From 3 Days to 15 Minutes: Marshall’s AI-Driven Cash Forecasting
Matt Tidd – Senior Vice President of Finance/CFO, Marshall University
Marshall University cut cash-forecasting cycle time from 3 days to 15 minutes by moving from manual, Excel-driven workflows to AI-powered forecasting. In this session, Marshall’s CFO and a practitioner partner unpack the journey: integrating multi-year financial, operational, and external data; modeling tuition cycles, population trends, and cash-spike events; automating what-if scenarios; and using generative AI explanations to build trust with stakeholders.
Attendees will leave with a practical playbook to modernize forecasting for speed, accuracy, and transparency—without ripping and replacing legacy ERP, student, or finance systems.
Marshall University’s treasury team faced timing misses and reconciliation errors in a historically driven, Excel-based cash-forecasting process. Partnering with an AI-enabled forecasting platform, they implemented a modular forecasting engine that ingests finance, operational, and macro-event data, then surfaces predictions with narrative explanations leaders can trust.
The following are the issues addressed and how they were identified:
- Manual prep created delays and error risk (vlookups, timing lags).
- Forecasts under-weighted tuition cycles, inflation, and one-time “cash spike” events.
- Stakeholders lacked transparent rationale behind projections.
Solution approach
- Data integration: Multi-year GL/treasury data + enrollment/operational feeds + external indicators (inflation, population, tuition/fee events).
- Feature & event modeling: Tuition bill dates, enrollment shifts, seasonal utilities, and known cash spikes.
- Scenario automation: One-click what-ifs for tuition changes, enrollment mix, inflation, and payment timing.
- Explainability: Generative AI creates plain-language rationales for forecast drivers to improve adoption.
- Delivery: Outputs to dashboards, reports, cube views, or Excel—meeting stakeholders where they work.
Challenges & lessons learned
- Mapping legacy fields and reconciling historical idiosyncrasies took time—establish a data dictionary early.
- Start with cash drivers you can explain; add complex signals later.
- Governance matters: Freeze feature definitions between cycles to reduce drift.
Change management strategy
- Pilot on one tuition cycle → prove accuracy → expand to full year and other cash drivers.
- Publish confidence intervals and side-by-side comparisons with prior method for 2–3 cycles.
- Train finance staff to shift from compilation to analysis.
Learning Objectives
- Map legacy ERP/student/financial data to a minimal AI feature set for cash forecasting and define quality checks.
- Run automated what-if scenarios (tuition, enrollment, inflation, timing) and interpret model explanations with stakeholders.
- Apply a phased change-management plan to expand from cash forecasting to adjacent use cases (enrollment, fundraising, utilities).
CPE Available
- 1 Credit: Information Technology
|