Expertise Scaling: The Mission Underneath CourseAgent
A few months ago, I was talking to the head of a small but important environmental charity. They’re one of the few organisations in the country with genuine, field-tested expertise in Biodiversity Net Gain assessments. A new law has just made BNG mandatory for most new developments in England, creating a sudden, massive demand for skilled ecologists.
This charity should be in a perfect position. They have the knowledge. They have the proof. What they don't have is a way to teach it at scale.
Their expertise lives inside the head of one senior ecologist. Let's call her Jane. Jane is phenomenal at her job, but she is maxed out doing the work itself – surveying sites, advising developers, training the two juniors on her team. She has no time to stop and design a course that could train fifty, or five hundred, other ecologists across the country. The charity's impact is capped by Jane’s calendar.
This isn't a unique story. I’ve seen the same pattern play out for twenty years, across every sector. In healthcare, it’s the senior consultant who has perfected a diagnostic technique but can't get it out to the junior doctors. In finance, it’s the one person in the firm who truly understands the nuances of a new piece of financial crime regulation. In law, it’s the partner with a career’s worth of experience in a niche area of contract law.
The expertise is there. It’s valuable. It’s often critical. But it’s trapped.
The bottleneck isn't content. It's craft.
When we talk about ‘scaling expertise,’ the conversation often defaults to a content problem. People assume the challenge is getting the words out of the expert’s head and onto a screen. So we give them a blank PowerPoint deck or a Word document and say, “Just write down what you know.”
It never works.
Jane, the ecologist, could write a 20-page document about BNG. It would be factually perfect. And it would be almost impossible for a novice to learn from. Why? Because expertise isn't just a collection of facts. It's a way of seeing the world. It’s knowing which details matter and which don't. It’s understanding the common mistakes and how to avoid them. It’s the instinctive judgement that comes from practice.
Simply writing facts down doesn't transfer this. You need to structure the learning. You have to design it.
You need to think about clear objectives. You have to break complex ideas into manageable chunks. You have to build in activities that let people practise, not just read. You have to create scenarios that mimic the real world. You have to assess whether they can apply the knowledge, not just recall it. This is the work of instructional design. It is a specific, professional craft.
And that craft is the bottleneck. Most subject matter experts are not instructional designers. They are busy people, brilliant at their own jobs, who are being asked to do a second, entirely different job for which they have no training and no time. The friction of authoring – the sheer difficulty of turning raw knowledge into a well-designed learning experience – is what stops expertise from scaling.
Organisations then face a stark choice: either pay a fortune to an agency to interview Jane and build a course over several months, or let her knowledge remain locked away. For a small charity, that's no choice at all.
Removing friction, not replacing people
This is the problem I wanted to solve when building CourseAgent AI. The goal was never to replace course creators or to automate expertise itself. The goal was to remove the authoring friction that keeps expertise trapped.
Imagine if Jane could take her messy field notes, a few internal reports, and a link to the government’s BNG guidance, and drop them into a system. And imagine if, a few minutes later, she had a draft course – not just a wall of text, but a properly structured learning pathway with introductory sections, core modules, interactive flip-cards to explain key terms, a scenario-based exercise to test decision-making, and a meaningful final assessment.
The AI hasn’t invented the expertise. It has simply applied the craft of instructional design at speed. It provides the scaffolding. It suggests the right kind of activity for a particular learning objective. It imposes a pedagogical structure that a non-designer would struggle for weeks to create.
Jane is still in complete control. She is the expert. She’ll edit the text to get the nuance right. She’ll tweak the scenario to make it more realistic. She’ll add a critical detail the AI missed. But she is now editing, refining, and validating – not wrestling with a blank page. The cognitive load has shifted from ‘how on earth do I build a course?’ to ‘is this course, which already exists, correct and complete?’
That is a fundamentally different and more achievable task for a busy expert.
This isn't just about efficiency. It's about access. It means the small charity can suddenly punch above its weight, scaling its impact nationally. It means that vital, hard-won knowledge no longer walks out the door when a key person retires. It means we can build organisational capability much faster, and keep it current when the world inevitably changes.
The mission underneath it all isn't to build AI tools for their own sake. It’s to find every place where valuable human knowledge is stuck, and build a better, faster, more accessible way to get it moving.