Back to portfolio

Managing AI learning assistants in Canvas LMS

Led the operational rollout and ongoing management of eleven LUMA assistants across ten university-partnered programmes, supporting learners and using engagement insights to improve the service.

At a glance

Context
FourthRev
Role
AI Enablement
Timeline
Pilot: February–April 2026 · Ongoing management: May 2026–present
Tools & Skills
LearnWise Canvas LMS ClickUp LTI AI configuration Testing and evaluation

Project overview

I led the operational rollout and now manage LUMA, FourthRev’s AI learning assistant, across ten programmes with Cambridge, LSE, King’s College London, and Monash.

Eleven assistants are live in Canvas LMS: one course assistant per programme, shared across cohorts, plus a shared career assistant that reviews CVs and cover letters within the Career Playbooks courses.

The challenge

The aim was to reduce repetitive support questions and give learners on-demand help within Canvas, including outside working hours. We also wanted insights beyond usage totals: what learners were asking, where they were getting stuck, and how their questions changed over time, to help teams improve course content and support.

Delivering this sustainably across programmes meant accommodating different content and support arrangements, keeping answers grounded in approved materials, and providing clear routes to human support.

About LUMA

Powered by LearnWise (opens in a new tab), LUMA explains concepts, summarises content, quizzes learners, and gives feedback on drafts and code. It also answers questions about course navigation, policies, procedures, and assessments. The career assistant provides structured CV and cover-letter reviews.

My approach

  1. Planning the rollout

    I planned the three-week pilot rollout in ClickUp and coordinated implementation with LearnWise’s implementation specialist.

  2. Defining assistant behaviour

    • Together with QA, programme delivery, student support, and helpdesk teams, we agreed how LUMA should answer common questions, communicate with learners, and direct them to human support when needed.
    • For the career assistant, I worked with the career coaching team to build structured CV and cover-letter review workflows. These checked length first, followed by structure, evidence, and suitability for the target role.
  3. Configuring, integrating, and testing

    • I configured the assistants and integrated them into Canvas through LTI, an API, and a JavaScript widget.
    • Working with colleagues across the product and delivery teams, I collected and coordinated the ingestion of approved course materials.
    • I organised testing with colleagues using likely learner questions and activities. We checked whether LUMA could explain concepts, guide learners through course content, and quiz them using the correct sources.
  4. Managing new deployments and maintenance

    • Following the successful pilot, I took responsibility for LUMA’s day-to-day management and deployment to new courses and cohorts.
    • I created a dedicated ClickUp space for deployment requests, knowledge updates, and staff access. Each knowledge update is tracked in a separate ticket linked to the relevant assistant.
    • I developed a testing process, deployment checklist, knowledge-base structure, and supporting documentation to make future rollouts more consistent. Documented procedures guide request handling and maintain a record of changes.
  5. Reporting on use and content gaps

    LearnWise produces standard monthly reports for programme teams. Working with the vendor, I expanded reporting beyond usage totals to include recurring questions and potential content gaps, helping teams review course information and learner support.

Introducing new features

As new LearnWise features become available, I coordinate their introduction into the service. These include voice mode, file uploads in chat, shortcut buttons, proactive notifications triggered by page location or time spent on a page, and staff access to live data.

One of the most anticipated updates was live data, which allowed staff to access trends, dashboards, and improvement suggestions within Canvas courses as well as through the LearnWise platform.

To support onboarding for new features, I create staff guides and walkthroughs using fictional data.

The impact

LUMA brought on-demand learning support into Canvas, while giving staff a shared service to manage routine questions and identify opportunities to improve courses.

  • Insights to improve course content

    Reports and dashboards gave programme teams visibility into recurring questions, learner difficulties, and potential gaps in course information. Trends and improvement suggestions helped teams identify where content or guidance needed review.

  • On-demand support for learners

    Reported return rates of 59–82% by programme showed learners coming back for support. Examples included project-related questions, critique on drafts, help with code, and learners returning to technical topics across several weeks. Support was available outside working hours, with clear routes to human help when needed.

  • Practical support for staff

    On some programmes, 42–45% of conversations concerned administration or finding resources, showing LUMA’s role as a first point of help for routine questions. Staff guides, walkthroughs, and dashboards helped colleagues use the service, while shared request processes supported consistent deployments and updates.

Reflection

“It was encouraging to see how staff engagement grew once colleagues could see LUMA helping with their own work. Reports about their learners made the value more tangible than a product demo. Colleagues began contributing FAQs, correcting answers, and requesting features without prompting.”

— Aida

Get in touch

Like this kind of work?

Happy to chat about full-time roles or projects — pick whichever option works best for you.