.

Today

  1. 01 · StudentUploads a transcript and waits✕ No date for a result
  2. 02 · EvaluatorLooks up each course by hand in a spreadsheet✕ Weeks of backlog at peak
  3. 03 · FacultyGets an email asking about unmatched courses, with no syllabus attached✕ Days lost chasing course descriptions
  4. 04 · StudentRegisters before the evaluation is done✕ Repeats courses or loses credit

Reimagined

  1. 01 · AIReads the transcript at upload and proposes a match for each course from the equivalency table
  2. 02 · EvaluatorApproves or changes each proposed match
  3. 03 · FacultyReviews only unmatched courses, with the course description already attached
  4. 04 · Student and advisorGet a plain-language result before registration
  5. 05 · RegisterEach approved match is added to the equivalency table for next time

What changed

  • Evaluators approve matches instead of looking them up
  • Faculty see only the courses that need judgment
  • Students get a result before they register
  • The table improves with every approved match

What it needs

  • An AI service your campus has approved for student records
  • A named Process owner in the registrar’s office
  • A test on last term’s transcripts before launch
  • Match accuracy checked for each sending institution

This is a level 2 or 3 workflow on the right-sizing scale. Apps Script can run a pilot inside Workspace; a production version usually belongs with the student information system.

5. Run a reimagine session

Use this in place of the improvement steps in the workshop when the process is limited by reading, sorting or reach.

  1. 01 · ThinkStart from the outcomeName the person served and what they need, not the steps you do today.
  2. 02 · ThinkList the limitsWrite down what the current process was built around: “we only read a sample,” “we wait for them to ask.”
  3. 03 · ThinkAsk which limits AI removesFor each limit, ask what the work would look like without it.
  4. 04 · DesignSketch the new flowMark every decision point. Each one gets a named person.
  5. 05 · ProveTest on past casesRun the new flow on last term’s cases, with no live data, and compare.
  6. 06 · ProvePilot with full reviewA person checks every AI output until error rates are known and agreed.
  1. 01 · ThinkStart from the outcomeName the person served and what they need, not the steps you do today.
  2. 02 · ThinkList the limitsWrite down what the current process was built around: “we only read a sample,” “we wait for them to ask.”
  3. 03 · ThinkAsk which limits AI removesFor each limit, ask what the work would look like without it.
  4. 04 · DesignSketch the new flowMark every decision point. Each one gets a named person.
  5. 05 · ProveTest on past casesRun the new flow on last term’s cases, with no live data, and compare.
  6. 06 · ProvePilot with full reviewA person checks every AI output until error rates are known and agreed.

↺ If the test at step 05 falls short, go back to step 04 and sketch again.

6. Guardrails

Build in from the start

  • Use only AI services your campus has approved for that data
  • Tell students and staff where AI is used, and how to reach a person
  • Link every summary or draft to its sources
  • Check results across student groups
  • Keep a manual path for when the AI service is down

Not ready to reimagine when

  • The process owner and decision points aren’t agreed
  • You can’t measure how the process performs today
  • No approved AI service covers the data involved
  • Nobody has time to review AI output during the pilot

The data and code rules are on Build it safely. For workflows where AI takes actions, not just drafts, see Campus AI Agents ↗.

For leaders

Run automation as a portfolio

For innovation, optimization and operations leaders. The rest of the site helps one office fix one process. This page covers the decisions above that: which processes to work on, who owns what, what evidence to ask for, and when a workflow has outgrown Apps Script. Same rule as everywhere else on the site. Don't automate a bad process.

1. Choose where to start

Look at every candidate process before picking one. Score each on five factors, then place it on a grid of impact against risk.

Volume
How many requests or cases a term, and how many people handle them.
Time and cost
Staff hours spent, plus how long people wait.
Effect on people
How much the delay or confusion affects students and staff.
Data sensitivity
Low, medium or high, using the same scale as the catalog.
Cross-office dependency
How many offices touch it, and how many rely on it.
High impact, low riskStart here. Good fits for Apps Script and the fast track.
High impact, high riskWorth doing, with the full seven steps and the right reviewers.
Low impact, low riskLet offices handle these themselves, within the guardrails.
Low impact, high riskUsually not worth it. Consider retiring the process instead.

Tag every candidate with its HERM capability. Grouped by capability, the list shows where offices are duplicating each other and which capabilities nobody has touched. For a fuller prioritization method, see Campus AI Compass ↗.

2. Right-size the solution

I'd keep most campus workflows at level 1. Move one up only when it meets the criteria below.

  1. Level 1 · This site's focusProcess plus Apps ScriptPeople do the work, with triggers handling acknowledgments, routing, reminders and records. One office owns it.
  2. Level 2AI-assisted stepsAI classifies, summarizes or drafts within a step, and a person approves every output before it is used.
  3. Level 3Orchestrated across systemsThe workflow connects to the SIS, CRM or ERP through approved integrations, with review points between stages. This belongs with IT and enterprise systems.
  4. Level 4Governed agentsAgents plan and act within set limits, with a named owner, a list of permitted actions and an audit trail. See Campus AI Agents ↗.

