Skip to main content
How Can We Help?Contact

Education

Reach the student who is slipping by week three, and give the board and the state the same numbers.

We join what your student, learning, aid, advising and finance systems each know. Advisors and teachers see who needs a call while there is still time, and enrollment, retention and completion are counted once for the board, the state and accreditors. Every AI tool that touches student records gets an owner and a privacy review before students meet it.

Talk through your student and institutional data See three results that transfer

Universities · Community and technical colleges · School districts · State systems and consortia

A campus hall with a clock tower behind a lawn; an application, a student reached by an advisor, a diploma and a report light in turn along the walkway, and the hall’s windows light at the end1234
  1. 1EnrollmentAdmitted, deposited, and still enrolled on census day
  2. 2Student successWeek three: missed classes and a hold, advisor called
  3. 3CompletionTransfer credit applied; no course taken twice
  4. 4ReportingRetention counted once, for the board and the state
Illustrative: one student, from the application to the report the state receives.

Where students fall through, and where the numbers do

Four index cards we could write at almost every college we meet, and most districts. Each one reads the same way: what students notice, what staff notice, the reason, and what it costs students and the institution.

  • Advisors find out after the midterm

    Students notice
    An email about a missed deadline weeks after it passed.
    Staff notice
    Advisors with hundreds of students each and no list of whom to call first.
    The reason
    Attendance, learning-platform activity, grades, holds and aid status sit in different systems and are joined, if at all, after midterms.
    What it costs students and the institution
    For students, a failed term, aid put at risk, or debt with no degree. For the institution, first-year students lost, and the tuition and state funding that would have followed them.
    Continue in DATA 230: early alerts on the degree map

  • Every office reports a different enrollment

    Students notice
    Little, until a program they rely on is cut on the wrong numbers.
    Staff notice
    Admissions, the registrar, institutional research and finance each with their own count, and a week of reconciling before every board meeting.
    The reason
    Headcount, full-time equivalents and census-day enrollment defined differently in each office, from different extracts.
    What it costs students and the institution
    For students, a program cut or a section cancelled on the wrong count. For the institution, budget decisions made on numbers nobody can reconcile, and state and federal reports that run late.
    Continue in DATA 201: counting one class three ways

  • Transfer students repeat what they already passed

    Students notice
    Courses they already passed, listed again on their degree plan.
    Staff notice
    Transcripts evaluated by hand, weeks after the student registered.
    The reason
    Course equivalencies kept in spreadsheets and people’s heads instead of in the degree audit.
    What it costs students and the institution
    For students, an extra term of tuition and a later start at work. For the institution, longer time to degree, lower completion, and students who leave for a college that took their credits.
    Continue in DATA 230: transfer credit on the degree map

  • AI reached the classroom before the policy did

    Students notice
    Different AI rules in every course, and no way to know which tools see their work.
    Staff notice
    Faculty using AI tools on student work, and vendors adding AI to the learning platform.
    The reason
    No inventory, no review path, and no answer yet on FERPA, accessibility or academic integrity.
    What it costs students and the institution
    For students, an accusation resting on a detector alone, and their work in tools nobody approved. For the institution, appeals over AI accusations, and a policy written after the incident.
    Continue in DATA 340: the AI rubric

Three results that transfer

No education engagement is published yet. These three come from other sectors, each doing a job colleges and districts need done: a reporting layer over a core system, AI literacy across a workforce, and an assistant for repeat questions at peak. Each is set out at full length with the client’s own figures; the clients are described, not named.

Transfer credit 1

From health and social care

Accepted towardInstitutional reporting over the student system

A multi-site care provider and nonprofit

A governed reporting layer over the core people system

Workforce data lived mainly in one core system, with no capacity for analysis and too little integration or access. The platform brought that data into one repository automatically, modeled it for reporting, secured it down to the row and column, and put a semantic layer and dashboards over the measures leaders relied on.

Coursework

  • Automated ingestion: from the core system into a central repository.
  • Modeling for reporting: each measure defined once for analysis and dashboards.
  • Row- and column-level security: sensitive records seen only by those entitled to them.
  • A semantic layer and dashboards: the priority measures, self-served.

Outcomes reported

  • 30%more efficient talent data analysis, with more timely decisions
  • 45%reduction in reporting errors
  • 70%of staff reported easier use and access

Reported by the client.

Case study: The people-data platform case

Transfer credit 2

From healthcare

Accepted towardAI literacy for faculty and staff

A medium-sized regional health organization

An AI and data literacy program for more than 300 employees

The organization saw data-driven decisions and AI as central to patient care and operations, but its workforce lacked the skills to use them. The program began with an assessment of skills and needs role by role, from doctors and nurses to administrative staff, then a curriculum built on the organization’s own challenges, taught in hands-on sessions and backed by online resources, with progress assessed throughout.

