AI and the Last 10 Percent

Technology’s Human Problem

Computers, innovation, and technology have been amazing and encouraging extensions of our humanity, our social interactions, and even lifesaving tools.

But technology hasn’t always lived up to its promise when it matters most to users.

I’m on my iPhone. I switch to my Plantronics bluetooth. As I leave the house, my iPhone inevitably disconnects from both the WiFi and my bluetooth headset. Almost every time. I have to say sorry to the person I was talking to, as it beeps while I attempt to reconnect to the headset.

Or as I’m driving and interacting with CarPlay, I want to play a song that isn’t quite well known by Siri. I try to have Siri play it, and it inevitably plays the wrong song or glitches out completely and doesn’t play anything.

In the tech world, this is considered technical debt, feature completeness (bloat), or even feature incompleteness. Just like how big new road projects get funding while everyday streets fall apart, flashy new features are prioritized over fixing existing bugs and gaps. We inevitably decide to take a different route and end up with a feature with a lot of unmet potential because businesses have focused on the next new shiny feature.

Technical debt is the same as old pavement. It’s rarely prioritized, but heavily used… Until users find another route to get to the same place.

The small issues that usually don’t impact the users workflow, but is annoying enough to make them huff in frustration slowly losing the users trust.

AI’s Business Expectation

We’ve heard loud and clear what businesses plan to do with the AI they are working on. The ideas technology growth businesses like Google, Amazon, and others have, go hand in hand with the “productivity gains” they are trying to achieve with their current employees.

Andy Jassy said: “As we roll out more Generative AI and agents, it should change the way our work is done… We will need fewer people doing some of the jobs that are being done today, and more people doing other types of jobs.”

Or from Barbara Peng, the Business Insider CEO: “Business Insider will lay off one in five employees and go ‘all‑in on AI.’”

These quotes are everywhere. These leaders aren’t wrong in pursuing efficiency, but many times these companies are acting on future gains before those gains have been fully realized.

Businesses are expecting AI to drive productivity and in turn profitability. But in AI’s current state, we shouldn’t use that as the main selling point.

Look at Uber's self-driving ambitions. They viewed autonomous vehicles as "existential" to their future, promising higher profits and better service. Despite major investments and bold predictions, technical challenges proved insurmountable. After a fatal accident in 2018, Uber sold its self-driving unit in 2020, abandoning this vision entirely.

You can see that already happening with Klarna, who in 2022, let go about 700 of their customer support agents to find that they ended up hiring back a good portion of their support agents because they realized it wasn’t the cost saving they were expecting.

Additionally, consider the sheer volume of code AI is generating. It's staggering. I've watched as senior developers share their experience with AI and its wildly varying levels of effectiveness. Even if a quarter of that code makes it in real world scenarios, the amount of additional technical debt will be insane. Worse off, because its generated, it’ll be hard to do a “find and replace” since it’ll all vary in outcomes based on its prompts.

The REAL Solution

There’s a solution for both AI and the employees AI is supposed to replace.

We shouldn’t be firing developers, because it solved the “quick cost cut win”.

We shouldn’t be using AI as a platform to bet on future potential gains.

We shouldn’t be using AI as your dearly developed “vibe coder” code.

We should, at least in part, use AI to solve “the last 10%” of technical debt, feature promises, and general code uniformity that we can’t afford to have developers focus on, but is just as critical to many users to fix or improve.

Lets keep those developers designing the highways and streets we need for an innovative future, and let AI re-pave the ever-increasing patchwork of technical debt, and feature incompleteness in a way that could never efficiently and efficiently be completed by humans.

It’s time to stop chasing the next big thing and start using AI to fix the ones we already promised.