How Can University Faculty in Calgary Use AI for Assessment Design?

Last updated: August 2026

University faculty in Calgary get the most out of AI not by having it grade student work, but by using it to design better assignments and rubrics. With a properly configured AI project, an instructor can run a draft assignment through an alignment check, a student-perspective audit, and a rubric calibration in about 30 to 60 minutes, work that used to take the better part of a day. The catch is that the value comes from the setup, not the prompt. Faculty who paste a one-line request into a chatbot get generic output; faculty who give the AI their teaching philosophy, course outline, and grading standards get a genuine thinking partner. This post walks through the exact workflow we teach University of Calgary faculty, the tools that work, and how Alberta employers can fund the training.

Start With the Workflow, Not the Tool

The single biggest mistake faculty make with AI is treating it as a better search box. The prompting era is largely over; context is now the lever. Asking an AI to "generate a rubric for my entrepreneurship class" produces something bland and unusable. Loading a project with your teaching philosophy, your course learning objectives, your Bloom's taxonomy expectations, and an example of your preferred rubric format produces something you can actually work with.

The reason this matters is the same reason most corporate AI fails. MIT's 2025 report, The GenAI Divide: State of AI in Business, found that 95% of enterprise AI pilots delivered no measurable impact, and the differentiator was almost never model quality. It was whether the tool was integrated into a real workflow with real context. Teaching is no different. A useful analogy we use in faculty sessions: a colleague who has taught alongside you for years already knows who you are, what you value, and where your students struggle. The AI knows none of that unless you tell it. The work is telling it well, once, in a place it can reuse.

Takeaway: Before you prompt anything, build the context. The 20 minutes you spend loading a project pays back on every assignment for the rest of the term.

The Assignment Diagnostic: Four Steps

At Solway we teach a repeatable workflow we call the Assignment Diagnostic. It takes a draft assignment and runs it through four passes before you finalize a single instruction.

Step one, alignment check. The AI maps your draft assignment against your course learning objectives and flags Bloom's taxonomy mismatches, the places where your instructions say "apply" but actually only test "understand," or where the task sets students up to describe when you wanted them to take and defend a position.

Step two, student-perspective audit. You ask the AI to read the assignment as a student who has never seen the material, and to surface the top five ambiguities that would send someone to office hours. In a faculty workshop we ran for a Calgary business school, this step surfaced the exact questions the instructor confirmed came up repeatedly in office hours every term: one verdict or three, what counts as an "academically supported" opinion, whether a table is allowed. The AI found them in ninety seconds.

Step three, decision summary. The AI summarizes every alignment question that requires your judgment and hands them back to you. It does not invent answers or fill gaps. You make the calls; it refines the assignment around your decisions.

Step four, export. Only once the content is right do you ask it to produce the final document. A useful token-saving tip: have it print the output in the chat first so you can review, then generate the Word file, rather than burning through a full document generation before you have even read it.

Takeaway: The Assignment Diagnostic catches misalignment and ambiguity before students ever see the assignment, which is precisely where most office-hours confusion originates.

Rubric Build and Calibrate

Once the assignment is locked, the same project moves to the rubric. The technique that makes this reliable is calibration by example. You ask the AI to write short excerpts of an A-plus paper, a B paper, and a C paper as lightweight markdown files, then you read them and tell it whether the grades feel right.

The diagnostic signal here is subtle and useful: if the B paper is hard to write, the rubric probably does not differentiate well between levels. In one session, the AI kept wanting to make its own B-grade sample better than the rubric's own threshold allowed, which exposed a genuine gap in how the criteria were separated. That is the rubric telling you it is broken before a single student is marked against it. Once the criteria are calibrated, the final table copies straight into the D2L rubric builder used across University of Calgary courses.

Takeaway: A rubric you have not tested is a rubric you do not yet trust. Calibrating against AI-written sample papers validates the separation between grade levels in minutes.

The One Rule That Never Changes: AI Does Not Grade

The hard line in every faculty workflow we build is that AI never grades, scores, or evaluates actual student work. University of Calgary, like the University of Alberta and most Canadian institutions, requires that generative AI use adhere to existing academic integrity and privacy policies, and student work carries privacy obligations that make uploading it inappropriate. A well-built project encodes this as a guardrail: load your institution's AI policy into the project instructions, and the AI will flag you if you try to feed it student submissions.

This is not only a compliance point. It reflects the deeper principle we build every engagement around, which we call the Goldilocks Zone: AI handles the horsepower steps, humans keep the judgment. Offloading evaluation to a machine is exactly the kind of judgment work that erodes competency over time. Design, diagnose, and calibrate with AI; grade with your own trained eye.

Takeaway: Use AI to build the instrument, never to read the students. This keeps you compliant and keeps your judgment sharp.

