AI for Digital Accessibility

View the Video

Read the Transcript

Orris Long [00:05]: Hey, everybody!

Thanks for joining us today. We’re going to take a moment while people finish logging in.

While we’re getting ready, we do have two ASL interpreters on the call today. Alex and Nicole, they’re in the window here. You can pin them if needed, if you need those services.

And thank you all for joining AI for Digital Accessibility. We’re really excited about today’s webinar.

All right, let’s go ahead and get started. Again, for anybody that’s newly onto the call, we have two ASL interpreters, Alex and Nicole. You can pin both of their windows to the screen if needed.

So, again, let’s get started. Hi, I’m Orris Long. I’m the SVP of AI Solutions and Sales here at Onward Search. I’ll be leading today’s conversation along with my colleague, Pete Bruhn.

Pete is our Director of Accessibility Solutions, and I’m going to hand it over to him to briefly introduce himself.

Pete Bruhn [01:43]: Great, thanks, Orris. As mentioned, my name’s Pete Bruhn, Director of Digital Accessibility Solutions here at Onward. I’ve got a 10-year background in the digital accessibility space, and I’ve been at Onward for a little over a year now, helping build out our service side of the business.

We’ll tell you a little bit more about what we offer in services at the end here, but I’m really excited to be here and to host this panel. We have some amazing guests here, and we are really looking forward to sharing our knowledge with you.

Orris Long [02:12]: Awesome. Thanks again, Pete. And thank you all again for being here. Let’s walk through today’s agenda.

So, here’s how we’re going to spend our time together. I’ll start by sharing some of the AI challenges that we’re hearing from clients overall, not just from an accessibility standpoint.

Pete’s then going to explain how those challenges are multiplying right now when it comes to accessibility, but [he’ll] also identify where there’s opportunities. And then from there, we’ve got a great panel discussion from experts to take your questions.

We’ll explore, do a deeper dive in terms of where we’re seeing AI for accessibility. We will open it up for a Q&A.

And then we’re going to close with different ways that Onward can support you through both your accessibility journey and your AI transformation journey. So, first off, let’s start off with what clients are saying right now.

As I had mentioned, I lead our AI Solutions practice, and clients across the board, within product, within marketing, creative — we’ve been doing a deep dive with those clients over the last couple of years.

The consistent themes that we hear are pretty consistent: there’s hundreds of tools, no clear guidance on which ones are worth adopting.

Even if clients knew which path was clear, most teams don’t have the bandwidth to execute right now. We all know internal teams right now are risk-constricted, there’s a lot more pressure, there’s a lot less resources.

Companies are tasked with doing more with less, and there’s a real concern around AI bias, hallucinations, trust issues.

That problem’s obvious across the board with new technologies, and it’s exacerbated across the accessibility community. And on top of all of that, there’s pressure from leadership to do something with AI without clarity into what’s actually worth doing.

So, as we talk about that shift to accessibility issues, these issues remain, but I think they’re multiplied from the accessibility side. And with that, I’ll let Pete dive into more in that area.

Pete Bruhn [05:31]: Great, thanks, Orris.

So, there’s definitely a lot to talk about when it comes to AI, and what it can and can’t do as it relates to digital accessibility. It’s one of the reasons why we’re here today, to have this conversation.

Just a few things off the top: AI can really help scale accessibility work. Think about things like alt text generation, screen reader compatibility, automated testing. Those are all areas where AI can be really helpful.

But on the flip side of that coin, AI can also scale inaccessibility if biases or poor training data go unchecked. So, there’s a lot of concern that needs to be taken from both sides of the coin here.

But to focus on the opportunity, we’re talking about increasing efficiency at scales really not seen before. We’re talking about being able to have more consistent results and deliverables. We’re talking about more reliability when you’re using AI correctly as a tool.

But when we think about accessibility, the core of it is really the interactions between humans and computers. There’s a lot of nuance there, and I don’t think we’ll get to a point where we ever really rely on machines to define what is a good user experience.

But it can really be helpful, again from a workflow perspective, to make things more efficient and much more scalable.

So, with that, let’s introduce our panel members and kick off the discussion.

Pete Bruhn [06:14]: All right, Dylan, you want to introduce yourself?

Dylan Isaac [06:17]: Hey everyone, my name’s Dylan Isaac. I was formerly at DQ, and now I run my own independent consulting company, Enablement Engineering. So, if you all are interested in the intersection of AI and accessibility, please hit me up.

