LMS at 30 Part 2: Learning Management in the AI Era

Our coverage at On EdTech is shifting to look more at big picture impacts and less on day-to-day changes in markets, particularly for the LMS market. Today Matt Pittinsky, co-founder of Blackboard and currently a board member at Instructure via the acquisition of Parchment, further explains his perspective for future developments based on his initial post. As an update, I have had two other founders / early executives of early LMS companies who accepted my request for further guest posts, and we will share those in September - October. Of course, this article is Matt’s and represents his perspective. You can find more of Matt’s writings on his site. - PH

I didn’t write “The LMS at 30” with a plan for it to be “Part One” of a two-parter. I simply sought to document an anniversary that I thought deserved acknowledgement, in large part because the project that gave birth to the LMS, the Instructional Management Systems (IMS) project, deserves more recognition for the rapid adoption of eLearning than it receives. Then a funny thing happened.

Phil Hill was the first person to point it out. What I actually ended up writing was an argument that the LMS paradigm hasn’t changed much over the past 30 years, that the product category remains course management in practice even as it rebranded to learning management, and that AI is poised to change the paradigm of course management to true learning management . . . at last.

I shouldn’t be surprised, then, that the response so far focuses more on my closing paragraph, from “you buried the lede,” to “say more.”

For those who haven’t read “Part One,” here is what I wrote towards the end:

In this Part Two, I’ll do my best to explain what I mean.

DIsruptive Technologies and Paradigm Shifts

It is commonplace to call AI a “disruptive technology.” It is also wrong, or at least incomplete. My reading of Clay Christensen’s Innovator's Dilemma is that a technology is neither sustaining nor disruptive in and of itself.  Rather, context matters. Take Google for example. AI is clearly disruptive to Google search as more and more people turn to AI chat services to ask and answer questions. In this way, you can say AI is enabling a paradigm shift in search, from users entering keywords and clicking on links, to users asking questions and receiving an answer. This paradigm shift has massive downstream consequences for content producers and advertisers, not to mention Google itself!

At the same time, AI is a sustaining technology to Google Cloud Platform (GCP). AI expands the value GCP can deliver through its infrastructure and provides a basis of competitive differentiation for Google, the third place player in cloud services, against Amazon Web Services (AWS) and Microsoft Azure.

In the case of LMS, AI is sustaining in the ways it makes course management workflows 100x better. For example, LMS providers have launched capabilities that allow an instructor to describe in plain language a series of actions they want taken and the LMS will execute them without lots of clicks across multiple screens – create an assignment, set the due date, give certain students extensions, etc.   

Where AI is disruptive is in its ability to introduce a whole new set of capabilities that are best described as personalized learning services. AI offers a new value proposition to the LMS, roughly the set of capabilities currently being developed in the AI Tutor / agentic TA segment. These new capabilities are so valuable given their impact on learning that I predict they will become the services with greatest engagement within a school or university’s “enterprise” instructional platform.  

In this way, by LMS paradigm shift, I specifically mean a shift from buyers valuing the product on its course-centric and course management capabilities, to valuing it on its learner-centric and personalized learning capabilities.

The Jobs to Be Done of an LMS

Consider two simple questions: What aspects of an LMS does AI make obsolete, and what new capabilities does it unlock?

Historically, LMS platforms have centered on three core functions:

1. Course Publishing & Management – Traditionally, this involved making as easy as possible (for novice instructors) a complex series of workflows for building course Web sites, uploading and sequencing content, developing assessments, setting permissions, and more. These are the primary jobs to be done by an LMS. When AI allows an instructor to engage in course publishing and management through natural language prompts, dynamically generating structured course materials, assessments, and other instructional assets with the uploading of a syllabus, a broad swatch of what the LMS does changes in form, if not function. Not the “jobs to be done”, but the click-after-click after-click workflows that are how these jobs are achieved. Generative AI eliminates much of the manual overhead, making course setup nearly instantaneous while also rendering traditional LMS workflows for content curation and deployment increasingly superfluous. Asking for a test bank that assesses the philosophical tradition Pragmatism is a 100x better experience than today’s LMS workflow approach.

