top of page

Bridging the gap to AI product adoption with scaffolding

Updated: Jun 29




Remember when smartphones first came out? Designers had to help people figure out how to use them. They used metaphors – creating camera icons that looked like real cameras with shading and curves. That’s the same reason you’ve got folders and a trash can on your laptop. Over time, as we got used to it, the icons got simpler and flatter.

A recent survey found that 37% of people use chatbots for work daily. Professor Ethan Mollick said we should “use AI for everything we legally and ethically can because that’s how you get experience with how these systems work.” I agree with him. But not everyone knows where to start, what AI can do, or how to make the most of the results.


One of the ways some of the big players are doing this is to focus on magic. Ever since I heard Todd Olson, CEO of Pendo, mention this on a podcast I see this approach everywhere. While magic might be a good marketing message, I’m not sure it will help skeptical users try these tools if how they work is mystical instead of being engineered. I also loved Fantasia and know how that whole pretending to be a wizard thing went sideways.


It’s on us designers and product folks to support users through this transition in a way that we can remove over time as they get more comfortable.


Borrowing the concept of scaffolding from educators

I’ve written before about scaffolding, a concept from education. Scaffolding gives learners support early on until they can do things independently. This idea is super relevant for product and design teams using AI. I’ve taught skiing for years, and these are the techniques to help students tackle tougher terrain while building skills without me always being there:


  1. Breaking learning into smaller steps

  2. Gradually giving less support

  3. Building on what they already know

  4. Personalized approach

  5. Showing demos

  6. Regular feedback

  7. Thinking out loud so the learner understands the thought process


Ways to build scaffolding

Over time, as Mollick said, AI will get good enough to guide us without too much hand-holding. But for now, we need to build scaffolding to support users. Here are key areas where scaffolding can help the most:


Helping users get started

  • Make a welcoming interface that encourages exploring. Microsoft is doing this with the idea of a copilot that is there to help and guide you. The concept of an autopilot has been around for a while, so people may be able to relate to it.

  • Provide prompts and examples to show what the AI can do. Most tools are offering prompts to get started.

  • Borrow from video game designers by introducing new and more advanced features as users explore. I haven’t seen this yet, but after you do a few tasks, you could start getting suggestions on more advanced techniques.


Helping users judge the results

  • Guide them on how much to trust AI-generated results. The disclaimers about how AI can sometimes be wrong are pretty lame.

  • Clarify where information comes from and how reliable it is. Understanding the sourcing and confidence would help a lot of people who are naturally more skeptical to lean in. It also supports your education and next steps. I read a couple of the cited works to get a better understanding of a topic.

  • Point out what kind of decisions this tool is good/bad for. We are not seeing this yet, but more purposeful models could help with this. I could imagine a personal finance tool being clear that it’s optimized for being low-risk and reducing debt instead of providing the most optimized advice.


Supporting next steps

  • Smoothly transition into users’ existing workflows. What is the next step? Can I suggest it? Can I do part of it for you? Tools like Perplexity are great at suggesting follow-up questions to keep exploring.


Guiding and coaching
  • One of my best mentors in teaching liked to always say, “You might try…” It was a way to introduce new ideas and give feedback without making someone feel bad.

  • Encourage multiple rounds to refine understanding. After they provide the first prompt, show them different ways to modify the results.

  • Constantly offer tips to get the most out of the AI. You could also give some more information about how the reasoning works to provide more insight into how things work.


Guide the user on their way

When designing AI tools, our goal should be creating scaffolding that supports users as they learn. With thoughtful design, we can build their confidence and smooth the way for AI to become part of their normal routine. Let’s innovate and guide, making sure our designs meet today’s needs but also set the stage for what’s coming.


Image: A sketch with acrylic paint markers

bottom of page