Equidox AI for Healthcare

[Paul Campbell] Hello everyone, welcome  to our webinar today for Equidox AI,   which is a fully automated PDF remediation  solution. We're very excited about this   cutting-edge technology that is solving  for challenges within the healthcare   and healthcare insurance provider  industries. By way of introduction,   today my name is Paul Campbell, and I will be  joined by Dan Tuleta for the next 30 minutes. Some quick logistics: if you have questions  during the webinar, please feel free to drop   them in the Q&A chat button at the bottom of  your screen and we will get back to you with   an answer. Additionally, you know, this webinar  will be recorded and will be sent to you after   the meeting in addition to the slide deck and  a short survey. We're happy to do a more direct   interactive session with you and your other  team members if they're not available to join,   and obviously, further our conversations  as to whether Equidox AI may be a fit for   your organization specifically  after you review here today. As far as the agenda today: first of all,  I'm going to, you know, go through who is   Equidox Software Company and, you know, where  have we been and where are we going. Second,   you know, the challenges that we've seen in the  healthcare market as it relates to remediation   of PDF documents and our solution to that  problem. I'm going to hand it over to Dan,   my colleague, to talk about why do we  make PDFs accessible and, you know,   what's driving this. This will be followed  by an overview of Equidox AI, how it works,   and finally, a demonstration for the last 15  minutes of the half-hour we have here today. So who is Equidox? Equidox has been in existence  for nearly a decade now. Way back when,   a Canadian citizen was trying to apply for a  government job posting on the internet but,   unfortunately, was unable to do so because of  her visual disability. She sued the government   and won her case, and the government of Canada  sought out a solution to accommodate, you know,   this use case in the future. But they  couldn't find one, and there was no one   really addressing it in the market, so they  asked the marketplace to respond. Hence why,   you know, we started building our remediation  solution and haven't stopped innovating since. So originally we created a software-as-a-service  solution, you know, way back, and now, you know,   hundreds of customers currently use our SaaS  product. This solution is a world-class, you know,   solution and is adding tremendous value to the  marketplace of digital accessibility for not   only enterprise organizations but, you know,  mid-level organizations, small businesses,   governments, and educational institutions as  well. You know, our customers love the product,   evidenced by the fact that nearly 100% of our  customers renew their subscription every year. And while this is all well and good, you know,  we started hearing from organizations that,   you know, had tens of thousands or hundreds  of thousands or even millions of pages in   documentation that needed to be remediated, and  there just wasn't a fully automated solution for   that type of daunting need. Our SaaS solution,  you know, is still a manual solution that has,   you know, some automated functions but doesn't  truly automate the full process, and you know,   nothing out there does with the quality. So  you know, the traditional service providers,   you know, at that time, you know, were sending  the documents to India and other countries,   but you know, there's really no efficient way  to scale. So companies were forced to settle for   a solution that is, you know, time-consuming,  expensive, and doesn't truly mitigate the risk   of a lawsuit with some of the software tools  and some of the, you know, service provider   solutions that were out there at the time. So enter Equidox AI, and you know, Equidox AI   is a fully automated PDF remediation solution that  removes the traditional manual remediation methods   and auto-tagging methods while still increasing  the quality, accuracy, and compliance. The Equidox   AI solution is utilized for use cases where  there are templated, reoccurring large volumes   of documents where manual remediation methods are  just too cumbersome and daunting to address. And,   you know, we've found over the past couple of  years that the healthcare industry can really   get the most value out of Equidox AI, and this  is because of Section 508 compliance mandates,   the large quantities of templated and reoccurring  documents that consistently need to be accessible   because of these compliance requirements and  the challenges with current processes today. The quality, speed, and overall vendor, you  know, just sprawl that these customers possess,   you know, currently at these organizations. So  the use cases, you know, may include but are   not limited to the biggest one that we've seen:  physician directories. You know, these can be,   you know, documents that are updated on a  reoccurring basis and can be very daunting   to remediate and complex. Additionally,  explanation of benefits, digital ID cards,   invoices, statements, etc., all these are  really great candidates for Equidox AI   because of the high volume, repetitive,  and templated nature that they possess. So, you know, we found that there are three  main challenges for healthcare organizations   when it comes to PDF remediation. Number one  is the costs. You know, multiple vendors,   we've seen outsource providers, and  also investment into internal personnel,   is very costly for PDF remediation.  