Align Your Program With Industry AI | ACBSP 2026

One evidence base. Two ways to use it.

One checked workbook runs everything. The dashboard is the program view; the Syllabus Check is the course view. Every output is a draft for human judgment.

AI Field Scan dashboard 42 use cases across 7 business disciplines, each with what students need to know and do, what the human adds, a classroom idea, and every source with its quote. 156 of 156 sources opened and checked, October 13, 2026. Open the dashboard →
Syllabus Check Drop in a syllabus, choose the field and course level, and get what the course already does well, plus up to three assignments. You'll need to sign in to Claude to run it. Open the Syllabus Check →
Build Your Own AI Field Scan (guide) The full method, step by step, with every prompt. Works in ChatGPT, Copilot, Gemini or Claude. Open the guide →

Using the Syllabus Check

Steps: drop in the syllabus (Word, PDF or text) or paste the text; choose the field and course level; click “Check this syllabus” and allow it to run. Start with “What's already working,” then the assignment cards.

Privacy: nothing you drop in is stored. Ask the instructor before checking their syllabus, and follow your institution's rules on AI tools.

Limits: it reads the syllabus only, not what happens in class, and an instructor may already cover something without listing it. It flags any quote it can't find word for word. Its reports are drafts for the instructor's judgment.

How to read the dashboard

AI Field Scan dashboard overview: summary tiles, readiness by discipline, and the use case table — **The program view.** 42 use cases, 6 in each of 7 disciplines. Every source was opened and the exact quote found on the page. Filter by discipline, teaching readiness, or evidence tier. At UW-Platteville, it's one input to our assessment conversations, school-wide and in each functional area.

The dashboard filtered to HR, showing six use cases with reported gains and K, S, A skill lines — **One discipline.** Each row shows the use case, the reported gain, and what students need to know and do: K for knowledge, S for skills, A for abilities. At least one line in every use case is about checking or overseeing the AI. Figures appear only when a source published them, and vendor figures are labeled.

Details for one use case: what the human adds, a classroom idea, and sources with quotes — **One use case up close.** Details opens what the human adds, a classroom idea in four moves, and every source with its exact quote and the date it was checked. (In this example, “the UW study” is the University of Washington.)

The method

AI does the legwork. A person makes the calls. Nothing counts on the AI's word alone: if a source can't be opened and the exact words found, it doesn't count.

What “checked” means

Flowchart of how the AI Field Scan is built: find, check, write and decide, then the dashboard, re-checked every term — AI does the searching and drafting, software checks every source and number, and a person makes every decision that matters.

The method, step by step

Step	What happens	Who
1. Pick the use cases	Name 5 or 6 concrete ways AI is used at work in each field, such as resume screening or invoice coding	You, with AI
2. Decide what counts	Set which sources count and how vendor sources are labeled	You
3. Gather	AI finds sources for each use case, with the exact quote and any published figure	AI
4. Check	A checking program opens every source and finds the exact words (or you do, with Ctrl+F); a second AI tries to disprove each item	Automatic check, then you spot-check
5. Write what students need	For each use case: knowledge, skills and abilities, what the human adds, and a classroom idea	AI drafts
6. Decide	You approve each field before anyone uses it	You
7. Use it	AI reviews a syllabus against the approved use cases and drafts assignments	AI, you judge

Source tiers

Tier	What it is	How it's used
A	Research, government, regulators	Use freely
B	Named research firms and industry surveys	Use, naming the publisher
C	Company examples, news reports, and vendors	Use, labeled; a vendor's own number is marked “vendor”
Not kept	Anything you could not open, or whose words you could not find on the page	Never used

Every assignment, four moves

Use the AI
→
Check it
→
Improve it
→
Prove what the human added

The four method prompts (gather, the skeptic, what students need, and review a syllabus) are in method-prompts.md and in the guide.

Take the evidence base. Build the version that fits you.

One workbook and two prompts. The workbook holds the checked evidence; the dashboard prompt turns it into a program view, and the Syllabus Checker prompt applies it to one course.

AI Field Scan starter workbook All 42 checked use cases and 156 sources, a Summary sheet by field, yellow rows for adding your own, and a “Check my workbook” sheet that flags mistakes as you edit. The “Start here” sheet walks you through it. Excel
Dashboard prompt Attach the workbook to a new chat in an AI tool that can make a web page, paste the prompt, and check a few rows against the workbook. Text
Syllabus Checker prompt Paste it as the instructions for a Claude Project, custom GPT, Gemini Gem or Copilot agent with the workbook uploaded. Then attach a syllabus and name the field and level. Text
The method prompts To research your own fields from scratch: gather, the skeptic, what students need, and review a syllabus. Text

Make it yours and check it works

Keep what fits. Delete the use cases you don't need, and their sources (same ID). Or delete them all and start fresh.
Add your own in the yellow rows at the bottom of the Use cases and Sources sheets. For more rows, copy a yellow row and paste it below.
Pick, don't type. Field, use case ID, Kind and the Yes/No columns are pick-lists, so labels always match. To add a new field, type its name in a yellow cell on the Summary sheet first.
One knowledge, skill or ability per line in a cell (Alt+Enter on Windows, Control+Option+Return on a Mac).
Mark a quote as found only after opening the page and seeing the words.

