AI for Higher Education
AI that shows its work
Argos Data Agent is built on Argos. It answers plain-language questions from the DataBlocks your campus already governs, with the SQL shown on every response. Staff get answers without a report request, and IT never loosens a permission to make it happen.
Built on the governed reporting layer institutions have trusted for two decades
The situation
AI is already touching your data. The question is whether it is governed.
Policies are lagging adoption. EDUCAUSE research found practitioners nearly evenly split on whether they are even aware of guidance covering their own work-related AI use, and reported that institutions have more AI tools running across their campuses than IT departments know about.
Knowledge management for safer AI ranks fifth on the EDUCAUSE Top 10 IT Issues list, framed as mitigating AI risk by integrating knowledge management into data governance, privacy, and ethics programs. Governance comes first, then the tool.
01
There is already more AI on campus than IT knows about
Departments adopted tools individually. Nobody mapped which ones touch institutional data, and the policy that would have covered it is still in draft.
02
An AI that hides its work cannot be checked
The dangerous query is not the one that fails. It is the one that runs perfectly against raw tables and returns a number that is wrong, with nothing underneath it for anyone to inspect.
03
Your board asked what you are doing about AI
Governance, data quality, and access control are what stand in the way, and none of them fit in a board meeting. The pressure to show something is real all the same.
04
A model pointed at your ERP has never seen your definitions
It does not know which of eleven date fields you use for census, or what your institution means by full-time. It will pick one and sound certain.
Governed objects
It queries what your team defined, not your raw database
Argos Data Agent is built on Argos and requires it. It does not run against your ERP schema; it runs against the governed reporting objects your team already built, where the joins, the filters, and the institutional definitions were settled by people who know the data.
A model pointed at a raw database schema has to infer what your tables mean, and it will infer wrong without telling you. A model querying objects your team already defined is working from settled meaning, and every answer shows the DataBlock and the SQL it used, so a wrong one can be caught by the person who owns the definition.
- Runs against governed objects, where the definitions are already agreed
- Reaches only the objects you enable for it, so an entire subject area can be placed out of scope
- Answers reflect your institution's meaning of a term, not the model's guess
- Backward compatible with reporting objects you have already built and already trust
SELECT COUNT(*) AS studentsFROM DataBlock: EnrollmentWHERE first_gen = 'Y'Argos X
Every answer comes with the reasoning attached
Verification is what makes an AI answer usable in a place where numbers get quoted to boards, agencies, and legislators.
Argos Data Agent shows the SQL it ran and the DataBlock it ran against, beside every response. An analyst can confirm the cohort definition matches the certified one before the number travels. For people who would rather work with the result directly, Pivot Mode and Explore Data reshape it without writing a query.
- Argos Data Agent for plain-language questions, with the SQL visible
- Pivot Mode and Explore Data for reshaping and exporting without a new request
- Answers grounded in governed objects your team controls, not an uploaded spreadsheet
- Dashboard Designer, Explore Data, and Pivot Mode are included with Argos at no additional cost
- Argos Data Agent is licensed separately, in Core and Advanced tiers
Governance
The controls your policy committee is going to ask about
Institutions adopting AI on student data face the same list of questions every time, and the answers decide whether anything gets approved.
Argos Data Agent operates under Zero Data Retention, so institutional data is not retained by Evisions or by the model provider, for training or for anything else. You decide which governed objects it can reach, which means a whole subject area can be excluded from AI workflows rather than argued about field by field. Sensitive fields can be masked or redacted on top of that.
- Zero Data Retention, so institutional data is not retained by Evisions or by the model provider
- You choose which governed objects the agent can reach, so finance or financial aid data can be left out entirely
- Masking and redaction set on the DataBlock and its columns
- Runs against governed objects instead of the raw database, which limits exposure by design
- One HECVAT and one Trust Center covering the platform
Adoption
Rolling it out does not start another governance project
The governance work is already done if your reporting objects are governed. There is no new data store to secure, no copy to keep in sync, and no second access review, because the agent reads what your team already vetted. The value shows up on the follow-up question, when a dean wants the same number cut differently and would previously have filed a second request.
Rolling this out well is a people problem more than a technical one. The EDUCAUSE Top 10 IT Issues list puts the human edge of AI second and notes that training is time and resource intensive for already overworked technology teams, and that peer-based approaches spread capability faster than centralized ones.
- Nothing new for users to learn beyond asking a question in plain language
- No new data store to secure and no copy of institutional data to keep in sync
- Report writers keep control of what the objects mean
- Start with one office and one set of objects, and widen the agent's reach as governance catches up
The AI answer you can check
Every answer shows the SQL it ran and the DataBlock it ran against. See it in a demo.
Case study
1 day
to set up on existing DataBlocks
240
trained users with Data Agent access
~15 seconds
to break down grades across 2,000+ graduates
Texas A&M put governed student data within reach of 240 users
“When it gives me the SQL statement, I go look at the where clause and see how it got there. It’s nice that it explains how it got to that point, not just here’s the question, here’s the answer.”
Josh Radnitzer · IT Manager IV, Enterprise Information Systems, Texas A&M
Related
Where the governed layer comes from
Reporting for Higher Education
The layer Data Agent answers from, and how it gets built.
Platform and Security
Authentication, permissions, deployment, and what your reviewers will ask.
Reporting for IT
Governance, access control, and getting out of the report queue.