Coursework

  • Needs assessment: skill gaps and training needs identified for each role.
  • A tailored curriculum: data analysis, visualization and AI concepts, set in the organization’s own work.
  • Hands-on sessions: instructor-led training for more than 300 employees.
  • Progress tracking: regular assessments and feedback against the learning objectives.

Outcomes reported

  • 80%of participating employees showed improved data handling and analysis skills
  • 30%increase in the use of data visualization tools for reporting
  • 15%improvement in the organization’s data quality

Reported by the client.

Case study: The workforce AI and data literacy case

Transfer credit 3

From government

Accepted towardAdmissions, aid and registration questions at the start of term

The 311 service of a major US city

An assistant for repeat questions at peak demand

As guidance changed week to week, 311 agents spent their days on the same questions, and some answers depended on where the caller lived. The assistant worked inside the existing call flow, answered from a knowledge base staff could change without code, sent each caller to the right local resource by ZIP code, and passed anything that needed judgment to an agent with the context.

Coursework

  • Inside the live channel: part of the 311 call flow, not a separate bot.
  • Answers staff can update: content changed without code as guidance changed.
  • Routing by location: the right local resource for each caller.
  • A hand-off with context: anything that needed judgment went to an agent.

Outcomes reported

One trusted, current source of answers at the moment demand peaked.

No figures were reported for this engagement; it is cited for what it shows.

Case study: The city 311 assistant case

Count it once: enrollment, retention and completion

Start with one entering class, and count its completion three ways. Then turn over a flashcard to see how each number can be counted, who it is reported to, the decision it moves, and what to agree before the next census.

One entering class, three measures of completion

A college enrolls 1,000 first-time, full-time students one fall, and 1,000 more who start part-time or transfer in. Six years later, each square below is ten students. Choose what is counted.

Count the rate

52%520 graduated here, of 1,000 first-time, full-time students

The IPEDS graduation rate: first-time, full-time students, finishing here within 150% of normal time.

65%520 here and 130 elsewhere, of the same 1,000

A completion measure built from national enrollment matching. It counts what the students achieved, and it is not an IPEDS rate.

51.5%1,030 graduated here, of all 2,000 who entered

Your own measure for the whole class. IPEDS Outcome Measures follow part-time and transfer-in students cohort by cohort; one combined rate is yours to define.

Illustrative figures, checked line by line. Each is correct for the students it counts, and each answers a different question; only the first is the federal graduation rate.

First-time, full-time: 1,000 students

Part-time and transfer-in: 1,000 students

Where the 2,000 students are after six years
OutcomeFirst-time, full-timePart-time and transfer-in
Graduated here within six years520510
Graduated at another college13060
Still enrolled here6090
Transferred out, no degree yet900
Left without a degree200340

Seven numbers on flashcards

Enrollment

Can mean

  1. Headcount. Every student enrolled on census day.
  2. Full-time equivalent. Credit hours divided by a full-time load, such as 15 a term.
  3. Which period. Census day, the end of term, or everyone enrolled at any point in the year.

Reported to

The state funding formula, the budget office, accreditors and IPEDS.

Lives in

the student information system, the census snapshot, the registrar’s extracts

The decision it moves

The budget, the state funding claim, and how many sections and staff to plan for.

Show your work

On census day 10,000 students are enrolled: 6,000 full-time at 15 credits and 4,000 part-time at 6 on average, 114,000 credits in all. That is 10,000 by headcount and 7,600 full-time equivalents, while 13,500 different students enroll at some point in the year.

Agree before the next census

Headcount or full-time equivalents, and on which day?

Retention rate

Can mean

  1. Fall to fall. First-time, full-time students who return the next fall.
  2. Fall to spring. The same students enrolled in the spring term: persistence.
  3. Anywhere. Students enrolled at any college the next fall, not only this one.

Reported to

The board, enrollment and student success leaders, and IPEDS.

Lives in

the student information system, census snapshots, enrollment matches from other colleges

The decision it moves

Where advising and first-year support go, and which students get a call before the spring term.

Show your work

Of 2,000 first-time, full-time students, 1,820 enroll in the spring and 1,560 return the next fall. Persistence is 91% and retention 78%; counting the 120 enrolled at other colleges, 84% are still enrolled somewhere.

Agree before the next census

Fall to fall or fall to spring, and here or anywhere?

Time to degree

Can mean

  1. Calendar time. Years from first enrollment to the degree.
  2. Terms enrolled. Only the terms the student actually attended.
  3. Credits. Credits earned by graduation against the credits the degree requires.