Move up a level when

  • Volume or quotas are outgrowing the script
  • Several offices depend on it
  • It needs data from enterprise systems
  • The process is stable, agreed and measured

Don't make it agentic when

  • It decides eligibility, awards, holds or discipline
  • Nobody can explain or reverse what it did
  • The process itself isn't agreed yet
  • A simple rule or trigger would do the job

Building custom software is covered by Campus AI Development ↗. Connecting to enterprise systems is an IT architecture decision: prefer approved integrations that work around existing systems over replacing them, and settle identity, access and audit logging before go-live.

3. Set up the operating model

Office-level fixes add up only when someone owns the standards. Settle these four questions once, centrally.

Who owns what
A central team or a network of local leads sets standards and keeps the inventory. Process owners own outcomes. Named automation stewards own each tool. A small central team with a lead in each division is a practical middle ground.
How requests come in
One short intake form, triaged weekly with the five scores above. Every request gets an answer: fast track, workshop, enterprise queue or not now.
What gets reused
Grow a shared library from the scenario catalog and the patterns. Before anyone builds, check whether another office already has the same tool.
Guardrails for staff builders
The rules on Build it safely, an inventory every production script must be listed in, and a review date on each one. With those in place, staff can keep building without leaving IT a pile of unowned tools.

People need skills too. Process owners need the one-hour workshop, builders need the prompting method, and leaders need enough AI fluency to ask the questions below.

4. Ask for evidence

Require a baseline before launch and a decision at every review. Across the portfolio, track:

  • Hours returnedAnd what staff did with them.
  • Cycle timeFrom request to outcome, against the baseline.
  • Exception rateCases the workflow couldn't handle.
  • AdoptionShare of cases that go through the new path.
  • Maintenance loadHours spent fixing and updating tools.
  • FairnessOutcomes across student groups.
Consolidate or retire whenseveral tools serve the same capability, maintenance takes more time than the tool saves, a tool has no owner, or an enterprise system now does the job.

Per-workflow measures are on Evaluation. For the full method, see the AI Cost, Use, and Outcomes Framework ↗.

Questions to ask any proposal

  1. Which HERM capability does this serve, and does another office already have a tool for it?
  2. Was the process fixed before anyone automated it?
  3. What's the baseline, and when is the review?
  4. Who owns it if the builder leaves?
  5. Where does a person approve, and where do they only review exceptions?
  6. What would make us retire it?

Step 7 · Measure and decide

Keep it, fix it, or turn it off

Every change ends in a decision, not a shrug. Record a baseline before launch, review at a set point, and compare cost, use, and outcomes. The approach follows the AI Cost, Use, and Outcomes Framework.

Three moments

Most of the work happens before launch. Skip it, and there's nothing to compare against later.

  1. {{ m.when }} {{ m.name }}
    • {{ x }}

What to measure

Pick one or two measures in each column. Measure them the same way before and after.

{{ c.name }}

{{ c.question }}

  • {{ x }}

Five possible decisions

Before launch, write down what result would lead to each one.

  1. {{ x.verb }}{{ x.when }}

Feed it back

Every scenario keeps a Log tab, and it shows how the process runs in practice. At each review, read it for signals and take them back to step 3, Fix it first:

  • Steps where cases still wait the longest
  • Exceptions that keep coming up, which may need a rule or a form change
  • Fields that are often blank or corrected later
  • Reminders that fire again and again for the same step

Leaders: across many workflows, this becomes portfolio evidence. See For leaders.

See it applied in the emergency aid walkthrough. Once you decide, set the next review date and start again at step 1.

About

About this site

Campus Workflows is a concept site. It shows how a campus office can improve a process, then use Google Apps Script where automation fits. Four ideas run through it:

  • Decide how a process should change before choosing a tool.
  • Start from a named capability, not an office or a system.
  • Keep judgment about people with people.
  • Measure each change, then keep, fix, or retire it.

What's here

Build it
Apps Script basics, building with AI prompts, {{ totalCount }} scenarios with build steps and starter prompts, and a fast track for low-risk starters
Keep it healthy
When it goes wrong and evaluation
Run a session
The workshop toolkit: a one-hour agenda, facilitator script, exercises and handouts
Other tools
Equivalent platforms for campuses on Microsoft 365 or elsewhere

Capability codes and names follow the Higher Education Reference Models (HERM) Business Reference Model. Scenarios, estimates, and walkthroughs are illustrative, not any institution's actual processes.

Working notes

Created by
Joe Sabado
Feedback
Questions, corrections or a campus story to share: joepsabado@gmail.com
Independence
This is an independent project. It is not associated with, endorsed by or written on behalf of my employer, and the views here are my own.
Related work
Campus Enterprise Model ↗: HERM Business Reference Model and capability mapping Campus AI Compass ↗: deciding which AI and automation work to prioritize Campus AI Agents ↗: controls for governed AI agents Campus AI Development ↗: how to build software with AI, from vibe coding to governed development AI Cost, Use, and Outcomes Framework ↗: how to judge whether an AI or automation investment is worth keeping
Last updated
November 2026 · {{ totalCount }} scenarios, 2 walkthroughs, workshop toolkit

Accessibility

This site aims to meet WCAG 2.2 AA. Everything works with a keyboard, color is never the only signal, and pages reflow at 320px wide and 200% zoom.

Page not found

That page doesn't exist. Go home or browse the catalog.