Claudio Luis Vera [06:35]: I’ll jump in next. I’m Claudio Luis Vera, and I’m a long-time UX designer who switched over to accessibility about 8 or 10 years ago. I’ve come from a point of usability, come from a point of human issues, and I’ve been doing a lot of writing and talking lately, particularly on the ethics of AI, the ethics of working with technology that doesn’t work 100% well.

Sam Gong [07:05]: Hi, everybody. I’m Sam Gong. I’ve been in UX for over 25 years. I was a senior UX manager and principal AI designer at my previous FinTech, where I created the first AI Center of Excellence, which we successfully launched.

The first AI VDR product in the fintech space was back in 2019, and since then, I’ve overseen the successful design and launch of four other fund management products, all integrated with AI and all WCAG compliant.

Pete Bruhn [07:40]: Alright, awesome.

So, let’s get into the panel discussion. Sam, first question, why don’t you kick things off? If you’re looking to use AI to elevate and accelerate your accessibility efforts, but you’re really unsure where to begin, how do you recommend getting started?

Sam Gong [08:00]: Your first step should be to take an honest assessment of the AI maturity within your organization. If you’re in an org with a dedicated AI or innovation leader, your road’s going to be a lot easier, but for the rest of us, the number one barrier for AI adoption is not being able to identify a use case. Fortunately, accessibility is an easily understood use case, so the next step for you is to seek out other functional leaders in your org to collaborate in defining the value to business. This will ensure that the value of your efforts will be understood and thrive at a cultural level within your enterprise.

Pete Bruhn [08:50]: That’s great.

Dylan Isaac [08:50]: When I think about getting started with AI, especially in accessibility, start where you’re comfortable. Handing off AI to do your repetitive tasks that you find tedious will both build efficiency and build intuition in yourself on what works with AI and what doesn’t.

The next thing I would go into once you have that intuition is to think about using AI as a thought partner, not as a search engine. Instead of saying, “What ARIA attributes fix this?” say, “Help me understand what I’m trying to accomplish. I got this issue, I sent this code, and then what are the decision-making processes I can come about this?” That way, you start building capability, and you’re not building a dependency on a tool.

Claudio Luis Vera [09:42]: Adding on to that, my background isn’t so much in coding, but a lot of it is doing data analysis, looking at the results from accessibility scans, and compiling all that. AI is really good for writing scripts that you sort of know how to write, but you’re not particularly good at.

If you’re somebody who could throw something together in Python, but you’d spend hours fixing it up — or same with JavaScript, or something like DAX, like you’re writing Excel or Google Sheets equations — AI is particularly good for writing short programs that have a limited function and aren’t going to get away from you.

I would use it for that, and maybe for summarizing content in an accessibility report, to build up your confidence on it.

Pete Bruhn [10:37]: That’s great. Sounds like starting small is a general good practice here, right? Don’t go in over your head. Think about things you already know how to do, and see how AI can support those things so you can be critical of it, as opposed to trying to do something you have no idea about and hoping that it’s correct.

Dylan Isaac [11:02]: Don’t go beyond your skill level. Stay in that in-between — where you are good, and a little bit beyond — so you can continue to grow. Don’t go beyond.

Pete Bruhn [11:17]: Great, good point. Alright, Dylan, in your experience, where can AI models and agents add the most value in accessibility workflows?

Dylan Isaac [11:26]: I think AI’s biggest value isn’t really finding violations. AI’s biggest value is being a translator between different ways of knowing.

My example is, when I was an accessibility coach at DQ, my main value for the people I was coaching was that I could speak accessibility, design, code, and business fluently. I was able to adapt those different perspectives and requirements into what mattered to each role at that given time. However, AI can do this at scale. The models are trained on so much information that all of those perspectives are contained within the model, so AI can bridge those gaps in the same way I was doing manually.

An example is using AI to adapt between the visual intention of a designer to what a screen reader user would need for something like a header. A visual designer might add some bold text, and immediately under it you have paragraph text. Visual users instinctually know proximity and scale put the relationship there and give the role to the header.

What AI can do is, since it can speak that visual language, and it knows how to write the code that screen readers interpret, it can adapt from one person’s way of expressing information into another way of expressing information. It’s acting as a translator.

Pete Bruhn [13:08]: Interesting. Claudio, do you have anything to add to that?

