The good, the bad, and the downright ugly of AI within AAC systems.
About seven weeks ago, I was out walking Winter around the vineyard thing near my apartment on a Thursday afternoon.
She’d just been decommissioned from actual walks, so we were both adjusting to whatever this new version of normal was supposed to be.
I was bored.
Then my phone lit up.
A Messenger notification from Mum. Three letters.
WTF.
This’ll be good, I thought.
It was.
In the worst possible way.
I opened the message and there it was. A perfect image, in the exact style my workplace uses for social media. Clean. Familiar. Corporate.
And then the words.
AI for iOS apps.
I work for an AAC company. One of the big ones.
Before you get your knickers in a knot, I’m very pro-AI. I have to be.
The gap between what I can produce and what most people can is only going to get bigger as this technology improves and spreads. I want it everywhere.
Just not in my AAC system.
Not yet.
I always knew this conversation was coming.
I just thought I’d have more time.
I don’t.
We’re already there.
And if my language is going to bother you, this is your warning.
I’ve thought about this too much, for too long, to clean it up now.
Grandma, you can close this email.
It gets worse from here.
The problem isn’t AI.
It’s what happens when you put it inside the one place I should always have control.
My AAC system.
My voice.
How I communicate, work, write, and access the world.
Because once you start mixing AI with AAC, this stops being a fun little innovation conversation and starts becoming an ethics one.
A big one.
AAC users have spent decades fighting to prove that our words are our own.
So if we’re still arguing about the legitimacy of Facilitated Communication forty-something years later, why in the fuck are we so eager to shove AI into the middle of this before most people even understand what it is?
People already question whether our communication belongs to us.
This just opens the door wider before we’re ready to deal with what walks through it.
And that’s the part that should make people uncomfortable.
Because I’ve been using AI constantly for the last two years.
I know what it can do.
And I know what it can’t.
It still can’t write like me without a lot of back and forth. A fucking lot.
Which means one of two things happens.
Either I spend even more time correcting it than I would just writing it myself.
Or I don’t.
And if I don’t, those aren’t my words anymore.
Yes, I used AI while writing this.
The ideas are mine. The decisions are mine. The words are still mine.
That line matters.
And that’s the clean version of this conversation.
The part I’m still trying to figure out is what this actually means for me.
That’s the part people don’t really talk about.
The moment where something sounds like you, looks like you, is close enough to pass, but isn’t actually yours.
I notice it immediately.
It’s subtle.
A sentence comes out faster than it should.
Too smooth.
Too clean.
For a second, it feels incredible. Like finally, something is working the way it’s supposed to.
Like I’ve been given access to the version of communication everyone else gets by default.
And then I read it back.
And I know.
I wouldn’t have said it like that.
Not the words. Not the structure. Not the tone.
It’s close.
Close enough that most people wouldn’t question it.
But I do.
Because I built every sentence I’ve ever said.
Word by word. Decision by decision.
That’s not a limitation.
That’s authorship.
People talk about AAC like it’s a workaround.
Something to avoid whenever possible.
Like it’s a slower version of something everyone else already has.
It’s not.
It’s a completely different relationship with language.
Every sentence costs something.
Time.
Energy.
Focus.
Choice.
You don’t just “say something.”
You decide it.
You commit to it.
You watch it appear in front of you before anyone else hears it.
So when something skips that process, it doesn’t just feel faster.
It feels wrong.
People assume speed is the goal.
It’s not.
Speed is convenient.
Control is everything.
Slower is how I know it’s mine.
Slower is where I decide what I actually mean.
Slower is where I catch myself before I say something I don’t want to.
Or decide to say it anyway, because fuck you.
If you remove that, you don’t just remove friction.
You remove authorship.
And that’s the part that makes AI in AAC different from everywhere else.
Because in most places, AI is helping you get to your words faster.
Here, it can replace the process that makes them yours in the first place.
And that is a very thin line.
One that’s easy to cross without realising.
Because it won’t feel like replacement at first.
It’ll feel like help.
It’ll feel like relief.
It’ll feel like finally catching up.
Until one day you realise you’re not correcting it anymore.
You’re accepting it.
And once you start doing that consistently, something shifts.
Not all at once.
Slowly.
Quietly.
Your sentences get easier.
Your decisions get fewer.
Your voice gets… smoother.
More predictable.
More acceptable.
Less you.
And the worst part is, it’ll still sound like you.
Good enough that no one questions it.
Good enough that you almost don’t either.
But that gap is still there.
Between what you would have said…
and what you let it say.
That’s the part I don’t think people are ready for.
Because this isn’t just a tech problem.
It’s an identity one.
AAC has always been about proving that we have a voice.
That we are the source of our own words.
That what comes out of these systems belongs to us.
So when we introduce something that can influence, generate, or reshape those words, we’re not just improving a tool.
We’re changing the relationship entirely.
And if we don’t handle that carefully, we’re going to end up back in a place we’ve already spent decades trying to get out of.
Where people question whether our words are really ours.
Where our authorship is up for debate.
Where someone else, or something else, is seen as part of the source.
That is not a hypothetical risk.
We’ve seen it before.
And we’re a lot closer to it than people want to admit.
I’m not saying don’t build it.
I’m saying don’t rush it.
Don’t prioritise speed over control.
Don’t solve one problem by creating a bigger one.
And don’t assume that what works everywhere else will work here.
Because this is different.
And if we get this wrong, we don’t just lose time.
We lose something a lot harder to get back.
Our voice.
There’s a version of this where I take the shortcut.
Where I let it finish the sentence.
Where I accept the version that’s close enough.
Because it’s easier.
Because I’m tired.
Because sometimes I don’t want to spend ten minutes building something I could get in ten seconds.
And that’s the part no one really wants to admit.
It’s not just about control.
It’s about temptation.
Because the trade isn’t obvious at first.
It feels like relief.
It feels like finally not having to fight for every word.
It feels like being able to keep up.
And I want that.
I really do.
But I also know what I’d be giving up.
Not all at once.
Not dramatically.
Just a little, every time I don’t push back.
Every time I don’t fix it.
Every time I let “close enough” be enough.
Until one day I’m not choosing my words anymore.
I’m approving them.
So the question isn’t whether AI belongs in AAC.
It’s what it looks like when it does.
Because it will.
The cat is out of the bag.
If this is going to exist in AAC, I still need to be the source.
It can suggest.
It can support.
But it cannot decide.
If I’m not choosing the words, it’s not my voice.
Not everything that feels slow is broken.
Some of it is doing exactly what it’s supposed to do.
Slower is where I decide what I actually mean.
Slower is where my voice exists.
If you remove that completely, you don’t just make things easier.
You erase the process that makes the words mine.
And people need to be able to trust that what comes out of these systems is actually ours.
Not because we owe anyone proof.
But because we’ve spent decades fighting to be believed in the first place.
We don’t get to go backwards on that.
And none of this works if the people building it aren’t the people living it.
Because this isn’t theoretical.
This is daily.
I’m not against AI in AAC.
I’m against losing control of my own voice.
And if those two things ever become the same thing, we’ve got a bigger problem than people realise.