Rethinking Apps: From Features to Outcomes. Part 1

AI’s Impact on UI

Last night, my brother and I explored a thought experiment about how we interface with apps today and how that might evolve. While the future may unfold differently, it was a fascinating exercise to consider the possibilities.

I don’t think this gets enough attention. We talk about how AI will change the way we interface with a computer, but rarely stop to look at how we currently interface with the apps we use.

Currently we interface with nested menus, buttons, lists, popup menus and more, in which many users get lost frequently. UI designers have to balance the use of many casual users with the use of power users, essentially frustrating both when the app matures past its initial offering.

UI designers have an impossible job of balancing and surfacing advanced features while also making those same features feature rich enough for those power users.

What if we flip the script. AI is great(ish) and getting better at understanding what a user is asking, and if it’s not, it can ask clarifying questions.

If we were to use this methodology we can:

1. Reduce friction for new users by surfacing just-in-time features.

2. Customize interfaces per user instead of designing for the lowest common denominator.

3. Price features by actual value delivered—not just seats sold.

Yes, talking to AI is slower than clicking buttons. However, discovering features in complex app often requires extensive Google searches or ChatGPT conversations. Users become "power users" only after spending hours of company time researching how to use app features, focusing on the tools rather than the actual outcomes they need to achieve.

For those apps that allow you to customize your own UI, yes, that is a good thing but I’m sure that is a feature only power users find helpful to make them that 3% more efficient than their less skilled users.

AI Use Example

Here’s how I’m thinking it would work.

Lets talk about a basic PDF document.

I’m using a PDF editor app. That app has a basic UI and feature list on its first screen.

I’ll open a PDF and my goal is to sign it. I can either tell the app “I need to sign this PDF”, and a signature box can show up and either draw or import my signature, close the app and be done, or tell it I need to fill out the PDF based on context (I know Apple was touting this feature in their current unreleased AI feature).

This is simple but effective. Getting shit done quickly, without any questions on “where’s the signature button”.

Here’s a more complex use case.

I’m trying to decipher a complex contract with charts, and requires specific knowledge of other documents that may not be in a form that PDF app has access to.

I want to highlight some text, make changes to one of the charts based on information in other formats or locations, and make a new page a fillable PDF, based on specific design requirements.

Instead of requiring the PDF reader to be connected to some complex contextless LLM (for now), perhaps we focus on giving me the ability to surface features that can get results rather than trying to understand the way the developer expected the features to be used based on complex UI studies.

Now, some of those tasks are too complex for current AI to do easily and correctly. So instead, lets just have the AI do what it can do well and surface the features for those it would struggle with, and put those features in the ribbon for easy access.

I can tell the AI highlight any verbiage around warranty and payment terms. Done.

I can tell the AI to surface the feature(s) required to create a fillable page with images, tables, and text.

The AI will now surface the text tool, the form tool, the table tool, the image tool and put them in a ribbon allowing you to create the page the way you want to create it.

Now the tools working for me rather than the other way around.

Part 1: Final Thoughts

Apps should work for the user, not the other way around. Democratizing and simplifying tools to the outcome rather than the skills and limitations of the tool will allow businesses to get shit done, rather than hiring experts that understand little of the expected outcome but understand the nuances of the tool, or the other way around.

AI innovation and LLMs will help us get that last 9% for ALL users rather than getting us 90% for most users.

Now for Part 2, eventually you will pay for the code that gets something done, rather than the app that locks you into something that may get you 90% of the way there. How will that work.