It can really be a runaway train of   cost because the industry standard price per  page can be exponential, and the manual work   at this capacity and scale is obviously very  expensive. With that is the quality, you know,   because of the cumbersome manual processes and  auto-tagging mishaps and issues with quality,   multiplied by, you know, the volumes of pages  and scope and complexity. It really leaves   these organizations exposed to non-compliance  with Section 508 and, you know, unfortunately,   lawsuits in some of these cases because of  these elements. And lastly, the speed, you   know, with a lot of the use cases we've seen in  healthcare organizations, you know, the demanding   legal requirements do require quick turnaround  times to get this information out to customers   consistently, but it's not really realistic to  accommodate with traditional manual processes   and the auto-tagging methods because of the  volume to manage when coupled with the quality. So, we, you know, obviously wanted to create  a better way to solve for these challenges,   and we did as mentioned with our, you know,  solution Equidox AI. Our experts have found   a way to truly automate the PDF accessibility  process for many of these healthcare use cases   involving high volumes. Equidox AI automation  allows good quality, usability, and compliance   every single time because of our unique model  creation. We don't auto-tag or cut corners or rely   on human elements to get the process done, you  know, training these models with our developers,   our data scientists, and our accessibility experts  to ensure that quality is key to the solution. You know, our Equidox AI automation also  accommodates aggressive timelines because   we're relying on technology. We can dictate  how fast the solution runs and turn the dial   up or down, so to speak, to accommodate these  timelines that may be required. And lastly,   Equidox AI automation allows for lower costs,  which everybody's looking for. You know,   to the bottom line to, you know, efficiently  ensure that they're compliant, efficiently,   you know, make the documents accessible but while  keeping costs down. And you know, we increase the   process improvements because of this and are  also able to consolidate, you know, vendor   relationships because of this automation. So with that introduction, I'm going to  pass it over to my colleague Dan Tuleta,   who is going to talk about, you know, why  it's important to make PDFs accessible and   a little background about that, and then you know,   just how our Equidox AI works in more  detail with a demonstration. Thank you. Dan? [Dan Tuleta] All right, thank you, Paul. So, hi  everyone. I assume that most people on this call   are familiar with accessibility laws such as  the ADA or Section 508 or the WCAG guidelines.   Since I'm not a lawyer, I'm not going to go  into the details and the legal of these laws,   but at a high level, just like there are  requirements to provide physical access,   like wheelchair ramps or elevators or Braille  signage, organizations also need to ensure that   their public-facing digital content, including  their PDF documents, is accessible to everyone,   including people with disabilities. Ignoring the  accessibility of your digital content opens up   your organization to legal risks, and there have  been thousands of organizations who learned this   the hard way when they were sued for exactly  this type of problem, and there are thousands   more who quietly pay out large settlements, and  yet they still have to circle back and fix their   accessibility deficiencies after that payout.  Long story short, we live in a very digital world,   and we rely so heavily on digital information,  so digital accessibility is not a fad,   and it's not going away, so it is always good to  be aware of it and address it in a proactive way. For anyone who is unsure of why this  digital accessibility stuff matters,   people with disabilities use various types of  assistive technologies to interact with digital   content, like PDF documents. A very common type  of assistive technology is called a screen reader,   which is capable of reading digital  content like websites, applications,   and documents. Screen readers use what are called  digital tags to navigate documents. The tags need   to be properly encoded into a document to organize  the content and make it compatible with the screen   reader. Think of the tags as a framework of the  document, which gives the screen reader user   the ability to navigate and interact with all of  the various elements within that PDF. Equidox, in   cooperation with the National Federation of the  Blind, surveyed about 250 blind and low-vision   individuals who rely on screen readers to interact  with PDFs every day. Based on this survey,   we found that at least two-thirds of PDF  documents are inaccessible to people with   disabilities. So if you just take a moment to  put yourself in the shoes of a blind person,   you can quickly imagine how frustrated  you would become if you could not read   two-thirds of the documents that you came into  contact with on a daily basis. On top of that,   imagine the privacy issues if you have  to go and ask your neighbor or a friend   to help you read private documents like a  banking or investment statement, an invoice,   a pay stub, or health insurance policy documents  or medical test results. It could be embarrassing,   and it's just a breach of privacy that  no one should have to live through. To further emphasize the points I  was making a couple of slides ago,   here's some additional information about  the volumes and the types of lawsuits that   organizations have faced and will continue  to face moving forward. Just to reiterate,   the digital accessibility requirements that  organizations must adhere to are not going away,   and there will continue to be an increase  in the attention that is paid to it by state   and federal mandates. The DOJ, disability  advocacy groups, and individuals who simply   want to be able to access their critical  information are going to continuously put   more and more pressure on organizations to  adhere to these accessibility standards. One of the main challenges with PDF accessibility  is that each PDF document is unique. We have heard   a lot of empty promises of fully automating PDF  accessibility, but there are so many things about   PDFs that require human interpretation to decide  how to tag specific elements within the content.   