Subject matter experts

In her article, “Has AI Finally Fixed L&D’s SME Problem?”, Dr. Philippa Hardman describes the SME bottleneck as one of the most persistent problems in learning and development. Her point is important: The bottleneck isn’t just caused by the fact that subject matter experts are busy people. It also arises because these experts are often unable to fully explain their own expertise.

Hardman distinguishes between two issues here. The first is access: L&D has to wait for interviews, documents, reviews, and approvals. The second is articulation: Even if the subject matter expert is in the room, much of their actual expertise remains unspoken. Experienced people rely on patterns, shortcuts, warning signs, and decisions that feel like second nature to them.

Which is precisely why the traditional solution falls short. Another interview guide, another reminder, and another AI-generated transcript speed up the process, but they do little to change the underlying issue. L&D remains dependent on sporadic availability and has to continue to develop an effective learning product from fragmented information.

Hardman cites specific figures from the State of Instructional Design Survey conducted with Synthesia: For 30 percent of the designers surveyed, SME delays were a speed issue, and for another 13 percent, they were a quality issue. She also explains that a typical eLearning project lasts 10 to 12 weeks and includes several SME review cycles, each of which takes one to two weeks.

The real problem, however, is more than just these numbers. The SME bottleneck is not just a minor hiccup in the process, but a structural constraint on the entire L&D function. The faster AI produces drafts, scripts, quiz questions, and course structures, the more expensive it becomes when human expertise is lacking or comes in too late.

That’s the future challenge for L&D. AI makes content production more affordable. But that doesn’t automatically make learning better. On the contrary: As production becomes cheaper, judgment, context, and quality become scarcer. That’s exactly where you’ll see whether L&D becomes more strategic or gets lost in a flood of mediocre content.

Hardman describes four AI use cases we are already seeing today: AI analyzing SME interviews, AI conducting asynchronous interviews, AI serving as a research assistant, and AI analyzing an expert’s existing output. These approaches make sense. They reduce friction, shorten wait times, and help L&D ask better questions before the next expert meeting.

But every approach has its limits. A perfect transcript isn’t complete if the conversation itself didn’t cover everything. An AI interview is still an interview and still depends on what the subject matter expert can articulate. A research assistant is familiar with the published literature, but not necessarily with a company’s actual practices. A virtual SME created from documents only knows what has already been filed away somewhere.

This is exactly where L&D needs to take a clearer stance. AI as an extraction tool is helpful, but defensive. It makes the old model more efficient. The greater opportunity lies in an offensive model: Departments becoming self-sufficient in production, L&D building the operating system to support this, and the authoring tool providing the technical work environment.

One possible next step in this development is to promote more responsible knowledge production at the point where knowledge is generated. Subject matter experts become authors, but not lone wolves. They work within a controlled system consisting of templates, AI support, instructional design guidelines, reviews, and centralized publication.

This is employee-generated learning in its mature form. It means that companies can free content creation from the SME bottleneck while at the same time making quality standards binding.

In this model, the work shifts:

• The subject matter expert provides real-world examples, specialist logic, common errors, exceptions, and decision-making scenarios.
• AI helps to develop structure, learning objectives, texts, media, questions, translations, and variations based on all this data.
• Authoring tools such as Knowledgeworker Create manage content, versions, reviews, and publication with a single, centralized process.
• L&D establishes criteria, trains authors, evaluates effectiveness, and co-authors content on sensitive or didactically challenging topics.

This is more than just process optimization. It’s a new architecture for corporate learning.

LVB shows something else, too: Decentralization only works if it’s tied to a shared platform. The one platform principle ensures that employees know where to learn. The user-centricity principle ensures that content remains discoverable and accessible to various professional groups. The make not buy principle ensures that internal practical knowledge is not replaced by generic content.

The result is strategically powerful because it resolves three long-standing conflicts. First: Specialization and scalability are not mutually exclusive. Second: Decentralized production and centralized quality can go hand in hand. Third: L&D doesn’t have to develop every learning module itself in order to maintain responsibility for learning quality.

And that is precisely where the vision for the future lies. In the coming years, L&D won’t be measured by how many courses the team produces on its own. It will be measured by how effectively it facilitates the flow of knowledge within the company. The subject matter expert is no longer treated as a bottleneck, but as part of a learning network.

If subject matter experts can create learning content themselves using AI, external support may seem less important at first glance. But this is exactly where it’s worth taking a second look. The question is: Which tasks are closely tied to subject matter expertise, and which benefit from experience, distance, and professional production routines?

External support is still useful when projects are sensitive, didactically challenging, or large in scale. When it comes to compliance topics, security-critical content, complex product training, or strategic learning programs, a quick course outline is rarely enough. On top of that, there’s learning architecture, media quality, accessibility, review logic, and a clear synthesis of many expert opinions to think about.

Partial outsourcing is also gaining importance. A company can empower subject matter experts with Knowledgeworker Create while still involving external partners for design, co-authoring, quality assurance, or scaling. This keeps internal knowledge internal, while external learning expertise provides support where it can enhance effectiveness, speed, or security.

In practice, this manifests in three useful ways:

• Subject matter experts create initial drafts while external instructional design professionals refine the structure, learning objectives, and knowledge transfer.
• L&D manages the program; external authors handle specific parts of the production when the volume is high or the timeline is tight.
• Critical courses receive full support when the risk, target audience, or visibility is particularly high.

This allows for more targeted use of external support. It helps companies build a scalable system in which internal authors improve their skills and L&D isn’t left to juggle quality, speed, and capacity all on its own.

The best place to start is with an issue where the bottleneck is already causing problems. Choose a project with a clear subject matter focus, a visible need for updates, and a target audience that directly benefits from the knowledge. Product training, technical instruction, onboarding, and process knowledge are particularly well-suited for this.

Next, select two to four subject matter experts who have strong technical expertise and are open to working with digital formats. Give them Knowledgeworker Create, a clear template, and a brief instructional design framework. L&D then provides them with guidance on initial drafts through co-authoring, identifies common pitfalls, and uses this information to develop standards for other authors.

This step is cultural. Don’t say: “We’re turning subject matter experts into mini L&D departments.” Say: “We’re making expertise shareable without taking it out of context.” This approach changes attitudes. Subject matter experts don’t see themselves as an additional production resource, but rather as masters of their own knowledge.

If you’d like to test this approach in practice, you can try Knowledgeworker Create for 30 days and get started with a real topic you’re working on at your company. The test phase is particularly well-suited for incorporating an initial subject area and realistically evaluating the process that takes you from expert knowledge to a finished online course.