Client course content stays with the client. What follows is how the work was scoped, what constrained it, and what I would change.
- Higher education
- Contract, renewed
- 2026
Rebuilding a ten-course Organizational Development program for Canvas
On paper this was a migration. In practice it was a redesign, because the source material was never built for a learning management system in the first place.
What I received per course was a faculty development document: instructor notes, a reading list, a general concept for where the course should go. No course shell, no module map, no finished assessment plan. The subject matter expertise was there. The instructional architecture was not.
That gap is the entire job. Ten sets of faculty intent, turned into ten consistent, accessible, functioning Canvas courses, without altering the academic substance the faculty own.
Constraints
- Two fixed delivery windows, six courses and then four, against a hard contract end date.
- Ten courses, one designer. Anything solved once had to be reusable across the other nine.
- Faculty input arrived as prose, not structure. Every course needed interpretation before it needed building.
- Inherited content debt: broken links, missing assessments, and accessibility problems that had never surfaced because the material had never lived in a system that exposed them.
Approach
- Template first, build second. Common module naming and sequencing, a consistent landing page, predictable placement for syllabus, readings, and assessments. Ten courses built ad hoc produce ten different student experiences. A template turns each later build into configuration rather than invention, which is what makes this scope survivable solo.
- Development document to module map. Each map lays out what the learner works through, in what sequence, and what they submit to show they have learned it. Where faculty notes described an outcome with no matching activity, I built the activity and confirmed it against the instructor's intent.
- Built the missing assessment layer. A large share of this work was creating assignments and rubrics that did not exist. In a face-to-face section, grading criteria often live in the instructor's head and get delivered verbally. Online, that fails. Every assignment needed explicit submission requirements and an attached rubric.
- Remediated for accessibility during the build, not after. Heading structure, descriptive link text, alt text, table structure. Fixing accessibility inside the build costs a fraction of retrofitting a published course.
- Repaired inherited content debt as encountered. Dead links and degraded resources. Unglamorous, and the fastest way to lose learner trust if skipped.
Status, August 2026: six of ten courses delivered, meeting the first milestone on schedule, with the remaining four due September 1. The client has since engaged me for a second contract.
What I would do differently
- Formalize the template on day one. I arrived at a consistent structure through the first courses rather than declaring it up front as a written build standard. That standard, agreed with the program lead before course one, removes rework and tells faculty exactly what they are getting.
- Audit source material as a discrete first phase. Broken links and missing assessments surfaced course by course. Cataloging that debt across all ten at the start would have produced a truer effort estimate and made the assessment gap a single program-level conversation instead of ten separate ones.
- Standardize the intake. Development documents varied in completeness. A short intake form covering objectives, required assessments, and non-negotiable sequencing would cut the interpretation load and the clarification cycles.
- Public university, Colorado
- Ongoing
- ADA Title II and state mandate
Building a faculty-led path to WCAG 2.1 AA conformance
Two separate mandates apply to this institution, and they point in the same direction.
Colorado HB21-1110 requires public entities, including state-supported institutions of higher education, to bring digital content into conformance with WCAG 2.1 Level AA. The original date was July 1, 2024, with a subsequent bill creating a one-year grace period to July 1, 2025 conditional on demonstrated good faith effort. Institutional work began after that date, so the state clock was already running.
Federally, the Department of Justice rule under ADA Title II adopts the same standard, WCAG 2.1 Level AA, and names online course content and digital documents explicitly. An interim final rule issued in April 2026 moved the compliance date to April 26, 2027 for public entities serving populations of 50,000 or more. DOJ guidance calculates that population at the state level rather than by enrollment, which places nearly every public university, including this one, in the earlier tier.
So the real problem is not a missed deadline. It is a live federal deadline roughly eight months out, in a content domain the rule names directly, at an institution that has not yet built the capability to meet it.
That framing dictates the strategy. Eight months is not enough time for one administrator to remediate a university's entire Canvas instance, and it never was. It is enough time to make several hundred content authors capable of remediating their own.
Constraints
- Two clocks, one already expired and one running. Whatever I built had to start producing results immediately rather than after a long design cycle.
- Scale against staffing. A university Canvas instance holds far more content than one administrator can personally remediate. Direct remediation does not scale and does not finish inside the window.
- Course content belongs to instructors. A central administrator rewriting faculty material is both a governance problem and a bottleneck.
- The federal rule covers content created continuously. Every new course, document, and video adds to the obligation, so a one-time cleanup decays immediately.
- No baseline existed. Nobody could say how many issues sat in the instance, which meant no one could size the work or show progress against it.
Approach
The central decision was to build capability rather than perform remediation. A model where one person fixes courses caps out at that person's hours, finishes nothing by April, and leaves the institution equally exposed the following semester. A model where instructors can find and fix their own issues compounds, and compounding is the only property that closes a gap this size in the time available.
- Built the training as a Canvas course. Instructors learn the standard inside the environment where they will apply it, which removes the translation step between abstract guidance and the interface in front of them. Building it in Canvas also makes it trackable, updatable, and available asynchronously to distributed faculty.
- Developing a supporting resource library. Reference material and job aids for the person who has a specific problem right now, as opposed to the person who wants to learn the topic. Both audiences are real and they need different things.
- Standardized on the accessibility checker already built into the platform. No procurement, no extra login. A more capable third-party scanner that instructors never open is worth less than a native tool they will actually use.
- Coach instructors to remediate their own courses. Deliberate. It keeps content ownership with the authors, scales past my available hours, and means the next course they build starts accessible instead of queuing for another remediation pass.
- Held the line on direct platform support. Alongside the enablement work I troubleshoot and repair courses and configure LTI integrations continuously. Faculty adoption of a new standard depends on the platform underneath being reliable.
Status: active program, not a finished project, running against an April 26, 2027 federal compliance date. Training course built and available, resource library in development, coaching model in place with instructors remediating their own courses.
What I would do differently
- Instrument it from day one. The largest gap in this program is measurement. Without a baseline of issues per course I can describe activity but not progress. That matters more under the federal rule than it did under the state one, because the limited exceptions turn on documented rationale and demonstrated effort. A measured reduction is evidence. "We are working on it" is not. Establishing the baseline first would have cost a week and would be paying for itself now.
- Attach it to a policy hook. Training that is merely available reaches the people already inclined to care. Training tied to something, a course publication check or an onboarding step, reaches everyone else. Voluntary adoption alone will not reach conformance.
- Address new content, not just the backlog. The existing backlog is the visible problem. Every newly authored document and uncaptioned video adds to it. Building requirements into the point of creation prevents more debt than remediation retires.
- Separate quick wins from deep remediation. A short list of high-frequency, low-effort fixes moves aggregate numbers faster than treating every issue as equal, and gives faculty an achievable first step instead of an intimidating one.