THE UNGLAMOROUS TRUTH: WHY CLEAN ERP DATA MAKES OR BREAKS AI IN HIGHER EDUCATION

ERP · Data Governance · AI

Every vendor pitch about AI in higher education eventually arrives at the same slide: a dashboard full of confident forecasts, flagged anomalies, and plain-language answers to questions that used to require a report builder and a week of waiting. What almost none of those pitches spend enough time on is the boring prerequisite underneath all of it. An AI model sitting on top of Banner, Workday, or PeopleSoft is only as good as the data it is reading, and at a lot of institutions, that data has quietly accumulated years of duplicate vendor records, inconsistent chart-of-accounts coding, and workarounds nobody ever documented. AI does not fix that. It amplifies it.

The Gap Between Feeling Ready and Being Ready

There is a specific and well-documented gap between how ready organizations believe their data is for AI and how ready it actually is. Industry surveys have found that roughly 80% of organizations believe their data is AI-ready, yet more than half still report ongoing quality problems once they actually try to put a model into production (Chimpiri, 2025). Higher education is not exempt from that gap, and arguably has more reason to worry about it. A university’s ERP touches admissions, financial aid, HR, procurement, and the general ledger, which means a data problem in one office rarely stays contained to that office. A duplicate vendor record in procurement can throw off a spend-forecasting model. A stale advising note can cause an early-alert system to miss a student who actually needs help.

What Bad ERP Data Actually Looks Like

It rarely looks dramatic. A recent industry roundup identified seven concrete, unglamorous signs that an organization’s ERP data is not ready to support AI: untrustworthy inventory or asset records, routine manual reconciliation between the ERP and a spreadsheet somewhere, staff regularly exporting ERP data to Excel because they do not trust the system’s own reporting, and duplicate customer, vendor, or in a university’s case, constituent records (ERP Software Blog, 2026). None of these are exotic problems. They are the kind of thing that accumulates slowly, department by department, as staff turn over and nobody owns the job of keeping the underlying data clean. The article’s central point translates directly to campuses: AI does not correct inconsistent or duplicate data, it produces confident-looking output built on top of it, which is often more dangerous than an obviously broken report because nobody thinks to question it.

Where This Shows Up on Campus

Enrollment forecasting is a good example. A model that blends historical registration data with softer signals like deposit rates or application volume is only useful if the underlying student records are consistent across terms, and if a transfer student’s history was recorded the same way in 2019 as it is in 2026. Financial forecasting has the same problem from a different angle. If two departments code travel reimbursements differently, or if a chart-of-accounts restructuring five years ago left a batch of legacy codes still floating around unreconciled, a variance-detection model will either miss real anomalies or flag noise as a crisis, and either outcome erodes trust in the tool fast. Student success analytics may be the most sensitive case of all, because a duplicate or fragmented student record can mean an early-alert system genuinely does not see a student who is struggling, not because the model failed, but because the data it was reading was incomplete.

Governance Is the Actual Project

The institutions that get real value out of AI-enhanced ERP modules are, almost without exception, the ones that treated data governance as its own project rather than an afterthought to the AI rollout. That means assigning actual ownership of master data, not just IT ownership but functional ownership, where the registrar’s office is accountable for the integrity of student records and procurement is accountable for vendor records. It means documenting what institutional customizations to Banner or PeopleSoft actually do, since a decade of local modifications is common at most colleges and rarely written down anywhere a new AI integration project can find it. And it means building in a review step before a forecasting or anomaly-detection feature goes live, where someone with real domain knowledge sanity-checks a sample of its output against what they know to be true, rather than trusting the dashboard because it looks polished.

Three Questions Worth Asking Before the Next Module Goes Live

A few questions tend to separate institutions that get durable value from AI-enhanced ERP from institutions that end up with an expensive dashboard nobody fully trusts. First, where does the underlying data actually come from, and has anyone traced it back through every system it passed through before it landed in the field the model is reading? Second, who is accountable when the model is visibly wrong, and how quickly does that error get corrected versus quietly ignored because nobody wants to be the one who says the new tool is broken? Third, and most practically, has the institution done the unglamorous cleanup work first, the deduplication, the chart-of-accounts standardization, the constituent-record reconciliation, or is it hoping the AI layer will somehow smooth over data that has been messy for a decade? Institutions that answer that third question honestly tend to be the ones with AI initiatives that actually stick.

Beidat LLC works with colleges and universities on data readiness assessments, governance frameworks, and ERP and AI implementation support, from the unglamorous cleanup work through vendor selection and go-live. If your institution is trying to figure out whether its data is actually ready for the AI features already sitting inside Banner, Workday, or PeopleSoft, reach out at support@beidat.com or 888.384.1992.

References

Chimpiri, T. R. (2025, November 17). Beyond the back office: How AI is reimagining ERP in higher education. EdTech Digest. https://www.edtechdigest.com/2025/11/17/beyond-the-back-office-how-ai-is-reimagining-erp-in-higher-education/

ERP Software Blog. (2026, June 24). Is your ERP ready for AI? 7 signs your data is holding your company back. https://erpsoftwareblog.com/2026/06/is-your-erp-ready-for-ai-7-signs-your-data-is-holding-your-company-back/

Let’s Data Science. (2026, June 24). ERP data is holding back AI readiness. https://letsdatascience.com/news/erp-data-is-holding-back-ai-readiness-7bd07627

Last updated on August 22, 2026