Skip to content
Securafy AI Lab by Securafy
AI Skills, Research & Trends

Universities Just Made AI Competency a Graduation Requirement. Your Team Still Isn't Covered.

Purdue, St. Bonaventure, and Pakistan's Higher Education Commission now require AI competency before graduation, and they define it as judgment rather than tool use. Jillian O. breaks down how those standards were designed and how a small or mid-sized company can adapt the same structure for the people already on its payroll.

In this article

If the graduates you hire next year arrive with a transcript that certifies AI competency, what standard will you hold your current team to? That question stopped being hypothetical in 2026. Purdue University, St. Bonaventure University, and Pakistan's Higher Education Commission each turned AI competency into a formal academic requirement, with defined outcomes, required coursework, and someone accountable for keeping the standard current.

Most employers have nothing comparable for the people already on payroll. That is the gap worth closing, and the universities have quietly published a usable blueprint for an AI competency standard that a 50-person company can adapt without building a curriculum office.

What does an AI competency requirement actually ask for?

An AI competency requirement asks people to show judgment, not tool fluency. The university versions define competency as using AI effectively in a specific field, understanding its limits, explaining and defending decisions AI helped shape, and adapting as the tools change. Employers can apply the same four tests to their own staff, by role, and verify them rather than assume them.

That definition comes almost word for word from the source. When Purdue's trustees approved the requirement on December 12, 2025, the university described it as a first-of-its-kind "AI working competency" graduation requirement covering every main-campus undergraduate, starting with students who enrolled in fall 2026. The stated goal runs "from exposure and awareness to skill-building and problem-solving mastery," and includes recognizing and communicating clearly about AI use, decisions, and limitations, including developing and defending decisions informed by AI-driven insights.

Notice what is missing from that list: prompt templates, product menus, and feature walkthroughs. Purdue's August 2026 rollout announcement says plainly that the initiative "is not focused on teaching students how to operate AI tools", and that it emphasizes critical thinking, responsible use, and deciding whether AI is appropriate for a task at all.

The common belief this challenges

Most companies treat AI training as a rollout task. A new tool arrives, someone runs a demo, a short policy gets circulated, and the box is checked. The working assumption is that competency follows access: give people the tools, show them the buttons, and skill will accumulate through use.

The universities rejected that assumption, and they had every incentive to take the easy route. Purdue could have required a single generic AI course. Instead it delegated the standard to the provost and the deans of every college, asked each college to set discipline-specific criteria, and asked each to stand up a standing industry advisory board on employers' AI needs so the curriculum gets refreshed every year. The result, by Purdue's count, was 22 courses identified, developed, or revised and more than 300 plans of study updated before the first cohort arrived.

The lesson for a business is uncomfortable but useful. If a major public research university concluded that AI competency has to be defined separately for a nurse, an agronomist, and an engineer, then a single all-staff AI webinar is not a competency program. It is an awareness event, and the gap between those two things shows up later as inconsistent work, unreviewed output, and policies nobody can prove were followed.

Purdue is not alone

St. Bonaventure University announced in March 2026 that every first-year student would complete an AI literacy unit inside its BONA 101 introductory seminar beginning in fall 2026. The unit covers how AI works, responsible use, academic integrity, and questions of bias, privacy, and authorship. Provost David Hilmey framed it simply: AI is becoming part of every discipline and profession, so every student needs that grounding from the start. The university also announced ChatGPT Edu licenses for all students, faculty, and most staff, so access and instruction arrived together rather than access first and instruction someday.

Pakistan applied the idea at national scale. In February 2026 the Higher Education Commission issued a notification making a three-credit-hour AI course mandatory in every undergraduate and postgraduate degree program at public and private institutions, beginning with the 2026 academic session. Universities can deliver it as an elective, an interdisciplinary course, or a supporting subject, but the baseline itself is no longer optional.

InstitutionWhat became requiredDesign choice employers can borrow
Purdue UniversityAI working competency in every undergraduate degree, starting fall 2026Role-specific standards, set by each college and refreshed annually with employer input
St. Bonaventure UniversityAI literacy unit in the first-year seminarA shared baseline for everyone, delivered alongside sanctioned tool access
Pakistan Higher Education CommissionThree-credit-hour AI course in every degree programA non-negotiable minimum, with flexibility in how each program delivers it

How far behind is the existing workforce?

Further than most leaders assume, and largely because nobody designed the learning. A Study.com survey of 1,000 employed workers, reported by eWeek in January 2026, found 89% use AI on the job but only about a third had received employer-provided training. Forty-six percent learned through self-teaching and trial and error, and nearly one in four said their employer offered no AI support at all, whether training, mentorship, or approved tools.

UK research points the same way. A survey commissioned by The Access Group and conducted by YouGov, covered by HR Grapevine in March 2026, found that among employees using AI at work, 70% describe themselves as experimenting with the tools, while only 19% have actively taken AI training or a course. Experimentation is how people discover what AI can do. It is a poor way to learn what it should not be trusted with.

Put those numbers next to the campus requirements and a strange inversion appears. Within a few hiring cycles, an entry-level hire may hold a documented, reviewed AI competency while the manager approving their work learned the same tools alone, unchecked, from whatever surfaced in a search. We have written before that the AI skills gap is not a tool problem. The university mandates make the same argument from the other side: they invested in standards, not software.

