AI and Neurodivergence: When Technology Removes the Barrier
AI and neurodivergence can be a powerful combination. Explore how AI can support people with dyslexia, ADHD and autism, plus the potential risks.
Most conversations about artificial intelligence eventually come back to productivity.
How much time can it save?
How many emails can it write?
How many jobs can it automate?
How much more can one person get done?
But AI and neurodivergence raise another part of the conversation that deserves far more attention.
Accessibility.
For someone with dyslexia, ADHD, autism or another additional learning need, AI may not simply make a task quicker. It might make a task that was previously difficult considerably more manageable.
It might allow someone to approach information differently.
It might remove the need to ask another person for help.
And in some situations, it can give someone greater independence over something they were already perfectly capable of understanding, but found difficult to express or organise in the format expected of them.
That distinction matters.
AI isn't removing neurodivergence. It isn't a treatment, and it certainly isn't universally helpful.
But it can change the interface between a person and a world that isn't always designed around the way they process information.
That could prove to be one of the most important benefits of AI.
Contents
- The Real Benefit Might Be Agency
- AI and Dyslexia
- AI and ADHD
- AI and Autism
- Asking AI Doesn't Carry the Same Social Cost
- AI Isn't Automatically Accessible
- There's Also a Risk of Dependency
- AI Gets Things Wrong
- Privacy Matters Too
- Is Using AI Cheating?
- Better Output Doesn't Mean the Difficulty Has Disappeared
- Using AI Doesn't Make the Achievement Less Theirs Either
- What This Could Mean for Small Businesses
- The Most Exciting Part of AI Might Not Be Automation
- AI Should Give People More Control, Not Less
- Further Reading
The Real Benefit Might Be Agency
Imagine someone with dyslexia has written an important email.
They know exactly what they want to say.
The ideas are theirs.
The decision is theirs.
But spelling, sentence structure or getting those thoughts onto the page creates an additional barrier.
Previously, they might have asked a colleague, friend or family member:
"Can you check this before I send it?"
Now they can ask AI:
"Can you check the spelling and make this clearer without changing what I'm saying?"
The obvious benefit is a better email.
The more interesting benefit is independence.
They didn't need another person to approve their communication before they could send it.
That's agency.
And this distinction runs through much of what makes AI interesting for neurodivergent people.
AI and Dyslexia
Technology was supporting people with dyslexia long before ChatGPT arrived.
Spellcheckers, text-to-speech, audiobooks, predictive text and speech recognition are all examples of technology reducing barriers around written communication.
Generative AI takes that further.
Someone can now give an AI tool a dense piece of text and ask:
"Explain this in plain English."
Or:
"Break this into five key points."
Or:
"I've written this quickly. Correct the spelling but keep my wording and tone."
Or even:
"Ask me questions about this document so I can check I've understood it."
That's fundamentally different from simply asking a machine to complete a piece of work.
The information is being transformed into a format that may work better for the individual.
Research into AI-driven assistive technologies is already looking at the potential applications across dyslexia, ADHD and autism, although the evidence base is still developing.
But There's a Risk Here Too
If AI automatically rewrites everything, where does assistance end and substitution begin?
Someone asking AI to correct spelling while preserving their meaning is very different from asking AI to decide what they should say.
There is also the problem of accuracy.
AI can simplify something incorrectly.
It can remove an important qualification.
It can misunderstand the original text.
And because AI tends to present its answers confidently, an incorrect explanation can look every bit as convincing as a correct one.
That's particularly important if someone is using AI precisely because the original material is difficult for them to check independently.
The tool can remove one barrier while potentially introducing another.
AI and ADHD
For ADHD, some of AI's most useful capabilities have very little to do with generating content.
They're about creating structure.
A large, undefined task can become:
- Do this first.
- Then do this.
- Ignore this until later.
A collection of thoughts can become an organised plan.
A long email can become three actions.
A project that feels too big to begin can be broken into something that can be started in ten minutes.
AI can effectively become an external place to organise information.
Research is beginning to explore how AI systems might provide this kind of cognitive scaffolding around planning, task awareness and self-regulation.
And in business, the same principle already has practical applications. AI can help turn unstructured information into something usable, whether that's handling repetitive enquiries, drafting responses or organising information as part of an automated workflow.
But ADHD also demonstrates the other side of AI particularly well.
AI Can Become the Ultimate Rabbit Hole
AI has almost no natural stopping point.
Ask one question and there are ten directions you could take next.
An idea can become a plan.
The plan can become a project.
The project can become a system.
The system can become an app.
And suddenly the original task has disappeared completely.
AI provides novelty, immediate responses and virtually unlimited possibilities.
So the same tool that can help someone focus on what they need to do can also provide an endless supply of more interesting things to do instead.
It can even make procrastination feel productive.
Spending three hours designing an elaborate system for managing emails certainly looks more productive than scrolling through social media.
But if the job was simply to reply to five emails, the outcome isn't necessarily much better.
The technology can support executive function.