Claudio Luis Vera [13:12]: Coming from a design point of view, Sam touched on this at the beginning: you spend a lot of your time defining your problem. If you’re working with AI, you have to write an incredibly elaborate or involved prompt. In doing so, what I find is a lot of times three-quarters of the work is there.

If you think of it as a double diamond, the first half is problem definition, and the second half is implementation or problem solutions. In that first half, just getting to understanding the problem and defining it well enough is a lot of times three-quarters of the way there.

Pete Bruhn [13:59]: Alright, Sam, do you have anything to add to that one?

Sam Gong [14:01]: Sure, my answer’s going to be different, because as UX, I don’t look at accessibility as a workflow the way a developer would. Instead, I’m going to look at how UX enables the integration of AI and accessibility into products. In my use case, AI and accessibility are two separate product initiatives, and they share a common roadblock to implementation: they both need some kind of scaffolding to ensure their sustainability.

For products, AI sustainability means UX needs to continuously evolve to support various lifecycle stages of data science, all while minimizing design disruption. If you think of it like the story The Giving Tree — where the design system is the giving tree and the AI is the child — there’s going to be a curious newborn phase of AI, where the user experience needs to enable the collection of as much data as we can.

Then, we’re going to have an unsure teenager phase of AI, where the UX might test multiple interactions with users within the interface in order to validate various algorithms. Finally, you’re going to have a wise old man phase where the algorithm doesn’t need more feedback or validation. A good design system helps this business problem by being modular.

UX would design templates that govern and confine any evolution of those changes to a specific area on a page, or to a specific page within a flow.

As for accessibility, sustainable means that the principles of compliance and accessibility are already baked into the design of every component.

Pete Bruhn [15:11]: Okay, yeah, that’s great, that’s a good point.

It’s interesting, because when you think about AI, there are models, there are tools. Especially with accessibility, I think most people are thinking about AI in the tools they’re using to support workflows.

From that perspective, Dylan, what are your thoughts on which AI-powered tools or models are the most reliable for accessibility?

Dylan Isaac [16:40]: Yeah. This is going to age quickly, just because of how fast everything’s moving, but my quick answer right now is Cloud Code. It’s the best AI tool because it can connect to anything, run anywhere, since it’s just a terminal application, and it’s programmable.

However, GPT-5 and Gemini are exceptional models with incredible multimodal understanding. If we’re going back to that translation capability between visual presentation and semantic presentation, that’s very important.

But beyond the staleness of specific answers, here’s my criteria when I look at tools. First: does this provide glass box versus black box reasoning? Does it show its work? Can you actually look at its reasoning steps and understand where the tool’s coming from? You want a cognitive partner, not an oracle. If it can’t explain why it came to a conclusion, don’t trust it.

Second: does it integrate with your current workflow? AI is only as good as the context you give it. You want the tool to embed into your current stack and hook into the data sources you need to make intelligent decisions.

Third: does it build your capabilities, or does it create a dependency? Does it teach you why, not just what? Does it make you smarter while it assists? If it doesn’t, then it’s just technical debt disguised as an accommodation or productivity boost. That’s my framework for evaluating new AI tools, because it’s moving too quickly.

Pete Bruhn [18:34]: Yeah, that’s a good point. In a month, six months, we might want to revisit these questions, and things could be different. But that’s interesting about glass box versus black box.

To elaborate, how would somebody who’s new to this find out whether a tool is black box? Is it just about asking?

Dylan Isaac [19:03]: Yeah, I would just start using it. Does it give just a straight answer, or in some AI chats can you see reasoning steps? You want to see that. What decisions were made behind the scenes, instead of just answers popping into your UI?

A lot of AI UX right now is trying to make everything “magic,” but AI can make mistakes, especially when it’s undefined in a workflow. You don’t want magic answers popping in, because you can’t trust it 100% of the time.

However, if you’re working on the edge of your capabilities — growing with AI, using it as a thought partner, seeing its reasoning steps — you can see where it goes off the rails. It’s usually a simple mistake, and you can just ask it to fix that one thing, clarify, or regenerate and try again.

Any tools that auto-populate and act like an oracle are the ones to avoid, because they’re not boosting your capabilities, and the tech isn’t 100% dependable.

Pete Bruhn [20:18]: Yeah. Sam or Claudio, do you have anything to add to that?

Sam Gong [20:25]: I’m going to defer, because this is more related to AI workflow. But I do want to say I love Dylan’s idea of AI as a partner. That’s an important consideration of voice and tone.