Projects and Skills: Make It Repeatable

The faculty who get real leverage do not rebuild this from scratch each time. Two features make the workflow repeatable. Projects scope the AI's working memory to a specific course, holding your teaching philosophy, course outline, AI policy, and rubric format in one place, with one chat per assignment to keep context clean. Skills save a multi-step workflow, so the Assignment Diagnostic itself becomes a one-click routine: open a new chat, invoke the skill, upload the next assignment, and go.

A practical granularity tip from our sessions: keep separate projects for distinct programs, because an undergraduate course and an MBA course sit at different Bloom's levels and the drift confuses the AI if you mix them. Within a single course, though, slides, assignments, and rubrics can happily coexist.

Takeaway: Set the workflow up once as a project with saved skills, and every subsequent assignment takes a fraction of the time.

The Bigger Shift: Assessment Itself Is Changing

There is an honest tension worth naming. Any assignment plus rubric, fed to a capable model, will produce an A-grade paper. The faculty we work with handle this less by trying to detect AI and more by designing around it: tying assessments to in-class discussions and guest speakers the model cannot know about, requiring application of specific sessions, and using formats like video walkthroughs where students talk through their reasoning.

This is a live conversation in business education. Accreditation bodies including AACSB are already flagging the need to rethink how learning objectives are assessed, and there is a credible view that traditional written-assignment-plus-letter-grade models have a limited shelf life. The consensus among the faculty we work with is not to eliminate AI use but to stay slightly ahead of it. That is a design problem, and it is one AI can help you solve.

Takeaway: The goal is not to AI-proof a single assignment; it is to redesign assessment for a world where students have these tools too.

How Alberta Employers Can Fund This Training

For Alberta organizations, including post-secondary institutions and departments training their staff, the Canada-Alberta Productivity Grant (CAPG) reimburses 50% of eligible training costs, up to $5,000 per employee per fiscal year and up to $100,000 per employer per year. AI and digital-skills training falls squarely within the program's digital and technological skills category. Training must be delivered by a third-party provider, which is where Solway qualifies, and you apply through the CAPG Portal before training begins.

Worth noting: CAPG has no minimum-hour requirement, so a half-day faculty workshop can qualify. (The 21-hour minimum belongs to the separate Canada-Alberta Job Grant, a different program.) Always confirm current details at alberta.ca/CAPG before applying, as guidelines are updated periodically.

Solway runs tool-agnostic, hands-on faculty workshops built around exactly this workflow. Shaheer Tariq, Co-Founder of Solway, has delivered AI briefings and workshops to organizations ranging from Global Affairs Canada to Calgary energy companies, and our faculty sessions focus on the practical mechanics of assignment design, rubric calibration, and repeatable teaching workflows rather than abstract theory.

Takeaway: A half-day faculty workshop can be largely reimbursed through CAPG, making practical AI training close to cost-neutral for eligible Alberta employers.

Frequently Asked Questions

Can AI grade my students' assignments?

No, and it should not. University of Calgary and most Canadian institutions require AI use to comply with academic integrity and student privacy policies, which makes uploading student work inappropriate. Use AI to design assignments and calibrate rubrics; keep the actual grading in human hands. This also protects your own judgment from atrophy over time.

What AI tool should faculty use for teaching workflows?

The workflow is tool-agnostic, but the features that matter are projects (to hold your course context) and the ability to save repeatable skills. In our faculty sessions, Claude tends to be more direct and critical when given strong project instructions, which is useful for diagnostic work, while ChatGPT is a capable alternative. What matters more than the brand is loading your teaching context properly.

How long does it take to build an assignment and rubric with AI?

Once your project is set up, an assignment diagnostic and rubric calibration typically takes 30 to 60 minutes, compared with several hours done manually. The initial project setup adds about 20 minutes but pays back on every subsequent assignment in the course.

Will students just feed the rubric to AI and get an A?

They can, which is why design matters more than detection. Faculty we work with tie assessments to in-class discussions, guest speakers, and specific sessions the model cannot access, and increasingly use formats like video walkthroughs. Students who skip the in-class elements tend to score lower even with AI assistance.

Is there funding for AI training for Alberta institutions?

Yes. The Canada-Alberta Productivity Grant reimburses 50% of eligible third-party training costs, up to $5,000 per employee per year and $100,000 per employer annually, under its digital and technological skills category. Confirm current eligibility at alberta.ca/CAPG.

How is this different from the University's own AI resources?

University of Calgary's Taylor Institute and library provide excellent policy and academic-integrity guidance. What they generally do not provide is a hands-on, step-by-step production workflow for building assignments and rubrics inside an AI project. Solway's faculty workshops fill that practical gap and complement the institution's own resources.

Can Solway customize a workshop for our department?

Yes. Solway designs faculty sessions around your programs, your Bloom's expectations, and your existing D2L setup, and can tailor examples to specific disciplines. Contact Solway to discuss a session and how CAPG can offset the cost.