It can also overwhelm it.
Both things can be true.
AI and Autism
This is an area where sweeping statements are particularly unhelpful.
Autistic people aren't one group who all think, communicate or experience the world in the same way.
AI will therefore be useful to some people and completely unhelpful to others.
But there are some interesting possibilities.
One is interpreting ambiguous communication.
Consider an email that says:
"It would probably be good if we could get this sorted fairly soon."
What does "fairly soon" actually mean?
Is it urgent?
Is the sender annoyed?
Is it a suggestion or an instruction?
AI can be asked to explain possible interpretations and identify where the language is ambiguous.
It can also work in the opposite direction.
Someone can write exactly what they mean and ask AI whether any part could be interpreted differently.
That doesn't mean AI has magically solved human communication.
It hasn't.
AI doesn't actually know what the person who wrote the email was thinking, and it shouldn't be treated as though it does.
But it can offer possible interpretations and provide another way of examining the communication.
Preparing for Unfamiliar Situations
AI can also be used to reduce uncertainty.
Someone could ask:
"I've never attended a business networking event before. Talk me through what normally happens from arriving to leaving."
They can then ask as many follow-up questions as they like.
What should I wear?
When do people normally introduce themselves?
What happens if I don't know anyone?
How long should I stay?
What do I say when someone asks what I do?
None of those questions require particularly advanced artificial intelligence.
But being able to explore an unfamiliar situation privately and at your own pace can make it more predictable.
Again, the benefit isn't necessarily that AI knows something nobody else does.
It's that the person can access the information in the way and at the time that suits them.
Asking AI Doesn't Carry the Same Social Cost
This might be one of AI's most overlooked accessibility benefits.
People can ask it the same question repeatedly.
They can say:
"I still don't understand."
Then:
"Explain it differently."
Then:
"Give me an example."
Then:
"Make that example relevant to my job."
Then:
"Now explain it in three sentences."
AI doesn't become impatient.
It doesn't make someone feel embarrassed for asking again.
It doesn't say:
"We've already been through this."
That can matter.
It creates a private space where information can be reshaped repeatedly until it becomes useful.
And unlike traditional software, generative AI can often adapt the explanation itself rather than simply presenting the same information again.
AI Isn't Automatically Accessible
This is where some of the excitement around AI needs tempering.
Just because a technology can improve accessibility doesn't mean every AI product has been designed accessibly.
An AI that produces enormous walls of text may make things worse for someone who is already struggling with information overload.
An AI assistant that gives 30 suggestions when three would have been useful can create more cognitive work rather than less.
A voice interface that regularly misunderstands its user isn't accessible simply because it has a microphone button.
A chatbot that responds to a straightforward question with unnecessarily complicated language hasn't removed much of a barrier.
Accessibility isn't achieved simply by putting AI into something.
The system still needs to work for the person using it.
AI has accessibility potential. That doesn't automatically make AI accessible.
There's Also a Risk of Dependency
If AI makes a difficult task easier, using it can make perfect sense.
We already accept assistive technologies throughout everyday life.
Spellcheck.
Calculators.
Predictive text.
Text-to-speech.
Speech-to-text.
Screen readers.
Navigation systems.
Technology has been removing unnecessary barriers for decades.
But generative AI is slightly different because it can potentially do much more of the thinking itself.
There is a difference between:
"Help me structure my thoughts."
and:
"Tell me what my thoughts should be."
There is a difference between:
"Correct my spelling."
and:
"Write my argument."
There is a difference between:
"Help me understand these options."
and:
"Make the decision for me."
Where that line sits will be different for different people and different tasks.
But it's worth thinking about.
The goal of assistive technology should ideally be greater agency, not greater dependency.
AI Gets Things Wrong
This point cannot be ignored.
Generative AI can produce incorrect information.
It can misunderstand context.
It can invent details.
It can confidently explain something that simply isn't true.
That doesn't make AI useless.
It means understanding its limitations is part of learning how to use it.
If someone asks AI to restructure an email they've already written, the consequences of a mistake may be relatively minor and easy to spot.
If they're asking it to interpret medical, financial, legal or similarly important information, the stakes are completely different.
AI literacy therefore becomes part of accessibility.
It's not enough to give someone access to the tool.
People also need to understand what it can do, what it can't do and when its answers need checking elsewhere.
Privacy Matters Too
There is another issue that deserves more attention.
People can end up telling AI an enormous amount about themselves.
Sometimes precisely because talking to a machine feels easier.
They might paste in an uncomfortable email.
Describe a difficult workplace interaction.
Explain something they're struggling with.
Upload documents.
Discuss their health.
Ask questions they don't feel comfortable asking another person.
That can make AI feel liberating.
But convenience shouldn't make privacy disappear from the conversation.
People need to think carefully about what information they're putting into AI systems, particularly when it involves employers, customers, medical information or other people.
Is Using AI Cheating?
This is going to become an increasingly difficult question in education and employment.