Dylan Isaac [20:45]: Yeah. Imagine we have access to little tiny data scientists in our pockets. Every accessibility tool lets you export out CSVs of your issues — your auto-scan issues. What if you could use your little data scientist in a pocket to make sense of this messy signal, with duplicated issues, or similar components?

What if it could use intelligence and data science to do some Excel kung fu, and come up with insights that would have taken you months to realize before? You don’t need an expert anymore. You’re enough. You just need your intention and the right tool.

Claudio Luis Vera [21:42]: That said, sometimes the insights are obvious — “thank you, Captain Obvious.” But I think, Pete, this is a good segue into mistakes to avoid.

Pete Bruhn [22:00]: Yeah, let’s roll with that. What are the most common mistakes to avoid when adopting AI and integrating it into digital accessibility workflows?

Claudio Luis Vera [22:16]: Number one, especially if you’re working with an LLM: it’s never going to tell you “I don’t know.” It’s like a high school sophomore who won’t hand in a blank paper — they’ll give you all kinds of BS. Expect that if the LLM has thin or weak knowledge. You have to be skeptical.

Dylan Isaac [22:42]: One thing I’ve done is regenerate the same answer two or three times. If you see consistency, you can feel more confident it’s correct. If it’s changing, that means the AI is guessing — like a high schooler on a multiple-choice test. It’s rewarded to give its best guess.

Claudio Luis Vera [23:19]: Another thing to watch out for: these are the same tools that can drive a person to psychosis. There are cases of teenagers using LLMs who got into toxic loops and took their own lives. You want guardrails: Is a minor accessing it? Is someone using it at 4 a.m.? Are certain topics taboo?

In an enterprise or workplace, you need guardrails and also think about data security. You don’t want your data to become training data for something else, especially if it’s confidential.

Sam Gong [23:13]: Yeah, that’s a good point. In high-maturity enterprises, the leading barrier for AI implementation is failure to establish security policies.

In my last AI project at FinTech, our users — investment bankers — wanted security and retention policies for training data in writing. From their perspective, data was part of their sensitive investor information ecosystem.

Another pitfall: forgetting to validate voice and tone with actual users. Voice and tone isn’t just copy — it’s the entire design treatment. In that project, corporate development bankers overseeing multi-million-dollar M&A deals had a mental model: if they used wrong numbers from AI, they’d get fired.

AI wouldn’t get fired. So the AI needed to sound like a junior assistant or intern whose work had to be checked, not a self-assured investment banker. That was their biggest concern and the deciding factor in adoption.

Dylan Isaac [26:12]: From the perspective of using AI, the biggest mistakes I’ve seen are people asking for way too large of work to be done, then wondering why results are inconsistent. If you just say something generic like, “Make this page accessible,” every vague assumption you put in is an opportunity for the model to hallucinate and go off into new directions. Confabulations compound quickly.

You have to be specific. Like: “Analyze this form component. Suggest ARIA labels for error states.” Make everything in your prompt specific, measurable, and validatable. That’s how you get good AI results.

AI is just going off the top of its head. If I say, “cat in the…” your brain probably auto-completed “hat.” That’s what AI does, about everything. But if you constrain it with data points, tests, and documentation, it can generate much longer without issues.

If you ask for large amounts of work without grounding techniques and specific measurable details, you’ll run into problems — like bankers reporting numbers they shouldn’t, or lawyers citing papers that don’t exist.

Sam Gong [27:59]: I see a question in the webinar chat from Steve Goldberg. He asks: “How do you feel about companies asking for people to be AI experts with such new technology that is so fluid, where learning is ongoing?”

This is important. A lot of companies are trying to determine a strategic reason for adopting AI. It’s important to view AI as a tool. There’s no intrinsic value in the tool itself — only in the problem it solves.

When I say there’s no intrinsic value in a tool, think of a screwdriver. In the hands of a baby, nothing gets done. In the hands of a master carpenter, it has great value. The question is: what’s the value of what you’re doing with the tool, and what tenets does your organization need to develop around it?

Orris Long [29:35]: I think this is a great question. I’d love to get everybody’s feedback, and I can add my own as well.

So Dylan and Claudio, did you guys get that question?

Dylan Isaac [29:53]: Could you repeat it one more time?

Orris Long [30:01]: Steve asks: “How do you feel about companies asking for people to be AI experts with such new technology that is so fluid, where the learning is ongoing?”