BI and Dashboards
When the answer should be a standing view instead of a question.
How it fits
Where this sits relative to what you already run
With your ERP
Argos Data Agent never touches your ERP schema directly. It queries governed objects that sit above it, which is what lets it answer in your institution's terms instead of the database's.
With Ellucian Insights
Insights aggregates data across Ellucian products on a managed data layer. Argos works directly against your own environment, covering dashboards through Dashboard Designer, conversational analysis through Argos Data Agent, compliance output, and scheduled delivery, on definitions your team controls. Most institutions run both.
With a general-purpose AI assistant
A general assistant is excellent at drafting and summarizing and has no knowledge of your data model. Point one at your student system and it has to guess what your tables mean. The two are complementary, and they should not be asked to do each other's jobs.
Trust and security
You take security seriously. So do we.
The Trust Center is published and the HECVAT is available on request, so a reviewer can start before anyone schedules a call.
Zero Data Retention
Institutional data is not retained by Evisions or by the model provider, for training or for anything else.
Permissions are enforced, never overridden
A question cannot reach data a report would not return for that user. There is no second access model.
Verifiable answers
The SQL it ran and the DataBlock it ran against are shown with every response, so an answer can be audited instead of taken on trust.
Field-level control
Masking and redaction set on the DataBlock and its columns, so a question about a population does not expose an individual.
Authentication
LDAP group sync, SAML SSO, and permission tiers, configured once at the platform layer.
Documentation
Trust Center, and a HECVAT (the standard higher education security questionnaire) on request.
Frequently asked questions
Questions we get asked first
Does it train on our data?
No. Argos Data Agent operates under Zero Data Retention.
Does it respect our existing permissions?
Yes. It honors the MAPS permissions already in place, so a user cannot reach data through a question that they could not reach through a report.
What stops it inventing a number?
It queries governed objects your team defined instead of your raw ERP schema, and it shows the SQL behind every answer so the result can be checked before anyone acts on it.
How accurate is it?
We publish the reasoning instead of a score. Every answer arrives with the SQL beside it, so your own team can confirm the figure instead of trusting a percentage we chose. Bring the questions your office asks and we will run them against your data.
Can we mask sensitive fields?
Yes. You can block an entire DataBlock from the agent, mask specific columns, or redact specific values by exact match or pattern. The settings apply to everyone who uses the agent.
What is included with Argos, and what is licensed separately?
Dashboard Designer, Explore Data, and Pivot Mode are included with Argos at no additional cost. Argos Data Agent is licensed separately, in Core and Advanced tiers.
What is the difference between Core and Advanced?
Reach and depth. Core is a campus-wide subscription with unlimited users, answering from one governed DataBlock, opened from Explore Report Data on any existing report. Advanced is seat licensed for analysts, IR, and report owners, reasons across multiple DataBlocks in one conversation, opens from any folder or dashboard in Argos X, and adds relationship analysis. Both tiers build charts, and governance is identical in both.
Do we need Argos first?
Yes. Data Agent answers from governed Argos objects, so that layer has to exist first. If you do not have it yet, start at the reporting page rather than here.
Can we turn it off, or limit who can use it?
It depends on the tier. Data Agent Core is licensed per institution, so once it is turned on, anyone with access to your Argos environment can use it. Data Agent Advanced is licensed per seat, so you decide who gets access by assigning seats. For both, what you control is what the agent can reach: you choose which DataBlocks are available to it. If you do not want it answering questions about finance data, you leave the finance DataBlocks out and it cannot answer them for anyone.
Which departments get the most from it?
Offices where the follow-up question is expensive. Institutional research, financial aid, enrollment, and finance are the usual first movers.
How does this coexist with Ellucian Insights?
Insights aggregates across Ellucian products on a managed data layer. Argos runs against your own environment on your team's definitions. Most institutions run both.
Our staff are already using a general AI assistant. Is that a problem?
It is the reason most institutions call. Pasting institutional data into a general assistant puts it outside your permission model and outside your retention policy, and the answer comes back with nothing underneath it to check. The practical fix is not a ban. It is giving people a governed place to ask, which is what this is.
Leadership wants an AI initiative. Where does this fit?
It is usually the first thing an institution can point at that is both real and defensible. It answers from objects your team already vetted, it shows the SQL, and it inherits the permissions you already set, so the governance conversation is largely already settled before it starts.
How long does implementation take?
If Argos is already in place, this is a license and a configuration instead of a project. Professional Services scopes it with you. The people doing that scoping ran campus IT and institutional reporting before they worked here.
What does it cost, and can we talk to an institution using it?
Licensing is per institution rather than per user, and price is driven by institution size and product mix rather than headcount. Argos, Argos X, FormFusion, and IntelleCheck all carry unlimited users and unlimited connections, with no per-connection or per-report charge and nothing billed for people who only read reports. The one exception is Argos Data Agent Advanced, which is licensed per seat. Data Agent Core is unlimited. Tell us your FTE and which products you are weighing and we will give you a real figure on the first call, then put you in touch with an institution of roughly your size and ERP. Ask them about support while you are there. It is in-house, and it is the thing customers raise unprompted more than anything else.
Next step
AI on the data your campus already governs
Plain-language answers, visible queries, and no new data layer to build or secure. Get a demo.
Ellucian Partner Award
2025: Customer First
Ethos Connected
since 2020
Ellucian partner
22 years