If someone with dyslexia uses AI to correct spelling, have they cheated?
If someone with ADHD uses it to break a large task into manageable stages, is that unfair?
If an autistic employee uses AI to help examine an ambiguous email, have they somehow avoided doing their own work?
Those uses are difficult to equate with simply asking AI to complete the entire task.
Move the line slightly, however, and things become less obvious.
What if AI writes the entire assignment?
What if it completes the analysis?
What if it makes the decisions?
We've been through versions of this conversation before.
Spellcheck.
Calculators.
Predictive text.
Text-to-speech.
Speech-to-text.
Screen readers.
Technology frequently starts by being viewed as an advantage before eventually becoming accepted as an accommodation or simply a normal tool.
Generative AI complicates that discussion because it is capable of doing considerably more than traditional assistive technology.
So perhaps the more useful question isn't:
"Was AI used?"
It's:
"What barrier did the technology remove, and who actually did the thinking?"
Better Output Doesn't Mean the Difficulty Has Disappeared
This is particularly important in workplaces.
Imagine someone with dyslexia who previously sent emails containing spelling mistakes.
They start using AI and suddenly every email is perfectly structured and grammatically correct.
That doesn't mean their dyslexia has disappeared.
The tool has simply changed the visible output.
Likewise, someone who previously struggled to organise a large project might use AI to create a clear task structure.
That doesn't mean the underlying difficulty no longer exists.
Technology can hide barriers remarkably effectively.
Employers, teachers and colleagues need to be careful not to interpret improved output as proof that support is no longer necessary.
Using AI Doesn't Make the Achievement Less Theirs Either
The opposite assumption can be equally unhelpful.
If someone uses AI to remove a barrier, that doesn't automatically mean the resulting work isn't theirs.
A dyslexic business owner who uses AI to correct the spelling in an email still had the conversation.
The ideas are still theirs.
The business decision is still theirs.
AI simply helped remove a barrier between the thought and the finished communication.
Tools change how people interact with a task.
They don't automatically remove ownership of it.
This is also why the distinction between AI as a tool and AI as a replacement for judgement matters so much. At Business Sorted Kent, for example, AI is part of how modern digital work can be produced more efficiently, but the final decisions and checks remain human. Our approach to AI and modern tooling explains that distinction in more detail.
What This Could Mean for Small Businesses
There's an important workplace angle to all of this.
Businesses have traditionally bought software and expected employees to adapt to it.
Here's the system.
Here's how it works.
Learn it.
AI creates the possibility of reversing some of that relationship.
Instead of requiring everyone to process information in exactly the same way, the information itself can potentially adapt.
One employee might want a detailed written brief.
Another might ask AI to turn the same brief into five actions.
Another might prefer a summary.
Another might ask questions about it conversationally.
The underlying work hasn't changed.
The route to understanding it has.
For small businesses, this doesn't necessarily require expensive specialist systems either. Modern AI tools and business automation increasingly allow information and repetitive processes to be handled in different ways depending on what the business and the people within it actually need.
That's potentially a much bigger shift than simply using AI to write emails faster.
The Most Exciting Part of AI Might Not Be Automation
Business conversations around AI tend to focus heavily on replacing repetitive work.
That's understandable.
Automation can save considerable time, and practical systems can already handle routine questions and enquiries while leaving more important conversations to people.
But accessibility might ultimately prove just as important.
The ability to say:
"Explain this another way."
"Turn this into manageable steps."
"Read this and tell me what matters."
"Help me communicate what I'm trying to say."
"Prepare me for what's going to happen."
Those are relatively simple uses of AI.
But for the right person, in the right situation, they can make a meaningful difference.
The OECD has also explored the potential for AI to support neurodivergent learners, including people with autism, ADHD and learning disabilities such as dyslexia, while stressing that the opportunities need to be considered alongside the limitations and risks.
AI Should Give People More Control, Not Less
That's probably the simplest way to judge whether AI is genuinely helping.
Does the technology give the person more control?
More independence?
More ways to communicate?
More ability to access information?
More choice over how they approach a task?
If it does, there's something genuinely valuable happening.
If instead it makes decisions for them, creates dependency, overwhelms them with information or encourages everyone to communicate and think in exactly the same way, then we've missed the point.
Neurodivergence is broad.
People are different.
Their needs are different.
The tools that help them will be different too.
AI shouldn't be about forcing everyone towards one supposedly "normal" way of working.
Its real potential might be the opposite.
Giving more people the freedom to work, communicate and understand information in the way that works for them.
That isn't AI replacing human ability.
It's technology removing some of the barriers that can get in its way.
Further Reading About AI and neurodivergence
For anyone interested in exploring AI and neurodivergence in more depth:
- OECD: AI to Support Neurodivergent Learners in Vocational Education and Training
- National Autistic Society: Advice and guidance about autism
- British Dyslexia Association
- ADHD Foundation
These are useful starting points for understanding the wider subjects of neurodivergence, accessibility and support beyond AI itself.