Executive summary

L&D professionals have always known something that the rest of the organization is only now catching up to: good learning does not happen by accident. It requires intentional design, clear objectives, and a real understanding of how people build capability. That expertise is not incidental. It is the foundation of every effective training program, every onboarding that actually sticks, every enablement effort that moves the needle.

The challenge has never been the quality of what L&D builds. It has been the gap between what L&D can build and how much the organization actually needs. Research consistently finds that the vast majority of how people develop at work, estimates range from 70 to 90 percent across multiple studies, happens outside formal programs: the Slack thread that explains a new tool, the colleague who walks someone through a process, the slide deck a team put together after figuring something out. That learning has always happened. L&D has always known it was happening. And until now, there was no practical way to bring the same rigor and quality to it that L&D brings to formal programs.

70–90%

of how people develop at work happens outside formal L&D programs

Read more

AI changes that equation. The volume of informal learning has grown exponentially as teams adopt new tools, rebuild workflows, and share knowledge faster than any formal program could keep pace with. This is the fundamental AI Acceleration Gap all businesses and leaders are facing: how do organizations plug the skill and knowledge gaps required quickly enough across their entire workforce to survive in the age of AI?

The good news: AI also makes it possible, for the first time, to embed instructional design principles directly into the tools that anyone uses to create learning. The expertise L&D has always held can now scale far beyond what L&D could ever produce directly.

The real opportunity inside the AI Acceleration Gap is not asking L&D to do more with the same resources, but giving L&D the mandate to architect quality across all learning in the organization not just the 20 percent they formally own. The shift is from creator of training to Capability Architect. And it follows three steps:

Enable.
Empower teams to create learning in the flow of work, with L&D leading that shift rather than having it happen around them.

Embed.
Put instructional design expertise into the creation tools so quality is structural, not dependent on 
review cycles.

Amplify.

Surface the exceptional learning already happening across the organization. Scale what works so the system improves over time.

In summary
L&D professionals have always known how to build learning that works. The constraint has never been their expertise. It has been their reach. Eighty percent of learning happens outside formal programs, and until now there was no practical way to bring L&D’s quality standards to that majority. AI changes that. For the first time, instructional design expertise can be embedded into the tools other teams use to create learning, extending L&D’s impact across the entire organization. Enable creation. Embed quality. Amplify what works. That is how 
L&D moves from the team that owns 20 percent of learning to the function that shapes all of it.

The AI acceleration gap

Ask any L&D leader what their biggest challenge is right now and you’ll hear some version of the same answer: the model we’ve relied on was never built for the pace of change we’re now facing. New AI tools are being adopted before training for the last ones is finished. Workflows that were stable for years are shifting week to week. Products are changing at an unprecedented pace. Employees are figuring things out on their own because they have to, and sharing what they learn through Slack threads, slide decks, or recorded calls.

This isn’t a failure of L&D, but the predictable result of a structural mismatch that has been building for years and that AI has now pushed into crisis. Organizations see the value of learning, and they’ve continued to add to learning demand without meaningfully investing in the capacity to meet it.

“Organizations have been adding demand without adding capacity, and AI just made that impossible to ignore.

The AI Acceleration Gap is the name for that mismatch: the growing distance between how fast work is changing and how quickly organizations can build systems to match it. And it’s widening in every direction at once. Teams across the world are each navigating new AI tools simultaneously, often arriving at different answers, with no shared standard for what good looks like.

Informal learning is already here

Eighty percent of how people learn at work happens outside formal L&D programs. It always has. The Slack thread that explains how a new tool works, the colleague who walks someone through a process, the ten-slide deck a team put together after figuring something out. This is how knowledge has always moved inside organizations. It is fast, it is motivated, and it is real.

The challenge is not that it exists. The challenge is that without instructional design behind it, quality is uneven. And as 
AI accelerates the pace at which teams are building new skills and sharing what they learn, the volume of that informal learning has grown exponentially — which means the quality gap has too. Three patterns show what this looks like 
right now.

80%

informal learning

20%

formal L&D

Read more

1. Teams adapt because waiting is not an option.
A procurement team figured out how to use AI to compress their sourcing cycle from 14 days to 3. They did it through trial and error and a shared Slack thread. They were motivated, resourceful, and entirely self-directed. The outcome was real. So are the questions nobody thought to ask: Is the method consistent across the team? Is anyone using a prompt that occasionally produces unreliable outputs? These are not reasons to discourage that kind of initiative. They are reasons to give it a better foundation.

