How UW-Madison and industry are working together to drive Wisconsin's AI future
AI has moved from theory to tooling. It's writing code, flagging anomalies in factories, assisting clinical teams, and quietly rewriting how work gets done. The question inside most Wisconsin companies isn't "if," it's "how." UW-Madison is helping answer that with research depth, cross-campus reach, and hands-on partnerships that turn ideas into working systems.
The university's expertise cuts across medicine, engineering, business, agriculture, and education. Its AI Hub for Business convenes leaders from Google, Walmart, Reddit, Medtronic, Kimberly-Clark, Bain, and Accenture-23 executives who bring real constraints and high-value use cases to the table. The School of Computer, Data & Information Sciences is reorganizing into the College of Computing & Artificial Intelligence, signaling sustained commitment to the field and workforce needs.
Learning by doing: where students and industry meet
Justin Hines - Director of Corporate Relations, CDIS: Companies are still figuring out where AI drives value in their operations. Capstone teams give them a low-risk, high-impact way to explore. Students pressure-test ideas, surface technical hurdles early, and de-risk adoption before production. The result: real business problems mapped to concrete solutions, while students build the skills employers expect.
On the ground, teams are refining tools for major health systems, streamlining aerospace supply chains, and building AI literacy inside Wisconsin firms. It's practical, fast, and aligned with actual constraints-data access, evaluation metrics, governance, and change management.
What leaders are piloting now
Matt Seitz - Executive Director, AI Hub for Business: Most leaders are off the sidelines. Early projects focus on customer chatbots and process automation for measurable efficiency. The more advanced work uses AI to build software and open strategic growth paths. Early movers are already seeing an edge.
- Customer support and knowledge assistants grounded in internal data (RAG over docs, policies, and tickets)
- Process automation across finance, HR, procurement, and compliance
- AI-assisted software development and testing to compress delivery cycles
The skill that matters: judgment
Tools generate confident outputs; leaders need people who can judge when those outputs can be trusted. Seitz's point is blunt: using AI is easy, verifying it is the job. Students and early-career staff must learn to question sources, validate numbers, and escalate uncertainty before it hits an executive's desk. That means clear evaluation criteria, audit trails, and human oversight where risk is high.
From problem statements to deployable systems
Kyle Cranmer - Director, UW-Madison Data Science Institute: Most real problems need translation before AI can help. The Data Science Institute convenes domain experts and AI researchers to convert fuzzy goals into datasets, model classes, and evaluation plans. That spans large language models, agent-based orchestration, and the data governance needed to make them dependable.
Trust is front and center: mitigating hallucinations, protecting privacy, and handling sensitive information responsibly. Frameworks like the NIST AI Risk Management Framework help teams align risk, controls, and measurement from day one. Understanding how large language models behave-strengths and failure modes-keeps deployments honest.
Faster discovery for Wisconsin industry
AI is speeding up the front end of discovery. In drug development, models help narrow candidates before lab work begins. In materials science, AI is identifying structures that may capture CO2 more efficiently. Experiments still validate the results, but the search space shrinks and timelines tighten.
For researchers building these capabilities, curated learning helps. See AI for Science & Research for methods and tools that bridge lab work and applied AI.
Why partner instead of build alone
John Garnetti - Managing Director, Office of Business Engagement: Universities and industry excel at different, complementary things. UW-Madison brings depth in AI and data science and a rare breadth across medicine, education, engineering, agriculture, and business. That mirrors how private-sector problems actually look-cross-functional and constraint-heavy. Collaboration with government and community organizations extends the impact and keeps solutions grounded in public needs.
Work in five years: human + AI by default
Expect AI to be embedded in most roles the way spreadsheets or web search are today. Tasks that took days-market scans, competitive analysis, financial models-drop to hours. Plant managers will predict equipment failures, sales teams will walk in with fresh account intelligence, and supply chains will reroute in real time. The scarce skill becomes judgment, creativity, and relationship-building-augmented by strong AI fluency.
Practical next steps for science and research leaders
- Pick two workflows with clear metrics and sufficient data. Stand up narrow pilots using retrieval-augmented generation and track precision, recall, and time saved.
- Establish an evaluation stack: red-teaming for safety, ground-truth datasets, human-in-the-loop review, and incident logging tied to the NIST AI RMF.
- Treat data like a product. Define owners, access controls, quality checks, versioning, and retention. Good data beats a fancier model.
- Upskill teams on prompting, verification, and basic model behavior. Make "don't trust-verify" a norm in analysis and reporting.
- Co-develop with external partners. Use capstones and applied research to test feasibility, integration paths, and change management before scaling.
The bottom line
Wisconsin's advantage won't come from flashy demos. It will come from disciplined collaboration, sharper judgment, and steady integration of AI where it measurably improves outcomes. UW-Madison is building that muscle with industry-problem by problem, lab to line, classroom to production.
Your membership also unlocks:
