Accessibility That Lasts
Transcript
Made automatically from the video's sound, so it may contain mistakes. Click a line to play from there.
Right, welcome to the grand finale explainer of our digital accessibility core course. So, now that we know how to order interfaces using automated tools, manual passes, screen readers, and paid disabled user research, which we covered in our last session, we're shifting gears.
We are moving way upstream to permanently embed accessibility right into our design systems, vendor procurement, CI pipelines, and well, our whole team culture. It's time to stop just finding bugs and start systematically preventing them altogether.
Let's have a quick look at the agenda. We'll start with the AI overlay allusion, then look at five modalities in production, before moving on to shift left design and procurement. After that, we'll cover automated CI regression gates, unifying the user journey, and wrap up with accountability and our companion labs.
Okay, diving straight into section one, the AI overlay illusion, when quick fixes fail in production. Now, you might be thinking, can't we just drop in one of those magic AI accessibility scripts?
Absolutely not. This isn't just theory. Over 700 accessibility practitioners, experts, and disabled users have signed the overlay fact sheet.
They clearly state that these one-line quick fixes actually fail legal compliance. Worse still, they can expose a user's sensitive disability status without their informed consent and actively harm people using native assistive tech. It's a bit of a disaster, really.
To put this into perspective, just look at this verbatim speech diff. On the broken side, the marketing overlay injects a massive amount of repetitive, unlabelled noise. Imagine trying to navigate a site, and a screen reader is literally shouting, accessibility menu, button, press Alt plus 9 for screen reader mode, unlabeled button, frame at you.
No thanks. Contrast that with the fixed side, where we've stripped out the overlay and just use natively accessible DOM elements. There's so much karma.
You get a logical, polite navigation flow. Skip to main content, link. Pathway course catalogue, main landmark. Upstream engineering wins hands down every single time.
Five modalities in production. Exposing post-launch breakages. Let's actually expose what happens post-launch when procurement drops an unvetted vendor chat iFrame onto our operations desk, or marketing adds that AI overlay to Pathway.
The physical breakage across strict accessibility modalities is undeniable. For keyboard users, that overlay injects a staggering 18-stop tab detour before they even reach the skip link. And at 400% zoom, which means a 320 CSS pixel reflow, colliding chat launchers completely obscure critical buttons like save and continue.
It proves we have to enforce governed viewport safe design tokens, not just bolt things on at the end. Empathizing through these extremes shows us the stark physical reality for users. On mobile, or when using voice control and switch devices, those floating overlay launches actively intercept bottom edge taps.
So if a user relies on iOS voice control and says tap save, the overlay blocks it. It physically stops the user dead in their tracks. We have to rip these widgets out and rely on paid disabled contributor release sign-offs and SLA tracked feedback instead.
So how do we stop these downstream disasters? Section 3. Shift left design and procurement, preventing 60% of defects upstream. The absolute game changer here is enforcing specific standards before a single line of code is ever written.
We're talking about embedding Figma annotations directly into your design systems. Things like token-baked 4.5 to 1 or 3 to 1 contrast ratios, dual-band focus visible outlines, explicit reading orders, and ensuring target minimums hit that 44 by 44 pixel sweet spot.
By making accessibility a core design primitive with proper heading levels and live regions, teams can systematically eliminate over 60% of accessibility defects right at the source. And it's not just our own designs.
We have to be just as strict with what we buy. When procurement is eyeing up a third-party vendor, say for a chat widget, we must ruthlessly scrutinize their VPAT. That's the accessibility conformance report based on Section 508 standards.
Nowadays, you'll increasingly see these authored in a machine readable open ACR YAML format. This is brilliant because it lets us rapidly cut through the vendor's marketing fluff and compare the actual hard accessibility data.
10 minutes. That is literally all it takes to expose those evasive partially supports claims that vendors love to use. We really need to arm ourselves with aggressive pragmatism here.
Do not sign a vendor contract without demanding a mandatory 10-minute live keyboard and screen reader smoke test. Seriously, trust but verify. Okay, let's look at the code.
Section 4, automated CI regression gates, enforcing elite engineering standards. To guarantee these elite standards actually survive into production, we wire a robust three step automated regression gate directly into our CI CD pipelines. First up, component testing with Storybook.
Second, command line automated scans using Perlevin. And third, end to end playwright testing using Axe Builder. We explicitly integrate this with escape key focus return assertions to physically block bad pull requests from ever being merged.
Here's a massive and incredibly simple automated engineering win for you. Configure your storybook add-on to fail the build on an error rather than just lazily flagging it as a to do. By doing this, you're creating a ruthless first line of QA that instantly catches up to 57% of blatant WCAG violations right at the component level.
It's a total no brainer. Unifying the user journey. End-to-end repair of pathway and operations desk. This right here is the climax of our nine-part core course.
We are finally synthesizing everything from episodes one through eight into a single cohesive user journey across our pathway and operations desk apps. We're tracing a complete release, starting with connecting logical headings and skip links, enforcing perceptual contrast and calm motion all the way through to frictionless pass key sign-ins and error summary enrolment forms.
And we cap it off with dialogue focus restoration, polite screen reader progress saves, and beautifully edited Web VTT captions with sortable cohort tables. This is what true end-to-end multi-layer verification looks like in the wild. And we have to bridge these fantastic technical wins straight back to inclusive human rights principles.
How? By publishing an honest gov.uk modeled accessibility statement, none of those deceptive overlay promises here. We are talking about transparent conformance status, openly listing known gaps with specific dates for when they'll be fixed, staffed feedback SLAs, and properly embedded AI guardrails, like an agents.markdown file.
It's all about accountability. Accountability and companion labs, bridging to advanced deep dives. With our core foundation safely locked in through systemic governance and CI pipelines, we're officially bridging into our four optional companion technical labs.
If you want to push your continuous engineering growth even further, these labs are where we get into the absolute weeds. We've got React, Focus Management and Roving Tab Index, Charts and Data Sonification, and native mobile development across iOS Swift UI and Android Compose.
It's gonna be brilliant. So this leaves us with one final vital question as we wrap up this core course explainer. Are you settling for downstream, reactive bug fixing and deceptive AI overlays just to tick a compliance box?
Or are you ready to step up and engineer upstream interfaces that truly effortlessly empower every single user? Because remember, accessibility isn't just an add on, it's the very foundation of brilliant engineering.
Thanks so much for joining me on this journey, and I'll see you over in the companion labs.