Prototyping with AI
A collection of listening, observing, communicating, chatting, and speaking from the Learning Technologies ’26 conference in London. There was one topic that dominated the Expo floor and nearly every session.
Human intelligence vs. artificial intelligence
Who is winning? Is this a competition? What is the hype and what is the reality today? Where is the learning going? Are we making a difference? What’s changing? What needs to change? Are we falling behind? Are humans interested in measuring impact, or the illusion of impact? Can we still connect as humans in the age of artificial intelligence (AI)? One word I took away from this experience: dialogue.
I wrote two screenplays. One of them was bad. But in the meantime, over the years, I’ve been working on my craft and creating good dialogue.
Dialogue is a conversation between two or more people or a written exchange between characters in literature, drama, or film. It serves as a tool for portraying characters, revealing personalities, and advancing the plot, and can also refer to a serious, collaborative exchange of ideas aimed at mutual understanding.
So, for a moment, imagine that we are characters in a movie. We all have backstories, belief systems, histories of failures and successes, biases (known or unknown), and more. Some characters in the story have human intelligence, while others have artificial intelligence. Our worldview is limited to the past, present, and future. Conversation takes place in scenes to move the plot. Every scene in a movie is important. As the plot progresses, personalities are revealed and help the characters grow.
reflected in the author’s sunglasses
What is not dialogue?
Speeches, downloads, mansplaining, lectures, content, information dumps, dashboards, Sharepoint sites…
Scene 1: Dinner with international speakers
Prior to the conference, some of the conference speakers and chairs met for an informal dinner. what did we eat? I don’t remember the food. But I remember the characters and the dialogue. Dialogue presupposes a common goal of mutual understanding. Mutual understanding does not imply complete agreement. Even if you completely disagree with someone’s opinion, you can still have a dialogue. But this can only happen if there is at least some mutual trust, respect and openness. Dialogue also includes listening. Active and open listening. I don’t wait for my turn to speak. Waiting for response.
We also touched on psychological safety, playfulness, food, travel, and of course learning-related topics. There were no slides, no task aids, no clicking “next.” Building connections through dialogue remains important in the age of AI.
Imagine two situations.
Your manager will send you a beautifully crafted note about your accomplishments on the project. Concise, concise, emotional, perfect grammar. However, it was clearly written by an AI. Your manager will send you a note about the same accomplishments. It’s not perfect, but it took some time and effort between two important meetings. There may be a typo.
Most people will automatically say they prefer authentic human messages and communications. But are we too? AI influencers with brand credibility will drive online traffic, chatbots are rated as more empathetic than human doctors, and customer service AI agents will replace long hold waits due to “unusually high call volume.”
I don’t have an answer, but I suspect that AI will dominate interactions if they are transactional, pragmatic, and don’t care about long-term relationships.
Scene 2: Reality vs. Hype
The current state of AI feels like the Land of Oz. Meanwhile, magical illusions rule LinkedIn, with experts in every field with rich frameworks. Every decent learning technology vendor now offers AI-driven capabilities, from content creation to simulation. L&D is still working on prompt engineering, but some leaders are moving to context engineering and the rest of the world is using OpenClaw to build their chief of staff.
Where can I find my results?
DX examined AI and engineering outcomes in a longitudinal study.
Many leaders feel that their organizations are falling behind in the race to achieve AI-driven engineering velocity. Vendor marketing and social media are expected to improve by 3x and even 10x. When leaders see more modest results, they think something is wrong.
To provide that complete picture, DX analyzed the velocity of engineering from November 2024 to February 2026 for a sample of more than 400 companies that have experienced exponential growth in AI adoption. We found that a 10-15% increase in PR throughput is a real benefit, but significantly less than most leaders expect.
The paper then delves into why the promise of improved performance from AI has so far been unmet. [1].
What about L&D?
There is currently a lot of research focused on the impact of AI on L&D. The findings from RedThread Research, Egle Vinauskaite, Markus Bernhardt and others provide guidance on what’s happening in L&D (and beyond) and how to take charge of the future.
As for my responsibilities, my session was very focused on rapid prototyping using AI tools. L&D has always had a problem with rapidly iterating designs to demonstrate a working model. Previously, this required technology expertise and often required IT assistance. Now, AI is accelerating the process, allowing learning experts to rapidly experiment, iterate, and learn through prototypes. I described this as a journey that requires a destination worth going to (a business problem or opportunity), a vehicle (an AI tool that meets the need through cost, speed, and control), and a map of how to get there (not a static map in the old sense, but more like a GPS route that shows you how to start the journey).