Check it works

The “Check my workbook” sheet runs seven checks as you edit: duplicate IDs, missing or unknown fields, use cases with no sources, sources whose ID matches no use case, sources with no link or quote, sources not marked as checked, and figures with no details. Each shows OK or Fix, with how to fix it.
The dashboard's four tiles should match the Summary sheet: use cases, fields covered, sources checked, and use cases with a published figure.
Open two or three Details pop-ups and compare them with the workbook.
For the Syllabus Checker, replace the workbook in the project and test it on a syllabus you know well.

Both prompts work with a customized copy: they skip empty rows and use whatever fields are in the workbook.

A test dashboard built from a customized practice copy of the workbook, including a made-up Healthcare management (TEST) field — **Example only.** Built from a customized copy of the starter workbook, with Finance removed and a made-up field, “Healthcare management (TEST)”, added. Its numbers are test data; the real dashboard has 42 use cases and 156 sources.

What you need to build this

The easy kit works in any AI chat tool. To make it work like ours, with automatic source checks and shared pages, the AI setup needs the abilities below. You don't need to code: the AI writes and runs the code.

Your AI setup needs to	Why	Easy kit	Full version
Search the web and open pages	To find sources and read them	Yes	Yes
Read files you upload	To use the workbook and syllabi	Yes	Yes
Keep saved instructions (a skill, project, custom GPT, Gem or agent)	So it follows the same method and honesty rules every time	Yes, for the Syllabus Checker	Yes
Make a web page	To build the dashboard	Yes	Yes
Run code that can reach the internet	The automatic check: open every source, find the exact words, check every number	No: check by hand with Ctrl+F	Yes
Keep files in a project folder with version history (such as GitHub)	Keeps the data, the programs and a record of every change	No: the workbook is enough	Yes
Publish and share pages	So colleagues can open the dashboard and the Syllabus Check	Optional	Yes
Build a page that can call the AI	The live Syllabus Check, where anyone drops in a syllabus	No: use the chat version	Optional
Run helpers in parallel	Research several fields at once	No	Optional
Run on a schedule	Re-check every source each term without being asked	No	Optional

Not tools, but just as necessary:

A person who approves each field, and a place to record it.
The institution's OK to use the tool, especially before any syllabus goes into it. Ask the instructor first.
An account level that includes these abilities, often a paid plan.
Comfort with a spreadsheet. No programming experience is needed.

How to test a tool: ask it to open one source from the workbook, using code, and find the quote's exact words on the page. Many chat tools run code offline. If it can't open the page, use the easy kit and check sources by hand.

What we used: Claude Code (claude.ai/code) with a GitHub project. It has every ability in the table.

What's in the workbook

Use cases sheet (one row per use case)

Column	What it holds
ID	A short ID, such as hr-01
Field	Picked from a list: the seven fields, plus any added on the Summary sheet
Use case	The use case's name
What AI does	One or two plain sentences
Published figure	The headline figure, if a source published one
Figure details	The figure as the source states it, with who measured what and when
Vendor figure?	Yes if the figure comes from a vendor's own data
Knowledge students need	“Knowledge of…” lines, one item per line
Skills students need	“Skill in…” lines, one item per line
Abilities students need	“Ability to…” lines, one item per line
What the human adds	Why a person stays in the loop
Classroom idea	An activity built on the four moves
Sources	Count of the use case's sources (formula)
Sources checked	Count of sources whose quote was found (formula)

Sources sheet (one row per source, matched to use cases by ID)

Column	What it holds
Use case ID	Picked from a list of the IDs on the “Use cases” sheet
Publisher	Who published it
Title	The source's title
Link	The URL
Published	Publication date
Kind	Research, Survey, Government, Rule or standard, Case study, News or Vendor (pick-list)
Vendor?	Yes if the publisher sells the product involved
Quote (exact words)	The words copied from the page
Quote found on page	Yes if the quote was found on the page
Checked on	The date it was checked

What checking caught

3 of 41

Our first AI-built dataset had 41 use cases. It looked polished. When we re-checked it source by source, only 3 came through with their numbers intact. In 25, at least one headline number was not in the source it cited. In 6 more, the number was on the page but came from an untraceable roundup or measured something else, and in 5 the cited page could not be opened. Each number was replaced with a published one or removed.

What the AI said	What the source actually says
Warehouse robots cut labor costs 65%	The linked page was dead, and the robot maker's own site says 50–90%, not 65%
AI answer summaries caused a “47% decline” in clicks, citing Pew	Pew reported click rates of 8% and 15%; the 47% was the original author's own calculation
An audit standard (PCAOB AS 1105) is an “AI rule”	Its Federal Register notice says it does not address “other technology applications used in audits (e.g., blockchain or artificial intelligence)”
Deloitte reports “92% accuracy”; McKinsey reports “up to 25% faster decision cycles”	The Deloitte figure is not in the cited release. The McKinsey page could not be opened, and no other copy of that figure was found
The DOL AI Literacy Framework covers bias	The full notice never mentions bias, fairness or AI agents
A study found AI resume screeners disadvantage older applicants	It found post-trained models 3.6% less likely than their own base versions to call back older applicants: a narrower claim

None of these would have been caught by reading the AI's answer. Every one was caught by opening the source.

Using AI to Align Your Program, Courses, and Syllabi With Industry AI Use