An AI Certificate That Starts by Asking If the Problem Needs AI

Waterloo built an AI certificate where the first skill is deciding whether the problem needs AI at all.

That is not how these announcements usually read. The standard version names a platform, a partner and a number of seats, and the implied promise is that students will come out fluent in a tool. This one names a partner and a number of seats too. Then the person running it says something that quietly undercuts most of what higher education has built in the last three years.

If you are finalizing a syllabus this week, that sentence is worth more to you than the certificate is.

What the AI certificate actually announces

On August 6, 2026, the University of Waterloo announced a partnership with Cohere, the Canadian AI company, to develop an AI Transformation and Change Management certificate through Waterloo’s Future of Work Institute. According to Waterloo’s own announcement, the program carries no course credit and is built from badges earned through hands on learning, workplace experience and collaborative problem solving. It is scheduled to launch in fall 2026. The first cohort starts in January 2027 with 20 structured co-op positions, working with private, public sector and non profit organizations, open to students in any faculty.

The modules are listed as human centred design, ethics, change management and business strategy, with students exposed to technologies from a range of companies rather than one. Development was led by the Future of Work Institute with the Centre for Work Integrated Learning and the GreenHouse Social Impact Incubator.

Read that list again and notice what is missing. There is no prompt engineering module. There is no platform certification. For a program built with an AI company, remarkably little of it is about operating an AI.

The sentence that does the work

Dr. Edith Law is Executive Director of the Future of Work Institute and an Associate Professor of Computer Science. In the announcement she says: “AI adoption is not the point, but solving problems is.”

She then describes what the certificate trains students to do, which is to think critically about bottlenecks, processes and challenges inside organizations, and to design solutions with those organizations. Her word for the output is socio technical, and she is explicit that the solutions will likely involve both technical and non technical innovations.

In plain terms: the students are being trained to diagnose first. The tool selection happens afterward, if it happens at all. A student who finishes this program and correctly tells an organization that its bottleneck is an approval chain rather than a missing model has done the job right.

Now hold that standard up against the AI initiative on your own campus.

Most campus AI work is answering a question nobody asked

The common shape of AI adoption in higher education goes like this. A tool becomes available. A committee forms. A policy gets written. A workshop gets scheduled. Faculty are asked to integrate AI into their courses, and the integration is measured by whether AI appears in the course.

At no point in that sequence does anyone name the specific thing that is broken.

This is why so much of it lands with a thud. An instructor adds an AI assignment to a course whose actual problem is that the third week has no scaffolding and half the class falls behind there every term. The AI assignment is not wrong. It is just answering a question that was never asked, while the real bottleneck sits three weeks in, unnamed, for the fourth year running.

The diagnosis is the hard part, and it is the part that gets skipped

There is a reason the diagnostic step gets skipped, and it is not laziness. Naming a bottleneck out loud is uncomfortable. It means saying that a unit you built does not work, or that an assessment you have given for six years is measuring the wrong thing, or that the reason grading takes all weekend is not volume but a rubric that forces you to write the same paragraph thirty times.

Adopting a tool is much easier. It is visible, it is fundable, and it can be reported upward as progress. Diagnosis produces a sentence that somebody has to own.

Waterloo’s design is interesting precisely because it makes the uncomfortable part the credential. Students are not certified in a platform, which would be obsolete in eighteen months anyway. They are certified in the discipline of walking into an organization and finding the actual constraint. Law calls the arrangement a living lab, where the students do the transformation work while the university keeps revising the method based on what happens.

That is a curriculum design decision, not a technology decision. Which is what makes it portable.

What this changes about the course you are building right now

You do not need a partnership with an AI company to run the same move. You need one course and one honest sentence.

Take a course you are teaching this term. Not the one you wish you were teaching. Find the single place where the most time goes or the most students stall. Then write one sentence naming it, and a second sentence saying what kind of problem it is. There are only three answers that matter. It is a design problem, meaning the sequence or the assessment is wrong. It is a capacity problem, meaning the work is correct and there is simply too much of it. Or it is an information problem, meaning you cannot see what is happening early enough to act.

Only the second and third have anything to do with AI. The first one is fixed with a pen.

This is the same order of operations Harvest Kernel calls review first, and it is why the Course Creation Pathway opens with a stage that produces an argument rather than a draft. You establish what the course is for and where it currently fails before anything generates a word of content. A tool that skips that step will happily build you a beautiful version of the wrong course.

Where this goes next

Diagnosing one course is a twenty minute job you can do with a pen. Diagnosing four preps, then turning each diagnosis into a dean ready case for what the course should become, is the reason it stops at one. Course Spark, the opening stage of the Course Creation Pathway inside the Harvest Kernel Faculty Toolkit, takes the problem you just named and turns it into a summary you could hand a chair, grounded in your Course Record so it already knows your course, your outcomes and your voice. You review and export. It is the stage that does this exact job.

See Course Spark in the Faculty Toolkit

If you would rather compare bottlenecks with other instructors doing this in the same week, that conversation is happening in the free Harvest Kernel community. If you are trying to run this across a department rather than a course, book a free call.

Twenty students will spend January learning to walk into an organization and find the constraint before anyone mentions a model. Your syllabus is due Friday. You already know where your course breaks. The only open question is whether you write it down.

Sources

• Partnering to lead AI transformation across organizations, Waterloo News, University of Waterloo, August 6, 2026. uwaterloo.ca
• University of Waterloo and Cohere develop AI transformation certificate with 20 co-op positions, EdTech Innovation Hub, August 13, 2026. edtechinnovationhub.com
• Cohere leadership team, listing Joelle Pineau as Chief AI Officer. cohere.com/about