But if we let AI drive this process and we just passively participate, it will be an expensive journey to learn how quickly we can get to places we never intended to go.
The reality is that AI is not a technology that L&D should “adopt.” At least not from that angle. And that is by no means the starting point. For example, you’ll want to show how using AI to automate content creation can improve efficiency. My challenge to all L&D leaders is to move away from faster content creation and measure impact. It doesn’t start with AI. It starts with understanding how we work today and how we should work tomorrow.
How will things go today? What is the workflow? Who makes what decisions? Who is responsible for what output? How do you define the quality of a particular output? How do you check quality? What are the outcome expectations?
Asking questions may feel like it slows you down, but it can help speed up your journey while reducing the number of dead ends you encounter.
Scene 3: Why prototype and what to prototype?
A common mistake is to treat a prototype as a cheaper version of the real thing. These prototypes often get stuck in the prototype stage because they aren’t scalable and don’t really answer any questions (other than “Can I build it?”).
The prototype is for learning purposes only. To learn something quickly and iteratively. Your prototype should focus on the most important parts of the experience you want to simulate. If this is your first AI chatbot to assist your employees, there’s no need to build a full-fledged application and find out that the content it produces isn’t relevant to your audience. Playtesting with real business problems and real users is key.
What do you do when you find out your idea doesn’t work? Well, you’ve saved resources and time to build something that will make it happen. I’ve seen far too many application “adoption problems” within companies because the team hadn’t prototyped the core experience. “If you build it, they will come” is not a strategy.
What to prototype?
Start with a business problem or opportunity worth solving. Efficiency is easy to aim for, but it can be counterproductive. Once upon a time, I created an automation that took text and created a PowerPoint deck from that content in minutes. I thought I had saved hundreds of HeH (human equivalent hours). In a sense. This allowed us to build out ineffective voice-over presentations more quickly. Again, make sure you have a business case for the future, not just the current stage.
Then start with the end point in mind: who is your audience and how will they access your solution? Prototypes don’t have to be perfect, but you should consider scalability and keep the final delivery in mind when creating prototype versions.
Who is the target audience?
yourself
This provides a practical application to help check proficiency and quality. For example, if you’re responsible for checking the quality of your assessment questions, that’s an easy target for a skilled AI agent. If you haven’t built an AI agent yet, but want to improve the user experience of the eLearning courses you create, that can also be a realistic goal. your companion
What if you could solve a bottleneck in your team’s current workflow? What if you could build something that enhances that process or replaces some elements? For example, if you’re using xAPI, you can create a statement builder for your team that follows standards and generates ready-to-use code. If you are still playing around with SCORM, you can build the same. your organization
What if you could solve cross-departmental workflow bottlenecks? What if you could use utility tools to help others do their jobs easier and faster, or find relevant information faster? What if you could retire old, outdated training courses and replace them with interactive assistants that provide real-time support? Employees (“learners”)
What if we could embed dialogue within the learning experience? Or simulation tailored to role, location, and previous skill level? Sometimes we need to be “innovative” in the sense of being resourceful. Since you already have an LMS that authenticates users and stores data (via SCORM cmi statements), you can deploy relevant, customized, and practical utility tools with deep link launches. Of course, a dedicated web server with single sign-on is better, but in the meantime you can prototype your tool.
Speaking of access, I suggested in the session that everyone should start with a plan, no matter how small their first prototype is. Specifically, plan the entire solution (not just the prototype) in a Product Requirements Document (PRD). All LLMs know exactly what a PRD is and can build the foundation for you. You can then deploy this document as one of your project artifacts.
No matter what AI tools you use (I alternate between Windsurf, Claude Code/Coworker, and Github Copilot), this basic PRD will help you make decisions and scope your prototype precisely with the final solution in mind. All of the above is related to one thing: dialogue. Repeated meaningful conversations between humans and AI.
Now let’s go make something!
PS If you’re wondering what this photo represents (beyond the reflection in the sunglasses), you need to look at the Banksy sculpture in the background. It was originally supposed to be about blind patriotism, with a man blinded by the flag stepping into free fall. For me, it brings parallels to AI. Learn and experiment responsibly. Don’t just blindly follow influencers.
Image credits: Photos in the article were provided by the author. References:
[1] AI and Engineering Velocity: A Longitudinal Analysis
Share with