Reported to

Students and families, the state, and degree-map and transfer policy.

Lives in

the student information system, the degree audit, transfer credit records

The decision it moves

Which degree maps, transfer rules and course schedules to change first.

Show your work

A student starts in August 2020, stops out for a year and graduates in May 2025: four years and nine months on the calendar, eight terms enrolled, and 132 credits earned for a 120-credit degree.

Agree before the next census

Calendar time, terms enrolled or credits, and are transfer credits in?

DFW rate

Can mean

  1. D, F and W. Every grade of D or F, and every withdrawal.
  2. Without withdrawals. Only grades of D and F.
  3. An FW rate instead. Where a D earns credit, only F grades and withdrawals, reported as an FW rate.

Reported to

Deans and department chairs, and where tutoring and course redesign go.

Lives in

the student information system’s grade history, the learning platform

The decision it moves

Where tutoring, placement and course redesign go, section by section.

Show your work

Of 400 students in a gateway math course, 60 earn a D, 50 an F and 30 withdraw. The DFW rate is 35%, and 27.5% without withdrawals; where a D earns credit, the FW rate is 20%.

Agree before the next census

Which grades count, and by section or by course?

Yield and summer melt

Can mean

  1. Yield. Admitted students who enroll, or who pay a deposit.
  2. Melt. Students who deposit but are not enrolled on census day.
  3. Which day. Measured at the deposit deadline, the first day of class or census day.

Reported to

The enrollment plan, the budget, and the size of the next admit pool.

Lives in

the admissions system, deposit records, the student information system

The decision it moves

How many students to admit, and which deposited students get a call over the summer.

Show your work

Of 8,000 admitted students, 2,400 pay a deposit and 2,100 are enrolled on census day. Yield is 30% by deposits and 26% by census day; summer melt is 12.5%, 300 of the 2,400.

Agree before the next census

Yield by deposit or by census, and when is melt measured?

Tuition discount rate

Can mean

  1. All institutional aid. Every grant the institution funds, against gross tuition and fees.
  2. Unfunded aid only. Endowed and restricted scholarships left out.
  3. Which students. First-time students only, or every undergraduate.

Reported to

The board, the budget and the enrollment strategy.

Lives in

the financial aid system, the general ledger, the student billing system

The decision it moves

The aid strategy, and the net tuition the budget can count on.

Show your work

Gross tuition is 60.0 million and institutional aid 27.0 million: a discount rate of 45% and net tuition of 33.0 million. With 6.0 million of endowed scholarships left out, the rate is 35%.

Agree before the next census

All institutional aid or unfunded only, and for which students?

Cost per credit hour

Can mean

  1. Direct cost. Instructional salaries and benefits only.
  2. With overhead. Department overhead, facilities and support shared in.
  3. Which credits. Credits taught, or credits completed.

Reported to

Program review, budget models, and the new federal earnings and cost reporting.

Lives in

the general ledger, the HR and payroll system, course enrollment records

The decision it moves

Which programs to grow, restructure or review, alongside what their graduates earn.

Show your work

A department spends 4.8 million on instruction and teaches 24,000 credit hours: 200 dollars per credit hour. With 1.2 million of overhead shared in, it is 250 dollars.

Agree before the next census

Direct or with overhead, and credits taught or completed?

Which number does your board ask about every term, and get a different answer from each office? Bring it to a 30-minute call, and we will show you what it takes to count it once and report it the same way to everyone.

Count one number with us

The work, laid out as a degree map

Four terms, following students from the first deposit to the report the state receives, then an elective track for school districts. Each course says what it is assessed by; where a published result exists, it is linked.

Term 1Recruit and enroll

Up to census day

  • DATA 231

    Yield and melt seen student by student

    Deposits, financial aid status, housing and orientation sign-ups joined, so counselors call the admitted students most likely to melt before census day.

    Assessed by: Summer melt, and census-day enrollment against deposits

    First step: AI & Data Assessment Sprint

  • DATA 232

    Net tuition modeled before aid is packaged

    Admit pools, aid offers and past yield modeled together, so the aid budget brings in the class the institution wants at the net tuition the budget assumes.

    Assessed by: Net tuition revenue per new student, and the discount rate by group

    First step: AI & Data Assessment Sprint

  • DATA 233

    Admissions and aid questions answered at peak

    An assistant on the admissions and aid pages answering from approved policies, checking who the student is before anything about their own record, and handing the case to staff.

    Assessed by: Calls and emails per applicant at peak, and time to answer

    Case study: The city 311 assistant case

Term 2The first year

From week one to the next fall

  • DATA 234

    Early alerts advisors act on

    Attendance, learning-platform activity, grades and holds joined every week into a short list for each advisor, with the reason each student is on it.