2. Communication & Collaboration – Discussion forums, messaging systems, and group collaboration tools have been mainstays of the LMS, designed to replicate and extend classroom interactions. For synchronous learning, the market has bounced between vertical tools such as Horizon Wimba or Blackboard Collaborate (now Class Technologies) and horizontal tools such as Zoom and Teams. Asynchronous tools have consistently been vertical and mostly LMS native. AI-enhanced platforms will function as intelligent intermediaries, synthesizing discussions, summarizing key takeaways, and even facilitating contextualized learning pathways based on engagement patterns. The LMS will evolve from simple facilitation to active, adaptive engagement through native and third-party tools. That said, I don’t see the relative “weight" of this function changing in an LMS paradigm defining way.

3. Instructional Orchestration – The most underdeveloped, yet arguably most critical function of the LMS, is instructional orchestration – helping students and instructors engage the learning process itself. Today, it remains surprisingly difficult for a student to know exactly what they need to do on any given day; what assignments are due, which late work policies apply, and how best to organize their learning. AI can transform this experience by dynamically interpreting course policies, pulling relevant deadlines from syllabi, and personalizing study recommendations based on performance data.  

For example, instead of a student having to manually check different sections of the LMS for assignment deadlines and late work policies, AI can proactively gather this information from various sources – Google Docs, course syllabi, or discussion boards – and present it in a clear, actionable way. As a second example, consider the custom course materials development capabilities of AI, put to work for each student based on their mastery of concepts to date.

You could argue that there is a fourth function of an LMS, enterprise management of learning, by which I mean the capabilities an LMS provides central IT and academic leadership to structure, observe, and more tightly couple what happens in courses with institutional policies and priorities. This may not be popular to say, but one reason the category hasn’t expanded to support more innovation around instructional orchestration is the way in which institutions purchase an LMS, prioritizing endless feature checklists and the needs of institutions, which incentivizes vendors in all the wrong ways. The LMS is overbuilt.

Whether three or four functions, my point is that the weights are changing. My very rough estimate is that today’s LMS is ~70% Course Publishing and Management, ~20% Communication and Collaboration, and ~10% Instructional Orchestration. 

In the AI era it is easy to imagine that the weights flip, with personalized learning services (aka instruction orchestration) being 70%. In that world, if you believe in the strategy of commoditizing your complements, you could imagine course management being given away free for the shell it creates to deliver personalized learning services – actual facilitation of teaching and learning. A true LMS experience could be organized around the learner, not the course, anyway.

I am fascinated to see if new entrants emerge with simpler, more affordable offerings on the course management side that satisfy this new value framework, expanding over time to subsume the capabilities of a traditional LMS. If this happens, the paradigm shift will have occurred.

Big Questions and Second Order Effects

My strong conviction is that “if this happens” is really “when this happens.” But there are a few questions and consequences that keep popping up where my thinking remains frustratingly fuzzy.

What will happen to the course materials market? In a world where generative AI is capable of creating course materials, and the LMS draws on content in the course to create derivatives in the form of flashcards, podcasts, and quizzes, is there still a commercial course materials market and how does it work? I wonder if AI will accelerate the adoption of inclusive and equitable access licensing arrangements, providing institutional AI models with high quality materials that can be deployed to all stakeholders in any format. As a second example, how does the regulatory environment respond?  In my home state of Arizona, K-12 parents have a right to review the core course materials their students are “assigned.” Does the law allow for dynamic, personalized content?

How do the preferences of institution, instructor, and AI LMS manifest? In a world where students learn from the LMS, from the instructor, and from the institution, all nested, how much of the agent experience is a reflection of each? And how does that get operationalized? I once wrote a piece advising entrepreneurs to “beware proximity to the instructional core” as it is the most political and fraught place to be, despite its central value to the “business” of the customer, a school or university. LMS 2.0 is smack dab in the middle! Can the LMS continue to be pedagogically agnostic? 

Are instructors willing and able to redesign for the new technology? To be blunt, AI is hollowing out learning. From essays, to classroom contributions, to research, it is difficult, if not impossible, for faculty to assess what represents mastery and engagement with concepts, and what is copy and paste. This compounds the effects of empty credentialism, which already drives grade inflation. Putting in place guardrails with school AI is meaningless when students can flip to the consumer Internet. Is the answer a sort of reverse migration from analog to digital back to analog? What does authentic assessment look like?  What does spoken word assessment look like? We need answers, quickly.