Dylan Isaac [30:05]: Great question. I wouldn’t have considered myself an AI person until recently. I’ve always been an accessibility person. But I see AI as an opportunity to scale accessibility in a way never possible before. I’ve been deep-diving into how it can work and multiply accessibility.

A lot of my machine learning friends don’t see the same possibilities. It’s challenging, because people who look good on paper aren’t necessarily up-to-date. What you should look for are people actively engaged with new research, trying to build things, showing things. The cost from idea to proof of concept is almost zero now. If someone can’t demonstrate, that’s a red flag.

Orris Long [31:28]: Claudio?

Claudio Luis Vera [31:38]: I think it depends whether you want AI experts to be teachers or individual contributors. Out of fairness, an expert should be creating workflows others can follow — not being a wizard behind the curtain.

For leadership, it’s less about credentials and more about how well they bring up the rest of the workforce, creating workflows that make productivity gains available to everyone, not just themselves.

Dylan Isaac [32:32]: Another thing: look for good communicators. People who can identify what’s important, explain why, and communicate intention to AI. AI doesn’t have human intention. It gives default responses, but you have a unique point of view. If you can encode that into a prompt and show AI how to manipulate information your way, you’re an AI engineer.

Orris Long [33:36]: I’ll step in. This is really good.

From my perspective — and my team at Onward is tasked with finding AI experts for clients — the implied question is: how can companies ask for AI experts when it’s a new technology? How do we identify experts?

The good news: if you feel behind, you’re not. A lot of people are just getting into this. Many don’t trust AI. But this isn’t a flash in the pan. The time to learn and get involved is now.

Also, AI’s been around a long time. Google is AI, and it’s been here 30 years. Generative AI like ChatGPT has only been around three years, but everyone got thrown into this pool at the same time. That’s reassuring.

When finding talent, what we look for are people who are passionate, who use these tools, who show it in their work. They don’t need robust experience — hiring managers are reasonable — but they want passion and proven workflows.

There’s also noise. People claim to be experts because they see where the market’s going. It’s similar to UX when it first emerged. A lot of people claimed to be UX designers, but weren’t. Passion, workflows, and examples are what matter.

Sam Gong [37:47]: Let me step in here. What’s fascinating is how this ties back to the use case. Dylan said he never saw himself as an AI person — because AI is the tool, accessibility is the use case.

When you have a use case, you’re open to any tool. If you focus on a tool, you’re limited. You might be an expert in a tool, but you won’t solve problems.

Think of a hammer. A hammer is not just a hammer. It’s metal made to drive nails into wood to build a house. Why not use a shoe? Because the interface matters — it needs to be held in the hand. Responsible tool users look at facets of tools: why, when, how, who should use them.

Too many people just say “AI” without thinking about the responsible factors that define its use.

Dylan Isaac [40:08]: Talking about hammer capabilities reminds me of ARIA attributes. Adding capabilities onto an existing element — it all comes together.

Sam Gong [40:10]: We’re all tool ontologists.

Pete Bruhn [40:23]: This has been fascinating. I want to make sure we discuss one of the most important questions: why is human expertise still essential for inclusion, context, and validation when talking about AI? Would love your thoughts.

Sam Gong [40:39]: From a UX perspective, this is simple: it’s the principle of a human in the loop. Technology is powerful, but it doesn’t always get it right.

For example, digital tools can simulate what it’s like to be colorblind, but they’re imperfect. Photoshop’s colorblind alerts sometimes tell me I shouldn’t be able to see something I clearly can see. That’s where human expertise comes in. Machines can approximate, but they can never validate lived experience.

Inclusion means recognizing the value of involving real people with real perspectives. Early in my career, I told Staples.com I was colorblind when I interviewed. Instead of seeing it as a limitation, they saw it as an advantage. Staples was bringing in close to $1 billion annually in online revenue, second only to Amazon at that time. By designing with colorblind users in mind — 7% of the population — they saw the potential to reach $70 million more in revenue.

That’s the value of human talent: lived expertise provides context and validation technology alone cannot.

Dylan Isaac [42:33]: I completely agree. Humans define intent, and AI translates it. Only humans know why a design decision was made, what they’re trying to communicate. AI is brilliant at translating between ways of knowing — via APIs or communication methods — but intention cannot originate from AI.