    Assessed by: Fall-to-spring persistence, and advisor contacts in the first six weeks

    First step: AI & Data Assessment Sprint

  • DATA 235

    Gateway courses with the highest DFW rates

    Which sections, prerequisites and student groups drive the rate, so tutoring, placement and course redesign go where they change it.

    Assessed by: DFW rate in gateway courses

    First step: Data Platform Accelerator

Term 3Through to completion

Until the degree is awarded

  • DATA 236

    Transfer credit applied before registration

    Incoming transcripts read and mapped to the degree audit, with the registrar approving each equivalency, so students register for what they still need.

    Assessed by: Credits at graduation, and time to degree

    First step: AI & Agent Pilot

  • DATA 237

    Course schedules built from degree-map demand

    Sections planned from what students on each degree map still need and from past fill rates, so a required course never stands between a student and graduation.

    Assessed by: Students blocked from a required course

    First step: AI & Data Assessment Sprint

Term 4Report and fund

Every term, and every October

  • DATA 238

    Census, IPEDS and state reports from one model

    Enrollment, retention and completion defined once, with lineage back to the student system, so IPEDS, the state and the board receive the same numbers.

    Assessed by: Days to produce a report, and numbers that differ by audience

    First step: Data Platform Accelerator

  • DATA 239

    Program costs and outcomes, program by program

    Completions, enrollment, cost per credit hour and contribution margin by program, ready for the federal earnings reporting and for your own program review.

    Assessed by: Programs with cost and outcome data at review

    First step: Data Platform Accelerator

  • DATA 240

    Programs matched to the jobs your region is hiring for

    Program enrollment and completions set against regional job postings, licensure pass rates and graduates’ earnings where the state shares them, so a community or technical college can show which credentials pay off.

    Assessed by: Completers working in their field, and programs that pass the federal earnings test

    First step: Data Platform Accelerator

Elective track

For school districts: the same work, in a district’s terms

Districts count different things, answer to different rules, and plan around different calendars. Six courses we would start with, each measured by a number the district already reports.

Show your work: chronic absence

Most states count a student as chronically absent after missing 10% of the days enrolled; a few count 15 days or more. A student enrolled all 180 days who misses 18 is chronically absent either way. A student who enrolls in January, is enrolled 90 days and misses 10 is chronically absent at 10% (11.1%), and not under a 15-day count. Whether excused absences and suspensions count differs by state as well.

  • DATA 251

    Chronic absence seen in the first month

    Daily attendance joined with course marks and family contacts every week, so each school sees which students are on track to miss 10% of the year while a call home can still change it.

    Assessed by: The chronic absence rate, and students reached in the first 30 days

    First step: AI & Data Assessment Sprint

  • DATA 252

    Screening results tied to the intervention that followed

    Fall, winter and spring literacy and math screening joined with the intervention each student received, so the district can see which programs move students and which only fill the time.

    Assessed by: Students moving up a screening band by spring, by intervention

    First step: Data Platform Accelerator

  • DATA 253

    Special education timelines tracked student by student

    Referral, consent, evaluation and IEP meeting dates in one view, with access limited to the people who serve each student, so case managers see who is close to a deadline.

    Assessed by: Evaluations and IEP meetings held within the state’s timelines

    First step: AI & Data Assessment Sprint

  • DATA 254

    Staffing and budgets set against enrollment by school

    Enrollment projected by school and grade and set against staffing, class sizes and spending, ahead of budget, boundary and closure decisions.

    Assessed by: Forecast error at the fall count, and class sizes against target

    First step: AI & Data Assessment Sprint

  • DATA 255

    State accountability reports from one model

    Attendance, assessment, graduation and staffing data defined once, with lineage back to the student system, so the state, the board and the school report card show the same numbers.

    Assessed by: Days to submit, and corrections after submission

    First step: Data Platform Accelerator

  • DATA 256

    Family messages in the language each family reads

    Notices and messages drafted in the languages families speak, with a person checking anything about a student’s rights, discipline or special education before it goes out.

    Assessed by: Families reached, and translation requests waiting

    First step: AI & Agent Pilot

In every office

  • One student record, with FERPA built in

    Student information, learning platform, aid, advising and housing records matched, with access by role and every disclosure logged.

    First step: Data Platform Accelerator. Taught alongside Data Engineering & Platforms.

  • An AI policy, an inventory and a review path

    Every AI tool used by staff, faculty and students listed with an owner, and a review before any tool touches student records or decides anything about a student.

    First step: AI Governance Quickstart. Taught alongside Strategy & Governance.