2. Informal learning spreads unevenly.
When a customer service team cut tier-2 escalation resolution time by 60 percent using AI, they shared what they had learned through a Slack thread and a slide deck. The knowledge spread. The team improved. But three people were doing it one way, four were doing it another, and no one could say with confidence which approach was more reliable or repeatable. The learning happened. The quality of it was a matter of luck.

3. The gap gets filled by whoever is available.
When a company launched new AI contract terms that required sales training and the L&D team was already at capacity, Legal stepped in and created the training. Sales got what they needed. But consider what that moment reveals: the demand was urgent, the need was real, and the organization routed around its L&D function. This is not because L&D was not capable, but because the system was not designed to scale this way. The question is not whether informal learning will happen. It is whether L&D will be positioned to shape its quality.

Why the old model can’t scale

The Quality Gap Is Widening
The gap between good learning and poor learning has always existed inside organizations. Formal L&D programs bring rigor, design expertise, and clear learning objectives. Informal learning, in the Slack thread, the recorded call, the slide deck a team put together themselves, has always been faster and more immediate, but far less consistent in quality. That tradeoff is not new.

What AI has done is widen that gap in three specific ways, simultaneously and at speed.

Volume has gone vertical.
Anyone can now produce a quick explainer, a prompt library, or an AI-generated guide and share it as training without any instructional design experience. The volume of informal learning content circulating inside organizations has exploded. More content is being created and consumed than ever before but the proportion of it that meets any real quality standard has not grown with it. The gap between what exists and what is actually good has never been larger.

Shelf-life has collapsed.
A workflow guide written three months ago may already be wrong. Best practices discovered last week may be superseded before they are formalized. The faster content is created and shared, the faster it goes stale and the less likely anyone is to notice or correct it. Poor quality learning is now not just more prevalent. It is also harder to track and harder to retire.

The stakes are higher than ever.
In low-stakes contexts, inconsistent informal learning is an inefficiency. But when the learning in question is how to use a new AI tool that affects output quality, how to handle a customer escalation, or how to navigate a product change that sales needs to communicate then the cost of getting it wrong shows up in performance, in revenue, and in customer experience. The quality gap is not just wider. The price of it is higher.

Work is changing at a molecular level

To understand why learning demand is so hard to predict or contain right now, it helps to zoom in on what AI is actually changing and at what level.

Take any task: writing a proposal, handling a customer escalation, reviewing a contract. Each of those tasks is made up of smaller components: research, drafting, fact-checking, decisions, handoffs. AI is rewiring the individual molecules that make these tasks up.

Research that took a day now takes an hour. Writing that started with a blank page now starts with a draft. Fact-checking that required expertise now requires judgment about an AI-generated output. The task looks similar from the outside. The work inside it has changed.

This is why the pace of learning demand is so difficult to anticipate. Dozens of small, repeatable components across every role are being quietly reshaped in real time. And each one of those reshapings creates a learning need that no formal program was designed to address.

In summary
The unit of learning is changing. The era of the annual, quarterly, or even monthly training programs and the static course library is giving way to something more continuous, more granular, and more tightly woven into the actual flow of work.

From Content Creator to Capability Architect

The shift L&D needs to make is not without precedent. The same trajectory has already played out in design, DevOps, and platform engineering. In each case, the pattern followed the same arc.

Before accessible prototyping tools, designers were a bottleneck. When Figma put design capability into more hands, non-designers made things that often looked terrible. Designers responded by building design systems that embedded visual and UX principles directly into reusable components. The result: designers became architects of experience strategy, their thinking present in every product whether they touched it directly or not.

DevOps followed the same path. Before cloud and automation, IT operations teams were ticket-takers. Access to infrastructure was gated through specialists. When self-service tools made creation accessible to more people, the quality was inconsistent. DevOps engineers responded by building continuous integration and continuous deployment  (CI/CD) guardrails and delivery pipelines that embedded their expertise into the process itself. The result: IT ops became architects of the systems that enable organizational velocity.

Platform engineering did the same. When internal developer platforms put infrastructure creation into more hands, platform engineers responded by embedding their expertise into the platform itself: constraints, guardrails, reusable capabilities. The result: platform engineers became architects of the systems other teams build on.

In every case, the pattern was the same. Make creation accessible. Embed expertise into the tools so quality scales with volume. Then design the system that amplifies what works. The specialists did not become less valuable. They became the architects of a much larger surface area.

AI is creating exactly this moment for instructional design. The tools that let anyone create learning content are already here. The question is whether L&D will be empowered to build the systems, the standards, the guardrails, the templates, that ensure quality at the point of creation, regardless of who is doing the creating.

L&D’s expertise doesn’t need to shrink to fit the current model. The model needs to expand to match what L&D already knows how to do.

