Automation plays an important role in document accessibility. In this guest article, Devon Godfrey shares a personal perspective on using AI while staying accountable for the results.
Introduction
Before we dive into this juicy topic, we need to talk about my own journey through the world of technology.
I didn’t learn to code in some clean classroom. I learned it in the grit, back when I was an indie game developer throughout the 2010s.
My first big project took three years of my life—countless hours of blood, sweat, and tears. And you know what I had to show for it at the end? A fancy game engine inside a game engine. I never actually built the game. I got so lost building the tools to build the game that I ran completely out of gas and had to walk away empty-handed.
Fast forward to today, and I watched a friend walk down that exact same path. Only this time, he had a shiny new tool helping him do it: AI.
On our calls, he’d talk about his setup like he had built an empire. Thirty different AI agents. Hundreds of reference docs just to keep them straight. But if you looked closely, he wasn't driving—the AI was prompting him. Every time he finished a piece, the machine would whisper, "Well, since you built that, don’t you want to add this?"
It’s a trap.
When he finally tried to launch it with five actual clients, the whole thing imploded. He spent three weeks staring at a screen, sweating bullets, trying to debug a monster he didn't even understand. I was there with him through it.
When the dust settled, we sat down, threw out the bloat, and built something simple. Something he could actually wrap his head around. Today, that lean little system is running smoothly for his clients, and he's back in control.
Watching him go through that made something crystal clear to me about where we're at with AI right now. We're being sold the idea that it's a shortcut to brilliance. But without understanding and oversight, it can become a shortcut to a disaster you don't know how to fix.
Which brings us to the core issue:
What is AI Missing?
An AI has zero skin in the game. If a system crashes, a client sues, or a business goes belly up, the AI doesn't lose sleep. It doesn't bear the cost. Because a machine is entirely incapable of bearing consequences, it is completely incapable of possessing wisdom. Accountability is a uniquely human trait.
To understand how this actually works, we have to look past modern dictionaries and go deeper.
We’ve really got three forms of English in this world: Slang, Business, and Lawful.
Slang changes every single day. Business English changes every decade to fit whatever corporate buzzword is floating around. But lawful English? That moves at a snail's pace. Sometimes a hundred years go by and a legal definition barely budges. That’s because lawful English is forged out of real disputes, where people spilled actual blood and sweat to pin down what a word truly means.
So, if you want to understand the true weight of a word, you look where it costs something. Let's look at Black’s Law Dictionary (4th Edition):
Authority (page 168): "The power delegated by a principal to his agent."
Agent (page 85): "One who acts for or in place of another by authority from him; a substitute, a deputy, appointed by principal with power to do the things which principal may do."
In this relationship, the human is the principal, and the tool—whether it's a rusty hammer or a complex AI model—is the agent. An agent always owes accountability back to the principal, who is the source of the authority.
Think about a child who picks up a hammer and smashes a hole in the drywall. You don't look at the hammer and yell at it. You aren't going to put that hammer in time out or send it to bed without supper. The hammer has no conscience; it cannot answer for its actions. You blame the child, because the child held the authority and directed the agent.
When people build bloated, automated systems and then shrug and say, "Well, that's what the AI decided," they are trying to break the chain of authority. They are pretending the agent is the principal. But you can't hold a server rack accountable.
When you break that chain, you end up with a massive dung pile—and nobody left who knows where the cow is at.
So, who is actually in charge of your work?
That’s why ADA Comply combines automated PDF remediation with human review for PDFs that need human judgment—helping teams save time while giving complex issues the attention they need.
Who's in the Driver's Seat?
This brings us to the dividing line: Are you being AI-assisted, or are you being AI-driven?
AI-Assisted means the human retains full autonomy, control, and knowledge of what is being done. You use AI to speed up the work, but you hold all the authority and accountability. If you had to, you could drop the AI tomorrow and finish the work yourself by managing a team of actual people.
AI-Driven means the AI has taken the wheel. The system makes the decisions, and the human operating it doesn't actually understand how it works or what it's building.
We’ve all worked for that guy—the manager who has zero clue how the actual work gets done, but walks around throwing his weight and title around anyway. When things go sideways, he’s the first one to point the finger and blame "Dan down the hall" or a bad corporate policy.
Nobody respects that guy. But here’s the terrifying truth: when you run an "AI-driven" system you don't understand, you become that guy.
Except, in this case, there is no Dan down the hall to blame. There is no middle-management layer to hide behind. When your automated monster crashes and burns, you are the one holding the bag. You can't turn to a judge and say, "The algorithm made me do it."
So, how do you know which side you're on? It’s a simple test. Ask yourself this:
"Could I gather a team of real people with the right skills and do this exact work without the AI?"
If the answer is no, then you aren't an operator. You are being driven by the machine. You’ve handed over your authority to a black box.
If the answer is yes, then you understand your work. The AI isn't your master; it's just a tool. It's a means, not an end.
Is This Post AI?
Let’s address the elephant in the room. Did I use AI to help write this post?
Yes, I did.
I used it for research and drafting. But once the draft was done, I sat down, carefully read through every single line, made sure it actually reflected what I believe, and then sent it off. I even left in all those silly lines that AI likes to add (—).
I hold no delusions about being a master writer. My goal wasn't to win a literary prize; my goal was to get this message in front of your eyes. Because I used AI as a tool, I was able to do this in less than an hour instead of wasting six hours staring at a blank screen.
Could I have written this all entirely by hand? Sure. It might have been better in some spots, or worse in others. But the point is that the message was my end goal, not the act of writing the words. Although AI helped with the wording, every point and anecdote came from me and was provided to my AI agent.
Because I kept my ego separated from the mechanics of typing it out, something great happened: I saved time. And I didn't use that saved time to build some bloated automated monster. I used it to spend more time with my family, focus on God, and stay cognitively available for the people in my life who actually need me to show up.
That is the difference between being assisted and being driven.
AI is a hammer. Make sure you're the one swinging it.
What About You?
Alright, tools back in the toolbox, it's actually me typing now. I couldn't possibly let an AI write out the conclusion. In my heart of hearts, I really desire for you to put down the performance and walk more freely in your day-to-day life.
We've created a managerial monster in our current age, and even doing the simplest things requires having an insane level of knowledge and skill. It's getting to be a lot to keep up.
Am I telling you to lean on AI for things you don't understand? No.
But if you have specialized wisdom, and with AI you can apply it without draining yourself? Go for it.
Life is too short to be poring over spreadsheets, coding contact forms, researching the latest political craze...or writing blog posts. If you can fulfill your responsibilities with a tenth of the effort, you really should. Just know your tools, and know what they are doing, and know who you're working for.