  • Advisors with the time and trust to act

    Lists sized to caseloads advisors can work, the reason shown with every flag, and a way to say a flag was wrong, so the alerts are still used in the second term. A model nobody has time to act on changes nothing.

    First step: Adoption Review. Taught alongside AI & Data Literacy.

  • AI literacy for faculty and staff

    Role-based training on your own data and tools, with academic integrity and student privacy built in, so the policy is understood where it is used.

    First step: AI & Data Literacy. Taught alongside Adoption Review.

The students you worry about most sit somewhere else on this map? Bring one population to a 30-minute call.

Talk through one student population

The required reading on student data and AI

Eight entries, each set out as a citation with who it applies to, where it stands this term, and what it means for your data. Each links to its source. It is a guide to what to check, not a ruling on what applies to you.

  1. Reading 1

    FERPA, the Family Educational Rights and Privacy Act (opens in a new tab)

    Applies toSchoolsColleges

    Where it stands this term
    In force. Vendors and AI tools usually receive education records under the school official exception. It holds only when the vendor does work you would otherwise do yourself, has a legitimate educational interest, is under your direct control in how it uses and keeps the records, and uses them only for that purpose. Consent and the other exceptions are separate routes.
    For your data, it means
    Contracts that set out how each AI tool meets that exception, access limited by role, limits on reuse such as training a vendor’s model, and a record of disclosures where FERPA requires one.

  2. Reading 2

    State student privacy laws: California SOPIPA, New York Education Law 2-d, Illinois SOPPA (opens in a new tab)

    Applies toK-12 schoolsTheir vendors

    Where it stands this term
    • California SOPIPA: operators of K-12 sites and apps, since Jan 1, 2016.
    • New York Education Law 2-d: districts, BOCES and their contractors; its regulations (Part 121) since January 2020.
    • Illinois SOPPA: districts and operators; the amended law since Jul 1, 2021.
    For your data, it means
    No targeted advertising, profiling or sale of student data by vendors; signed data privacy agreements, published in some states; and notice of breaches to families.

  3. Reading 3

    COPPA, as amended in 2025 (opens in a new tab)

    Applies toApps used by children under 13Schools that consent for them

    Where it stands this term
    Amended rule published Apr 22, 2025; compliance required since Apr 22, 2026. The duties fall on the operator of the service. Under FTC guidance rather than the rule’s text, a school may consent in place of parents only for a tool used solely for school purposes.
    For your data, it means
    Before a school authorizes a tool for children under 13: the operator’s consent process, separate consent before children’s data goes to third parties, including for advertising or training AI, a written retention policy, and a security program.

  4. Reading 4

    ADA Title II rule on web and mobile accessibility (opens in a new tab)

    Applies toPublic districtsPublic colleges

    Where it stands this term
    Equal access is already required; the interim final rule (Apr 20, 2026) set the dates for the technical standard. Apr 26, 2027 where the government a college or district belongs to serves 50,000 people or more; Apr 26, 2028 for smaller ones and special districts. Size is the jurisdiction’s, not enrollment: a state university counts its state, a county college its county, an independent district its Census estimate. Private colleges answer to Title III and Section 504 instead.
    For your data, it means
    Websites, apps, course content in the learning platform, documents, videos and dashboards meet WCAG 2.1 Level AA, and any AI assistant meets it too.

  5. Reading 5

    The GLBA Safeguards Rule, as Federal Student Aid enforces it (opens in a new tab)

    Applies toColleges in federal aid

    Where it stands this term
    Colleges in federal student aid count as financial institutions under the FTC’s Safeguards Rule. The Department of Education checks it through the Program Participation Agreement and annual compliance audits, as announced in February 2023. Breaches of 500 or more people go to the FTC since May 13, 2024.
    For your data, it means
    A qualified individual, a written risk assessment, multi-factor sign-in and encryption for student financial information, oversight of vendors, and an incident response plan.

  6. Reading 6

    Federal earnings accountability and STATS reporting (opens in a new tab)

    Applies toColleges in federal aid

    Where it stands this term
    Final rule published Jul 1, 2026; the earnings measure applies from Jul 1, 2027. The Department’s update of Oct 8, 2026: data for the 2024 to 2026 reporting cycles due by Jan 15, 2027, with no further extensions, and colleges may report through STATS early. Results are to be published in 2027; loss of loan eligibility is possible from 2028. It replaces the gainful employment and financial value transparency reporting.
    For your data, it means
    Program-level enrollment, completer and cost data that you can stand behind, checked against the Department’s lists before it is submitted, and warnings to students in programs that fail.