Humans provide the why. Human plus AI collaboration provides the how. And AI gives scale. Human expertise defines vision, AI scales it, and collaboration gets it ready.

Claudio Luis Vera [43:39]: I’d bring up two issues AI doesn’t get right. First, big picture: bias. LLMs and decision engines have biases from training data, scoring systems, or design. Don’t fall into automation bias, where you trust AI more than human judgment. Even with both, humans may defer to AI to avoid arguing every time. Humans in the loop must be respected.

Second, a smaller point: I wrestle with Grammarly all day. It wants to turn everything I write into bland corporate speak. AI is fantastic at that. But if you want spark, originality, or brand voice, a human brings that.

Orris Long [45:02]: I’ll chime in. On the creative side, I believe you get what you put into AI. If you train it with the best creative voices, you can get killer output. We’ve done webinars on this. The output can be phenomenal.

But the human is most important. It’s not about removing humans — it’s about enabling teams to be better. AI has bad PR, thanks to Frankenstein and Terminator stories. At the end of the day, it’s a calculator, another abacus, another tool. There’s opportunity for people to take their voice to the next level.

When people first oppose AI or distrust it, once they recognize its limitations and opportunities, they get excited.

Dylan Isaac [47:11]: To Claudio’s point, there are big representation gaps in models. Image generation struggles to produce disabled people. I’ve been working on a UX tool simulating personas with disabilities. The default decision-making isn’t accurate. It comes back to “nothing about us without us.”

We need more accessibility training data in models. Right now, there’s not enough good accessibility information, so we can’t rely on defaults. If AI labs build accessibility into models, those results will get distributed by default.

Orris Long [48:41]: Great feedback. Let’s get a few more questions from the group.

Dan asks: “How do you handle a situation when you’ve identified the best AI tool, but it’s not allowed by the company? Do you get a second laptop?”

Dylan Isaac [49:05]: Depends how much you value rules.

Sam Gong [49:15]: I’d say don’t subvert rules. From an enterprise view, that shows you’re not collaborating to define a business case. If it’s valuable, measure it, explain it, build a hypothesis, and find the right people responsible for policies. Maybe it’s time to revisit procurement.

Pete Bruhn [50:21]: Sounds like three options: build your own tool, find an alternative, or make a business case for the one you found.

Sam Gong [50:30]: Exactly. If you leave, the tool leaves with you unless you build culture around it. It’s always people.

Pete Bruhn [51:04]: Next question, from Aaron: “Is there an open-source API that mutates HTML on-page load to solve accessibility issues — to ensure code matches intended semantic meaning?” Dylan?

Dylan Isaac [51:49]: Pros: code is free now. Having AI fix a clear accessibility issue, like adding a label to a button, would be incredible. But at scale, misrepresentation is a risk.

I dream of a community-driven app: AI agents look at webpages, identify issues, write JavaScript fixes, and users confirm yes/no. Snippets are saved for reuse. Issues get forwarded to the people who need to solve them. Companies could pay for that data.

Claudio Luis Vera [53:09]: Sounds like Waze for webpages.

Orris Long [53:15]: We’ll wrap this part up and move into the next. Pete, overview?

Pete Bruhn [53:28]: Onward Accessibility is the only accessibility company doing both staffing and solutions. It’s a model that helps organizations shift left and build sustainable programs.

You need internal folks, but many orgs don’t have headcount or want third-party perspective. We [can provide you] support with staffing experts, inclusive hiring, audits, services, training, and consulting.

Training and remediation are huge. Audits alone don’t help. We want fixes, and we want to train developers and teams. That’s how we make the internet more accessible.

Orris Long [55:53]: Thanks, Pete. Talking about Onward Search and AI offerings, we have three main solutions.

First, hiring AI-skilled professionals. Second, training and upskilling teams to use generative AI tools in their workflows. Third, implementation: discovery, prototyping, deployment support.

We help find value, create use cases, and deploy custom or off-the-shelf tools with real experts.

As for what’s next: you’ll get a full video recording and transcript of today’s webinar in your inbox in the coming days, plus a short survey. Your feedback is valuable.

Finally, thanks to Dylan, Claudio, and Sam for your time and effort, and thanks to Pete for leading Onward Accessibility’s great work.

Claudio Luis Vera [58:19]: Terrific. Thanks, everyone.

Pete Bruhn [58:21]: Thanks. Bye, everybody.

Dylan Isaac [58:21]: Thank you all. Appreciate it.