AI and the Engineering Graduate
Board of Governors
College of Engineering, University of St. Thomas
Jote Taddese
2 October 2026
Everyone in this room has run this play
95%
of enterprise generative AI pilots produced no measurable impact
on the P&L
MIT NANDA, The GenAI Divide: State of AI in Business, 2025.
150 executive interviews, 350 employee surveys, 300 public deployments.
The report attributes this to integration and organizational learning,
not to the quality of the models.
1,000+
AI use cases in a single enterprise portfolio.
The measurable value was concentrated in a handful of them.
From the presenter's own experience.
Ideas were never the scarce resource
No owner
A direction everyone supports and no one is accountable for.
No definition of done
"Adopt AI" cannot be finished, so it is never finished.
No signal
Nothing observable tells you whether a given change worked.
Ask a capable group for ideas and you will get ideas.
Ideas are not what was missing.
MIT's committee on AI reached the same conclusion about its own recommendations: that they are not a checklist of initiatives to be implemented one at a time, but a set of changes that have to be undertaken together.
Report of MIT's Ad Hoc Committee on AI Use in Teaching, Learning,
and Research Training, August 2026.
The number that should concern this board
19%
Employment among workers aged 22 to 25 in highly AI-exposed occupations now sits roughly 19% below where it would be had it kept pace with their less-exposed peers.
Stanford Digital Economy Lab, Canaries in the Coal Mine?
Brynjolfsson, Chandar and Chen, revised August 2026.
Experienced workers show no comparable gap.
It operates through reduced hiring of young workers, not increased separations.
The declines concentrate in occupations that rely on codified knowledge: formal, documented knowledge of the kind taught through education and written procedures.
Read that last line again
The occupations losing entry-level hiring are the ones built on knowledge that a curriculum transmits.
If this holds, it is not just a hiring-market problem.
It is a question about what a degree is for.
The market has already repriced the skill
62%
Average wage premium for roles requiring at least one AI skill, across all sectors studied.
Source: PwC, 2026 Global AI Jobs Barometer.
Under review
ABET's Computing Area Delegation released proposed program criteria for Artificial Intelligence and Machine Learning in October 2025, released for a year of review and comment.
Source: ABET. Proposed, not yet adopted, and computing criteria rather than engineering criteria.
The accreditation floor will arrive eventually.
Waiting for it is a decision to be late.
The same shape, here
Department chairs were asked to develop ideas for how AI could enhance the curriculum.
Traction has been limited.
MIT faced a similar question
In January 2026, MIT charged a committee of faculty, students and staff with three tasks: assess current AI use, identify innovations in teaching and assessment, and propose an AI use policy.
Five months later the committee concluded that the faculty must take on deeper questions about the structure, meaning and value of an MIT education.
Report of MIT's Ad Hoc Committee on AI Use in Teaching, Learning,
and Research Training, 13 August 2026.
Lack of traction is not always inertia. Sometimes it is a correct read that the question was the wrong size.
The capability is already here
Center for Applied AI
MS in Artificial Intelligence
Manufacturing AI micro-credential
TommieBot, built by St. Thomas students and faculty
This is not a college that lacks AI capability.
The gap is diffusion, not capability
Most of what exists sits in graduate programs and a research center.
What has not happened is the harder thing: changing what every undergraduate engineer can do by the time they graduate.
What a decision looks like
Ohio State launched its AI Fluency initiative in fall 2025. Every undergraduate, beginning with the class of 2029, is expected to graduate fluent in their field and in the application of AI to that field. It is embedded in core undergraduate requirements and majors.
Source: The Ohio State University.
Not a model to copy. An illustration of what it looks like when an institution answers the question rather than studies it.
What a governance answer looks like
A named owner
One person accountable, with the authority to decide.
A defined outcome
Stated as what a graduate can do, not as what a course covers.
An observable signal
Something you can look at in twelve months and know.
None of this requires knowing which AI tools matter in 2030. That is the point.
What MIT's committee recommended
An ongoing AI and education committee, responsible for strategy, monitoring and policy.
AI Leads named at the school, college or department level.
A funded implementation team and fellows, plus a pilot fund instructors can draw on.
Metrics, tracked deliberately, including post-graduation feedback.
Report of MIT's Ad Hoc Committee on AI Use in Teaching, Learning,
and Research Training, August 2026.
Owner. Capacity. Money. Signal. The scale is MIT's. The structure is not expensive to copy.
The question only this board can answer
What should a St. Thomas engineer be able to do in 2030 that a graduate of a peer program cannot?
The faculty can design toward that answer.
They cannot supply it.