  7. Reading 7

    IPEDS, and its admissions supplement (ACTS) (opens in a new tab)

    Applies toColleges in federal aidACTS: four-year colleges

    Where it stands this term
    IPEDS is required of every college in federal aid. ACTS was collected from Dec 18, 2025 to Mar 18, 2026; the Department revised its explanation on Sep 14, 2026 after a court remand, and it continues for 2026-27.
    For your data, it means
    Admissions, aid, grades and completions by race and sex for several years back, which only works if the records were kept consistently; small groups need disclosure protection, and IPEDS definitions kept apart from your own measures.

  8. Reading 8

    State requirements for AI policies in schools: Ohio and Tennessee (opens in a new tab)

    Applies toK-12 schoolsTennessee public colleges

    Where it stands this term
    • Ohio: every public district, community school and STEM school, by Jul 1, 2026, with a state model policy since January 2026.
    • Tennessee: districts and public charter schools since the 2024-25 school year; public colleges and universities since 2025.
    For your data, it means
    A written policy on acceptable use, privacy, vendor review and academic integrity, and an inventory of the AI tools it governs.

For orientation only, not legal advice: other federal, state and local rules may apply, and what applies depends on whether you are public or private, K-12 or higher education, and on the data and the program. Checked October 14, 2026.

Score an AI use case before students meet it

Pick a use case to see what it does, how it scores on the rubric, what to read first, the conditions for using it, and who would usually sit on the review.

Student success

Early alerts for students at risk

What it does: Flags students showing signs of trouble from attendance, learning-platform activity, grades and holds.

The rubric
Uses student education recordsYes
Affects admission, a grade or aidAt times
Faces students or families directlyNo
Used by children under 13At times
Could treat groups of students differentlyYes
Accessibility review neededAlways

Read first

  • FERPA: only school officials with a legitimate educational interest see the list.
  • Civil rights law: flags and outreach checked across student groups.
  • State student privacy law, if a vendor runs the model.

Conditions for use

  • Advisors decide what to do; the tool only flags.
  • Every flag shows the signals behind it.
  • Accuracy and flag rates compared across groups each term.

Review panel (suggested)

Chaired by: Dean or directorStudent success leadRegistrarCampus privacy officeInstitutional research

First step: AI & Data Assessment Sprint

Admissions

AI in reading and scoring applications

What it does: Summarizes applications, or scores them for the readers.

The rubric
Uses student education recordsYes
Affects admission, a grade or aidYes
Faces students or families directlyNo
Used by children under 13No
Could treat groups of students differentlyYes
Accessibility review neededAlways

Read first

  • Civil rights law, and the 2023 Supreme Court decision on race in admissions.
  • IPEDS admissions reporting (ACTS) on the outcomes by race and sex.

Conditions for use

  • A reader decides every application.
  • No score built on protected traits or close stand-ins for them.
  • Outcomes reviewed by group after every cycle.

Review panel (suggested)

Chaired by: Cabinet or boardEnrollment leadGeneral counselCampus privacy officeInstitutional research

First step: AI Governance Quickstart

Student services

An assistant for admissions, aid and registration questions

What it does: Answers questions about deadlines, aid and registration, and points students to the right office.

The rubric
Uses student education recordsAt times
Affects admission, a grade or aidNo
Faces students or families directlyYes
Used by children under 13No
Could treat groups of students differentlyAt times
Accessibility review neededAlways

Read first

  • FERPA: the student is identified before anything about their own record is shared.
  • ADA Title II: the assistant meets WCAG 2.1 Level AA like the rest of the website.

Conditions for use

  • Answers only from approved, dated content.
  • Tells students it is an assistant, and offers a person.
  • Anything about a student’s own case goes to staff.

Review panel (suggested)

Chaired by: Dean or directorEnrollment servicesRegistrarDisability servicesIT security

First step: AI & Agent Pilot. Case study: The city 311 assistant case

Teaching

AI tutors and course assistants

What it does: Answers students’ questions about course material and gives feedback on drafts.

The rubric
Uses student education recordsAt times
Affects admission, a grade or aidNo
Faces students or families directlyYes
Used by children under 13At times
Could treat groups of students differentlyAt times
Accessibility review neededAlways

Read first

  • FERPA, if it keeps students’ work or questions.
  • COPPA, for users under 13.
  • ADA Title II, for course content delivered through it.

Conditions for use

  • The instructor sets what it covers.
  • Students are told when they are using it.
  • No student work used to train a model.

Review panel (suggested)

Chaired by: Dean or directorAcademic affairsFaculty senateDisability servicesIT security

First step: AI & Agent Pilot

Teaching

AI-assisted grading and feedback

What it does: Suggests scores and comments on assignments against a rubric.

The rubric
Uses student education recordsYes
Affects admission, a grade or aidYes
Faces students or families directlyAt times
Used by children under 13At times
Could treat groups of students differentlyYes
Accessibility review neededAlways

Read first

  • FERPA: grades and feedback are education records.
  • Your grading policy and grade appeal process.

Conditions for use

  • The instructor assigns every grade.
  • Students are told, and can ask for a review.
  • Suggested scores checked against the instructor’s on a sample each term.

Review panel (suggested)

Chaired by: Cabinet or boardProvost’s officeFaculty senateRegistrarGeneral counsel

First step: AI Governance Quickstart

Academic integrity

AI-writing detection

What it does: Estimates whether submitted work was written with AI.

The rubric
Uses student education recordsYes
Affects admission, a grade or aidYes
Faces students or families directlyNo
Used by children under 13At times
Could treat groups of students differentlyYes
Accessibility review neededAlways

Read first

  • Your conduct code and its due process.
  • Civil rights law: detectors misjudge some groups of writers more often than others.

Conditions for use

  • Never the only evidence in a case.
  • A person reviews every flag, and the student can respond.
  • False-positive rates checked before the tool is used.

Review panel (suggested)

Chaired by: Dean or directorConduct officeFaculty senateStudent affairsGeneral counsel

First step: AI Governance Quickstart

Advising

Degree planning and transfer credit evaluation

What it does: Reads incoming transcripts and maps courses to the degree audit.

The rubric
Uses student education recordsYes
Affects admission, a grade or aidAt times
Faces students or families directlyNo
Used by children under 13No
Could treat groups of students differentlyAt times
Accessibility review neededAlways

Read first

  • FERPA for the transcripts and the audit.
  • Your state’s transfer and articulation agreements.

Conditions for use

  • The registrar approves every equivalency.
  • Students see how each course was mapped.
  • A clear route to appeal a decision.

Review panel (suggested)

Chaired by: Dean or directorRegistrarAcademic departmentsCampus privacy office

First step: AI & Agent Pilot

Scheduling

Enrollment forecasting and course scheduling

What it does: Forecasts enrollment and section demand, and proposes the schedule.

The rubric
Uses student education recordsAt times
Affects admission, a grade or aidNo
Faces students or families directlyNo
Used by children under 13No
Could treat groups of students differentlyAt times
Accessibility review neededAlways

Read first

  • FERPA for any student-level records the forecast uses.

Conditions for use

  • Chairs approve the schedule.
  • Forecasts checked against census each term.
  • Only aggregated data leaves the student system.

Review panel (suggested)

Chaired by: System ownerRegistrarDepartment chairsInstitutional research

First step: AI & Data Assessment Sprint

Schools

Classroom AI tools for students under 13

What it does: Learning apps with AI features used by younger students in class.

The rubric
Uses student education recordsYes
Affects admission, a grade or aidNo
Faces students or families directlyYes
Used by children under 13Yes
Could treat groups of students differentlyAt times
Accessibility review neededAlways

Read first

  • COPPA, as amended, in force since Apr 22, 2026.
  • State student privacy laws and their data privacy agreements.
  • Your district AI policy, required in Ohio and Tennessee.

Conditions for use

  • Only tools on an approved list, each with a signed data privacy agreement.
  • No advertising and no model training on student data.
  • Families told which tools are used.

Review panel (suggested)

Chaired by: Cabinet or boardSuperintendent’s cabinetDistrict data privacy officerCurriculum and instructionSpecial education

First step: AI Governance Quickstart

Institutional

Drafting reports and accreditation self-studies

What it does: Drafts narrative sections from approved data and evidence.

The rubric
Uses student education recordsAt times
Affects admission, a grade or aidNo
Faces students or families directlyNo
Used by children under 13No
Could treat groups of students differentlyNo
Accessibility review neededAlways

Read first

  • Your accreditor’s standards for evidence.
  • FERPA, for any student records in the exhibits.

Conditions for use

  • Every figure drawn from the governed reporting model.
  • The committee reviews and signs the narrative.
  • Sources cited for every claim.

Review panel (suggested)

Chaired by: System ownerInstitutional researchAccreditation liaisonSelf-study committee

First step: AI & Agent Pilot

A first score for your review committee, not a legal opinion or a pre-approval. What applies depends on whether you are public or private, K-12 or higher education, the data and how the tool is used; your counsel, privacy officer and registrar decide. Accessibility is a fixed row because every tool people use needs that review, under the ADA Title II rule at public institutions and Section 504 wherever there is federal funding. “At times” on records turns on what the tool actually receives: whether it identifies students, and who sees it. The full reading list: DATA 310.

Who chairs, and who takes part. Several decisions may be separate: admissions, privacy, counsel, disability services, faculty governance, IT security and academic leadership each have a part. The level says who chairs the review; each use case lists the offices that usually sit on it.