What will students choose? We have a tendency in B2B EdTech to assume that what schools adopt, students will use. Yet we live in a world of increased EdTech abundance, by which I mean more and more substitution options on the consumer Internet for what schools used to provide, on a monopoly basis, as part of their learning environment. Where will students engage in their “jobs to be done” for learning as AI becomes more central to the learning experience? The LMS, or elsewhere?

How big can the LMS market become? This is a question not just for the supply side (aka, “vendor” side), but for schools for the reasons described in Part One – the smaller the core market the more LMS companies need to find growth in other segments and other products. It seems logical that an LMS that is deeper in the learning process is an LMS that can access more spend than just enterprise software for course delivery and support, and even enterprise software for higher education (or K12) generally. We should expect LMS spend to “eat” spend on tutoring, course materials, and even labor. AI allows software to do more, and education is a sector that is relatively under-penetrated by software spend. There is even an outside chance that an LMS, conceived as a supra-institution platform, could aggregate learners, learning activity, and institutions into a single network – one consistent with Ben Thomson’s Aggregation Theory, where demand attracts supply, marginal costs approach zero, and scale feeds on itself. If that happens, the LMS would no longer be just infrastructure for education, but the platform through which the entire sector organizes.

Closing Thoughts

Paradigm shifts carry an air of inevitability looking backward, but are in many ways contingent processes in their realization. A particular entrepreneur. A particular path dependence. A particular product-market fit. Look no further than the addition of chat to GPT 3!  

If the LMS plays an active role in the teaching and learning process, “agentic LMS,” how will that active role look, taste, and feel? I think you can believe that AI will be transformative and also believe AI is a normal technology, meaning the paradigm shift underway will develop within certain familiar contours.  

In architecture and campus planning there is something called the ‘desire paths’ approach, where after architects add a new building to a campus, rather than pre‑designing all the walkways, they leave the grass intact. They then wait for students to naturally walk to and from the building, relying on those paths as the eventual walkways.

When I think about the behaviors most likely to shape the specifics of agentic LMS, I try to ground myself by remembering what I consider the first principles of education.

1. Education is a social process - We will expect the LMS to be an agent in the learning community, as well as foster richer interactions between instructors and students, and among students themselves. For example, dynamic establishment and facilitation of peer interaction.

2. Education is an experiential process - We will expect the LMS to support a greater emphasis on "learning by doing" and iterative trial and error. For example, AI-generated simulations, adaptive problem-solving environments, and feedback loops that guide students through mastery of concepts. Not for nothing, I still believe in the potential of "spatial computing!”

3. Education is a developmental process - We will expect the LMS to support institutional models for recognizing and guiding learners’ cognitive, emotional, social, and moral growth. With agentic systems like ChatGPT already forming emotional bonds with students, institutions must define the developmental frameworks they want encoded, and ensure educators remain central in interpreting and directing growth.

4. Education is not the same as school or schooling - We will expect the LMS to operate across formal and informal contexts, since learning extends beyond classrooms and institutions. This means integrating internships, personal projects, and even learner-permissioned external apps into a coherent record of progress.

5. Education has multiple social purposes - We will expect the LMS to flex across education’s multiple, often conflicting goals – economic development, individual mobility, and democratic equality. The closer the LMS gets to actual learning services and agentic engagement, the more it needs to be aware of these goals, prioritized differently by different institutions. For example, it should look and behave differently for a community college focused on workforce preparation than for a liberal arts college centered on civic and intellectual formation, or for a neighborhood school versus a boarding school.  Here is the trick… without tracking.

6. Education’s charter is the credential - We will expect the LMS to strengthen education’s core function of credentialing by tracking skills and competencies in more granular ways. AI can help students assemble dynamic portfolios and new credentials that remain socially recognized and legitimate, ensuring their achievements are valued beyond the institution. And, equally importantly, ensure the university’s charter – legitimizing purpose – remains intact, if not strengthened.

Wrapping up this two-part very long-form article, I can’t help but feel incredibly lucky and grateful. Lucky to have been born when I was born: old enough to have experienced the Internet’s impact on education, and young enough to now experience AI’s impact. And fortunate for the many mentors, colleagues, and yes, business competitors, from whom I’ve learned to make the most of this luck. 

The main On EdTech newsletter is free to share in part or in whole. All we ask is attribution.

Thanks for being a subscriber.