Why this is a business problem, not an HR curiosity

The first consequence is hiring leverage. Candidates who have been assessed against a defined AI standard will reasonably ask what standard your company uses. A firm that cannot describe one, let alone show how it verifies it, looks behind the curve to exactly the early-career talent that is easiest to develop.

The second consequence is consistency. When every employee learns AI on their own, every employee also invents their own rules about client data, review steps, and when to trust an output. That variation is where errors and exposure tend to hide, and it is very hard to explain to a client, an insurer, or an auditor who asks how you know your people use AI responsibly. The federal government has already signaled where expectations are heading; our review of what the Department of Labor's AI literacy standard means for your team covers that side of the picture.

The third consequence is leadership credibility. Our guide to the AI skills every SMB leader should cultivate makes the point that leaders set the bar by what they model. If managers cannot explain how they evaluate an AI-assisted deliverable, a written competency standard for their teams will not carry much weight.

How can a 50-person company build its own AI competency standard?

Start small and borrow the university structure rather than its scale. You need a shared baseline for everyone, a role-specific layer for the people whose work carries the most risk, and a named owner who reviews the standard on a schedule. Everything else is detail you can add once those three pieces exist.

In practice, the workable version usually looks like this:

  • Define a baseline everyone must meet. Approved tools, what data never goes into them, how to check an output before it leaves the building, and who to ask when unsure. This mirrors the St. Bonaventure model of one shared unit for all.
  • Write role-specific criteria for the highest-risk roles. A finance lead reviewing AI-drafted reconciliations needs different judgment than a marketer drafting campaign copy. This is the Purdue model of discipline-specific standards.
  • Test for judgment, not attendance. Ask people to evaluate a flawed AI output, explain a decision AI informed, or identify when a task should not use AI at all.
  • Assign an owner and an annual refresh. Purdue built employer advisory boards to keep its criteria current. Inside a company, that is usually an operations or department leader paired with whoever manages your IT and security environment.

A baseline measurement makes all four steps easier. A short AI readiness assessment shows where your team actually stands before you write criteria around guesses. Our walkthrough on AI training for employees versus AI training for leaders is a useful companion when you separate the baseline from the leadership layer, and if you plan to issue credentials internally, it is worth reading what makes an AI certification worth sharing before you design one.

What should the standard avoid?

Avoid treating a vendor's completion badge as proof of competency. A badge confirms someone finished a module about one product. It says little about whether that person can apply AI inside your workflows, under your data rules, with the review discipline your clients expect. Also avoid a one-time sign-off. The universities built annual refresh cycles into their standards because the tools change faster than any curriculum, and a workplace standard written once in 2026 will be out of date well before 2028.

How Securafy helps teams get there

From the skills and research side of our work, the pattern we see most often is not resistance to training. It is the absence of any definition of "good enough." Teams are willing to learn, but nobody has written down what competent AI use looks like for an accounts payable clerk versus a project manager versus an owner. We help leadership teams draft that definition by role, measure the current baseline, and connect the standard to the guardrails that already govern your systems, so training, acceptable use, and access controls describe the same expectations instead of three slightly different ones.

Because Securafy also manages the security and IT environment for many of the companies we train, we can tie competency criteria to real conditions, such as which AI tools are approved in your Microsoft 365 tenant and which data classifications cannot leave it. That keeps the standard practical. If you are still sorting out the bigger picture, our leadership playbook for AI adoption is a good place to frame where a competency standard fits.

Where To Go From Here

The universities have shown that AI competency can be defined, taught by role, and checked on a schedule. Your next step is to write down what that standard means for your own roles and measure your team against it before the next hiring cycle makes the gap visible.

If your team is moving faster with AI than your guardrails are, start with structured training rather than another tool. AI University gives your people role-based AI training with security built into the material, not bolted on afterward.

If you would rather talk through your specific environment first, book a strategy call with Securafy and we will walk your current AI usage, exposure, and the fastest path to safe adoption.

Jillian O.
Jillian O.

Jillian Oco is the Chief Marketing Officer at Securafy, where she leads brand strategy, content, search, AEO, technical SEO, and the way complex technology and risk are communicated to real people.

With more than 10 years in digital marketing, she writes about the overlap between cybersecurity, AI, online trust, reputation, and business growth. Her work is especially focused on making technical subjects easier to understand without flattening them into generic advice or marketing noise.

She is currently learning to live slowly and consciously in a small surfing town with her tiny human. Her self-care must-haves are an Alan Watts mixtape, iced coffee, and a good end-of-week draft beer.

Writes about: Cybersecurity awareness, brand protection, AI risk, online trust, reputation management, AEO, technical SEO, practical security education for SMBs

More from Jillian O.

Learn AI by building with it

AI University helps teams move beyond AI curiosity through practical lessons, secure workflows, guided experiments, and real projects built for everyday business use.

Explore AI University

Stay current on practical business AI

Get practical updates on AI security, governance, tools, compliance, and implementation without the daily hype cycle.

Join the conversation

Have a question or a different take on this? Add it below.