Don't tell HR, but AI did my compliance training...

... and then inspired me to revolutionize the world of ladder safety.

Last week I received a reminder email for compliance training…

Normally this means I need to complete yet another online sexual harassment training, but this time it was different.

Apparently, it was essential that I, a remote online professor of astronomy, must complete a training course in ladder safety.

As a remote online professor, I will never set foot on this college campus, nor any of its ladders. But the training is still required.

Such an obviously ridiculous request seemed like the perfect opportunity to try an equally ridiculous experiment. And so I asked OpenAI’s AI agent, Operator, to complete the training for me.

(For an introduction to Operator, check out my prior post, “I built an AI agent to respond to student discussion posts.”)

Operator gets a B+ in ladder safety!

Operator dutifully watched all of the videos, completed the 9-question quiz, and earned an 88% on the course. A few noteworthy things stood out to me.

1. Operator worked for 19 minutes straight (the full length of the training) without stopping once to ask me a question. It completed the whole course without intervention.

2. Operator got just one question wrong (hurray!). But it makes me wonder… why did it get any questions wrong at all?

Curiously, Operator got one of the questions wrong on the final assessment. I don’t know why.

Implications for the future of corporate e-learning

This simple experiment has some interesting implications for the world of compliance training and corporate e-learning - areas where content must be delivered at enormous scale with no facilitation.

On the one hand, you can view this experiment as an existential threat to this kind of online training. If employees can cheat so easily, then this kind of training serves virtually no purpose.

On the other hand, compliance training generally sucks and may deserve to die.

My positive vision for an AI-powered future is that compliance training could someday be a conversation between the employee and an AI trainer. The conversation could start by exploring the user’s prior knowledge and then move into specific scenarios that are relevant to the user’s role in the company.

That’s a cool idea, right? I wonder if I could build it…. Seeing as how this is the “Experimenting with AI in education” Substack, I’m going to try.

Let’s build an AI-powered ladder safety training experience that eliminates cheating, not because it is impossible to cheat, but because the training is so engaging, relevant, and efficient that employees are happy to complete it.

My hope is that this simple experiment will demonstrate how AI can be used as a powerful tool for increasing engagement in education, even compliance training. After all, if you can make ladder safety interesting, then any other subject should be a breeze.

Andrew steps away from writing for 20 minutes to create the bot…

Meet LadderSafetyBot: The next generation in ladder safety compliance training.

This is LadderSafetyBot. And yes, he is breaking the #1 rule of ladder safety - keep at least 3 hands/feet on the ladder at all times.

Ok, I’m actually shocked how quick and easy it was to build my prototype and equally surprised how much it improved the learning experience over existing, market-leading solutions.

After introducing myself to the chat bot, the promise and potential of using generative AI for compliance training became immediately obvious.

I can’t wait to show you, but first, let me tell you how I created my tool:

1. I went to OpenAI’s GPTs page and clicked +Create to make my own custom GPT (This requires a ChatGPT Plus account.)

2. I used regular ChatGPT to create detailed instructions for the bot. Beyond the obvious request (make a ladder safety trainer), I also asked it to limit the conversation to 5-7 interactions and then provide feedback to the user.

3. I Googled OSHA ladder safety resources, downloaded a few PDFs, and uploaded them as Knowledge to my custom GPT.

That’s it. Now I’m ready to have a conversation about ladder safety and see what I can learn. Here is the start of my very first discussion with LadderSafetyBot:

---

“Hello. My name is Andrew, and I am an online professor of astronomy.”

“Welcome, Andrew! It's great to meet an astronomy professor interested in ladder safety—especially since accessing telescopes, adjusting ceiling-mounted projectors, or changing lighting in observatories or lecture halls might occasionally involve ladder use.

Let’s kick things off with a quick question: imagine you're in a campus observatory, and you need to replace a light fixture that's just out of reach. You grab a ladder from the supply closet. What are three things you’d check before stepping onto it?

Take your best guess! We'll use this to build your ladder safety savvy from the ground up.”

---

I’m embarrassed to admit this, but until getting to this point, it never occurred to me that astronomy professors do, in fact, have some special cases where they use ladders more than other faculty: working in observatories.

Ok, you may be thinking: “Duh, Andrew! That should have been obvious to you.” But the entire time that I was preparing this blog post and creating my custom GPT training bot, I thought ladder safety was so totally irrelevant for me as to be laughable.

The required ladder safety course (the one Operator completed for me) had videos of men wearing hard hats and carrying ladders outdoors to climb on rooftops. There were janitors replacing ceiling lights in the hallways. There were harnesses and safety equipment I’ve never seen before - surely this training was NOT for me.

It was only when LadderSafetyBot suggested I may need to use a ladder in an observatory that I suddenly realized I had been entirely mistaken - ladder safety training IS for me!

(In fact, the simple act of providing a relevant example reminded me that just two weeks ago, I was in an observatory, climbing a very steep ladder.)

Personalized learning in action

We’ve dreamed about personalization in education for a long-time. That used to mean giving students more choices in their coursework and assessments. With AI in hand, we’ve begun dreaming about a new kind of personalization, where every student can have a unique and customized learning experience, based on their prior knowledge and interest.

To me, the promise of AI-powered personalized learning has always felt like an abstract idea that lives somewhere in the future (or on some EdTech entrepreneur's pitch deck).

But my simple experience of interacting with the LadderSafetyBot has been a powerful illustration of how personalizing a topic can make it so much more engaging and relevant for the learner (in this case, me).

---

The future of compliance training is bright

I continued my conversation with the cute little compliance training chatbot for another couple of minutes. LadderSafetyBot used analogies that felt tailor-made for an astronomy professor - relating the “three-point contact” rule to a telescope’s tripod and telling me that I was on the right “wavelength” in my responses.

All-in-all, I was impressed by the experience. I found it to be much shorter, much more engaging, and honestly more impactful than most of my prior compliance training experiences:

• It forced me to an active participant in the learning, which quickly revealed the limits of my knowledge

• It was customized to my particular role at the college, which improved my engagement and made it feel relevant.

• It diagnosed gaps in my understanding and specifically worked to improve those areas.

Try LadderSafetyBot and tell me what you think.

In the spirit of experimentation, I would love for you to try my LadderSafetyBot for yourself. Introduce yourself and provide a description of your role at an organization.

I would love to hear from you about your experiences or your reflections on this blog post. If you are not comfortable posting a public comment, then email me at avheuv@gmail.com with your personal reflections and experience.

I appreciate you taking the time to read this post. If you liked it, then please consider subscribing or sharing it - that’s all I ask in return.