This is what it means to be a Capability Architect: not producing every piece of learning, but designing the conditions under which all learning, formal and informal, meets a standard worth trusting. It’s a bigger job. It’s a more strategic job. It’s one of discernment. And it’s the only approach that can actually scale with the pace of change.

The Model: Enable → Embed → Amplify

This shift hasn’t appeared out of nowhere. The same trajectory has played out in DevOps, design, and platform engineering. In each case, the pattern followed three consistent steps. For L&D, those steps are:

Enable: Make creation accessible
The first step is for L&D to lead the shift to decentralized creation.Teams are already creating informal learning. The question is whether they’re doing it within a framework that L&D designed, or entirely on their own.

“Enabling means getting AI-assisted authoring tools into the hands of subject-matter experts with L&D setting the conditions for how those tools are used.

Decentralizing creation is how knowledge moves at the speed of business, and it’s also how L&D finally gets leverage proportional to the demand.

Embed: Put expertise into the tools
Enabling creation at scale without quality standards just accelerates the existing problem. The second step is where L&D’s expertise becomes structurally irreplaceable.

Embedding means designing the guardrails, templates, and frameworks that live inside the creation tools so that when someone outside of L&D builds learning, the default output already reflects sound instructional principles. Quality can be  baked into the process before a single piece of content is published.

This is the design systems analogy made real. And it’s the work that no other function can do, because it requires a depth of instructional strategy that L&D uniquely holds.

Amplify: Scale what works
The third step is to identify the exceptional learning already being created across the organization and make it visible. Showcase quality examples. Surface them as models. Make the best learning experiences easy to find, replicate, and build on.

This creates a flywheel: enable creation, embed quality, then amplify the examples that demonstrate what good looks like. Over time, the system gets better. The floor rises. And the learning that happens outside formal programs stops being a liability and starts being an asset.

The playbook in one sentence
Learning is already happening everywhere—so leverage L&D experts to embed quality standards, amplify what works, and build the scalable structure that turns organizational risk into capability.

Proof: What this looks like in practice

Two organizations have already made this shift and the results are measurable.

In both cases, the L&D professionals became more valuable because their expertise now shapes ten or thirty times more learning than they could ever produce directly.

What L&D leaders should do next

This is your moment to step forward, not wait for permission. The Capability Architect role will not be handed to you. It has to be claimed. Here’s where to start.

1. Name the gap and make it visible to leadership.
The quality gap in informal learning is a business problem, not just a learning problem. You are often the first to see it clearly, but rarely the one with the budget authority to close it. Bring the evidence to the table: where is informal learning happening? What are the quality and risk implications? What would it look like to actually resource a response? L&D leaders who frame this conversation in terms of organizational risk and competitive readiness are the ones who 
get heard.

2. Make the case for the architect role.
Your ask isn’t simply more headcount. Articulate clearly what it means to operate as a Capability Architect: setting the standards, designing the systems, and enabling distributed creation, rather than producing all content centrally. This reframe matters for budget conversations, for executive alignment, for cross-functional collaboration and creation, and for how L&D is positioned in the organization. It may be a harder conversation than asking for one more hire, but it’s the right one.

3. Find and elevate the capable creators already in your organization.
In almost every organization, there are subject-matter experts already producing genuinely good informal learning. Find them and recognize them. Use their work as the visible standard for what quality looks like in practice. This builds credibility for the Capability Architect role, engenders trust from peers across the business, gives other creators a model to work from, and starts the flywheel of amplification without waiting for a perfect system to be in place first.

In summary
The organizations that build capability at the speed of change are the organizations that win. L&D holds the blueprint. The question is whether L&D leaders will step forward to build it.

From shadow learning to quality at scale

Learning happening outside formal L&D programs isn’t going to route back through a central queue. The pace of 
AI-driven change won’t slow down to match existing production cycles. And L&D teams who are already stretched thin cannot close the AI Acceleration Gap by working harder within a model that wasn’t built for this moment.

What they can do, given the right mandate and the right tools, is architect the conditions under which quality learning happens everywhere. Not by executing all content creation, but by designing the standards that make good creation the default. Not by being the sole source of learning, but by being the strategic force that ensures all learning is 
worth trusting.

That is the role of the Capability Architect. It’s more visible, more strategic, and more commensurate with the expertise L&D professionals have always brought to their work. It’s also the only approach that can actually scale with an organization navigating genuine, ongoing AI-driven change.

The gap is real. But the opportunity is larger. And L&D leaders are better positioned to close it than any other function in the organization. Their expertise has always been there. The tools to amplify their expertise are now here.

The organizations that build capability at the speed of change are the organizations that win. L&D holds the blueprint. The question is whether L&D leaders will step forward to lead the shift.