What Data Do Universities Need to Migrate?
Most university system projects involve four types of data, each with its own challenges:
- Curriculum data – programme specifications, module descriptors, learning outcomes and assessment, often held in unstructured Word documents.
- Student records – data in Ellucian Banner, Tribal SITS or SAP Student Lifecycle Management.
- Finance, HR and ERP data – including payroll, procurement and reporting.
- Business intelligence and reporting – data warehouses and the feeds into them.
How We Migrate University Data
- Discovery – agree the target data model, mapping rules and responsibilities, and identify every source of data.
- AI extraction – extract data from unstructured sources such as Word documents, field by field.
- Transformation and validation – apply your mapping rules with ETL tooling, using student records data for lookups and checks.
- Iterate with your team – each run is reviewed by your subject experts, and the rules are refined before the next run.
- Load and reconcile – produce load-ready files, reconcile every record and support validation in the new system.
AI Where It Adds Value, Proven Tools Where It Matters
AI is exceptionally good at reading unstructured documents, but on its own it can be unpredictable: results can change between runs, and it may fill gaps rather than flag them. Our approach uses AI to write field-by-field extraction scripts, so every run produces the same result. A second AI layer validates the output, and transformation runs in established ETL tools. Our data architects review the results and refine the rules at each stage.
The outcome is AI-level speed with the accuracy, repeatability and audit trail a university migration needs.
Case Study: University of Lancashire
The University of Lancashire was implementing a new TimeEdit curriculum management system alongside a move to Banner SaaS. Its curriculum data was held in more than 10,000 unstructured Word programme specifications and module descriptors, built on templates that had changed many times over a decade.
After a proof of concept, we delivered three iterations between October 2025 and February 2026. AI generated field-by-field extraction scripts, an AI validator checked the results, and ETL tooling transformed the data against the university's schema, using a snapshot of Banner data for lookups.
Results:
- Over 99% extraction accuracy
- More than 1,000 course records and 9,500 module records created without manual data entry
- A repeatable process – the full extraction can be re-run in seconds as rules are refined
Webinar: Migrating 10,000+ Curriculum Documents with AI
In this UCISA webinar, the University of Lancashire and Prolifics explain how the migration was approached, what worked, what didn't, and what they would advise other universities to do.
Key moments
- 3:12 – The university's curriculum data challenge
- 19:07 – Proof of concept and objectives
- 24:06 – How AI and ETL were combined, iteration by iteration
- 35:05 – Results and data validation
- 42:12 – Lessons learned for curriculum system projects
Lessons for Curriculum Management System Projects
Lessons shared by the University of Lancashire for universities planning a similar migration:
- Allow ample time for the data model and mapping – it took around six months, not the two originally planned.
- Complete the initial system configuration before mapping where possible, so the target is not moving.
- Appoint a dedicated data migration lead, separate from process design.
- Plan a curriculum freeze, with an exceptions process for essential changes.
- Build in time for quality checks between iterations.
University Data Migration FAQs
Can AI migrate curriculum data from Word documents?
Yes. For the University of Lancashire, AI-generated extraction scripts and automated AI validation achieved over 99% accuracy across more than 10,000 programme specifications and module descriptors.
How long does a curriculum data migration take?
It depends on volume and data quality. Allow time for the data model and mapping first. The University of Lancashire's three extraction iterations ran over around five months, with review time between each.
Do we need to clean our data before migrating?
Not entirely. Profiling and cleansing are part of the migration. But agreeing definitions and ownership early makes every later stage faster.
Can you work with our student records system data?
Yes. We use student records data from Banner, SITS or SAP SLCM for lookups and validation, so migrated records align with the systems they feed.