  • Cabinet or board: the president’s or superintendent’s cabinet, counsel and the privacy officer.
  • Dean or director: the academic or student affairs lead, with IT security and the registrar.
  • System owner: a named owner, access controls and a check each term.

A tool faculty already use, or a vendor feature switched on over the summer? Thirty minutes is usually enough to see how it scores and what it needs.

Score one AI tool with us

Register for one of three sections

Three first steps for three different readers, listed the way a schedule lists sections. The first two are our assessment and governance offers, sized to one student population or one policy, and neither commits you to the next.

  1. Section 01 Recommended

    Find where one group of students gets stuck, and what to fix first

    Suited to
    A provost, vice president for student success or institutional research director with one population to help first.
    Length
    Two weeks, one student population

    One population’s records traced across the student system, the learning platform, aid and advising to where students stall; the gaps on one list with what each costs in retention; and a plan in order, ready for the budget cycle.

    Start with one student population

  2. Section 02

    Review the AI already in classrooms and offices, before an incident does

    Suited to
    A CIO, provost or superintendent’s cabinet with AI already in classrooms and offices.
    Length
    A policy, an inventory, one review

    One AI use case taken through FERPA, accessibility, academic integrity and your state’s rules to a decision, with the inventory and review path the rest of your AI can follow.

    Ask about the AI policy review

  3. Section 03

    Data Maturity Assessment

    Suited to
    An IT or institutional research lead who wants a quick read first.
    Length
    Free, roughly six minutes
    Offered as
    A self-serve assessment

    Fifteen questions on lineage, quality, ownership, access and platform, scored on the spot and with no email asked for.

    Take the assessment

After either of the first two: a first build for one population or one office, measured against a number you already report, such as fall-to-spring persistence, a DFW rate or the days it takes to produce a report, then a decision to scale, change or stop.

AI & Agent Pilot · Data Platform Accelerator

Faculty and staff first? Role-based AI and data literacy runs alongside either section, or on its own: AI & Data Literacy.

Office hours: what provosts, registrars and CIOs ask

The questions we hear most from academic, student affairs and IT leaders, answered.

Can you work with our student information system as it is?

Yes. We read from the systems you run, model the data once, and leave the system of record where it is. Replacing a student system is a decision for another day and another budget.

Is our student data safe to use with AI?

It can be, though FERPA’s school official exception is only one test. The others are the purpose, who can see what, the vendor’s terms, how long prompts and outputs are kept, the security controls, whether anything trains a model, who handles an incident, and which state law and COPPA apply. We work through each with your privacy officer and registrar, and build inside your environment where your policy requires it.

Do you work with community colleges and school districts, as well as universities?

Yes. Most of this page is written for colleges; districts have their own track in DATA 230, with chronic absence, early screening, special education timelines and state reporting. The rules differ too: districts add COPPA for younger students, state student privacy agreements with every vendor, and in some states a required AI policy.

What does the accessibility rule mean for our dashboards and course content?

Equal access is already required. The technical standard, WCAG 2.1 Level AA, applies from April 26, 2027 where the government a public college or district belongs to serves 50,000 people or more, and from April 26, 2028 for smaller ones; private colleges answer to other rules. It covers websites, apps, documents, videos and the content in the learning platform. We rebuild dashboards to that standard, check them with a keyboard and a screen reader from start to finish, add text alternatives for charts, and offer the data behind each one as a download.

How do you keep early alerts fair?

Leaving out race, sex and other protected traits is not enough, because other signals can stand in for them. Before launch the model is tested on your own students, group by group, for the students it misses as well as the ones it flags wrongly. Advisors decide what happens and see why each student was flagged, students can ask about a flag, and every term we check whether outreach and support reach each group equally.

Can you help with the new federal earnings test?

With the data you have to report: program-level enrollment and completers each October, checked against the Department’s lists, and the cost and outcome data your own program review needs. The earnings themselves come from federal tax data, not from you.

Do you train faculty and staff on AI?

Yes. Role-based AI and data literacy, built on your own tools and policies, with academic integrity and student privacy part of every session rather than a separate module.

Can a small IR or IT team keep it running?

That is planned from the first week: documentation, training for the people who will own it, and tools built on what your team already supports.

Bring the number your board asks about every term.

Fall-to-spring persistence, the DFW rate in one gateway course, a census report that takes three weeks, or the AI policy faculty keep asking for. Thirty minutes with the lead who would run the work, on the first thing worth fixing.

Book 30 minutes for your campus

Advisors finding out too late? Thirty minutes on student success, reporting or your AI policy.

Book a call for your campus