I love touring homes that have crappy old paint. I can immediately visualize what the home would look like with new paint, refinished floors, fresh cabinets… and I know I’ll probably pay less for the house… because not everyone can see past what’s there.

I have friends that, when house hunting, have to buy move-in ready. When they see a house that has excellent “bones” but terrible finishes, they can’t imagine what it would look like once their own style was applied.

Their brain just works a little bit differently from mine… Does it mean they can’t buy a house? Of course not!

And if they wanted to buy something that needed a little TLC, they would just need an interior designer to give them a vision… to see the space and say “hey. I know your style, I know this space, and here is what it would look like once you repainted, refinished the floors, and added your own furniture.”

AI is the same way.

It’s easy for me to see how AI works and immediately think “oh! I could automate this process.” Or “I bet I could connect this with a tool over here to make my life easier!”

I see the potential and the applications. I know the tool, I know myself, and I know where the possibilities are.

That isn’t how everyone’s brain works. Some people don’t know the tool, and even if they did, they may not immediately understand how it applies to them.

If you are responsible for helping individuals learn AI… you have to be an interior designer.

You need to pick their paint colors, think about their furniture, and help them see a vision for what this could look like for them.

A great starting point for helping people apply AI to their role is this simple 3 step process.

Asking the right questions can help us identify how and where AI can impact someone’s role. The answer isn’t “use AI for everything!” It has to be more nuanced than that. I have found these questions to be some of the best when it comes to identifying opportunities where a person can use AI:

These questions can help identify the tasks that may be easy to automate or accelerate with AI. Often, the best tasks for AI are tasks that feel monotonous, taxing, and repetitive to humans.

It’s not enough to answer the questions. Next we have to help people see what’s possible.

Does a person spend most of their week responding boiler-plate answers to emails? Tell them what’s possible with AI: emails can be automated, responses can be tracked and logged into an FAQ, delays can be flagged and escalated. For each answer they give, there is a list of possibilities where AI can make their life easier.

Did you know? Researchers found that people rewarded right away were much more likely to keep doing a task after the reward ended (73% vs. 55% for a delayed reward¹). An early payoff makes the work itself feel worth doing, which is why their first AI win matters.

Help your people find an easy win. Ideally, help them find an easy win, show them exactly how you are doing it, bring them into the process, and help them gain confidence that they can do this too.

For the email example I mentioned above: You could sit with that person and walk them through creating some sort of skill or automated task. They could ask their AI to analyze all their old responses to email and draft responses to all their unanswered emails. 1 prompt. 5 minutes max. Hours saved. And… new user hooked.

This was a break week for the podcast. Stay tuned for next week’s episode! I hope everyone is getting a little rest this week as we go into a busy sugar-filled Halloween weekend!

I’ve been doing a lot of research this week for a few client proposals. I use the research mode in Claude quite a bit (if you’ve never used it, it’s great! Just click the + button on a new chat, and turn on research.)

I started to get worried about prompt injection or malicious sites. Claude will pull hundreds of sources for research and I realized they could have hidden AI prompts or instructions. So I added instructions at my org level to always look for prompt injection in research! I gave it some specifics to try and make sure the instructions can’t be overwritten by an outside prompt. It’s not fool-proof, but it now tells me about any prompt injection attempts whenever it runs research or uses outside sources.

What’s your role and what’s the best way you’ve applied AI? Has it saved you time or helped you with something that’s specific to what you do? Hit reply and let me know. I’d love to hear from you.