I have been working in the PDF accessibility  market for over seven years now, and I've   seen a lot of organizations who come to us, but  they assume that they have accessible documents   because their documents have some tags, but they  quickly learn that these tags are not necessarily   usable or compliant, and they are still open to  litigation. So please be aware of this concept of   autotagging. Auto-tagging technology is really  just masking or masking a solution to fully   automate PDF accessibility, but these auto-taggers  are capable of putting tags on a page,   but there's always going to be accuracy issues  with them, and the inaccuracy of those tags can   lead to a ton of confusion and frustration  for a screen reader user. Additionally,   auto-taggers can and will leave organizations  open to further litigation because there is no   guarantee of compliance with WCAG standards.  So even paying to outsource your huge batch   of documents to companies who auto-tag them, you  are not mitigating your risk of litigation because   auto-tagging simply falls short of true compliance  with accessibility standards. Oh, excuse me,   I hit the wrong button. The alternative of  outsourcing the remediation work to third parties   who are almost exclusively located overseas also  introduces a mountain of data privacy issues,   and even if you can work around that, the sheer  volume is impossible to keep up with. These   outsourced remediation providers will cut corners  and do the absolute bare minimum amount of work   to make the document pass an automated checker,  but they are not actually making the documents   compliant because it simply takes too long to meet  the deadlines when you're dealing with that type   of volume. Incorporating artificial intelligence,  more specifically computer vision and machine   learning, into high-volume PDF remediation  allows our Equidox accessibility expert   to train an AI model to accurately identify and  tag all of the elements in the document template,   and the use of AI developed by our data  scientists paired with that human element   of our trained accessibility experts allows for  incredibly accurate, usable, and compliant PDFs   to be returned to the customer in a fraction  of the time because AI works exponentially   faster than humans manually tagging each  page. AI doesn't need to take vacations,   it can work 24/7, 365 without breaks, and  also AI doesn't need to cut corners to meet   a deadline. And of course, we're going well  beyond just a simple autotagging process. So, how does any of this work? Well, our  accessibility experts use example documents   of customer templates to properly identify the  various elements on the page. These elements might   include text or paragraph structure, various  levels of headings, lists and tables, graphs,   and images, in the reading order of the content,  which is very, very important. This training data   is then fed to an AI model to apply what it has  learned en masse to many thousands or millions of   pages that have similar templates and formatting.  Although the mechanics of how this AI technology   works is a little bit abstract and much more  complicated than what I'm capable of showing   you on a simple PowerPoint slide, here are just  a few high-level examples of how we can visualize   the AI at work. In this scatter plot, each of the  green dots represents a page in a PDF document. If   you can see, they are grouped together based on  similarities that the computer vision finds. So,   for example, this cluster will contain  all of the pages containing pie charts.   In this example, you can see there are different  multicolumn text layouts that the AI will also   use to recognize different pages and group  them together appropriately. The AI will also   pick up on font styles, sizes, and colors  to help it establish accurate tags on a   page. We can even train the AI to identify  the many potential variations in tables,   which are extremely complicated to tag accurately,  so the different numbers of rows and columns,   table headers versus table data, and even tables  of varying sizes that might span across multiple   pages. All of this can be handled through the  AI. The result of all of this extensive document   analysis and training of the model, and feeding  the data to the AI, is creating a fully compliant   PDF document without any human remediators, who  are both expensive to employ or outsource to,   and they are also, of course, liable to make  errors or be forced to cut corners just to be   able to meet an unattainable deadline due to  the volume demands. We are also reaching full   compliance because this is not autotaggers  throwing sloppy tags on a page and saying   "quote-unquote good enough." And then, of course,  beyond compliance and passing automated checkers,   the bonus of using AI for high-volume,  hyper-fast remediation is that it will   produce incredibly accurate and very usable  documents for people with disabilities,   so your customers who rely on assistive technology  will not be filing complaints or lawsuits or   calling your headquarters to complain about  documents that they can't navigate or understand. So we're just about ready to jump into  a demonstration of this AI technology,   but before we do, I just want to make it clear  that this underlying technology can be deployed in   several ways to align with your organization's  requirements. We have built an interface,   which you will see during the demonstration, that  could allow theoretically your employees to run   this process from start to finish with just a few  clicks by uploading the document or documents,   running the batch, and then downloading the  finished PDF or multiple PDFs. We can also embed   the AI models into an existing document creation  and delivery system through the use of APIs.   This could be critical for customers needing  to download private documents, like a monthly   statement or an explanation of benefits or  medical test results or something along those   lines. Lastly, Equidox can operate the entire  process on your behalf as a managed service,   so we can take care of the remediation as well  as the validation of that remediation to ensure   that everything exceeds all accessibility  requirements, and then also deliver that   fully compliant PDF back to your  organization to be posted and distributed. Just one more thing to note: Equidox AI is tagging  the PDFs at the post-processing stage, which you   will see during the demonstration, and what I mean  by that is that these PDFs are already created,   and we are applying the accessibility  as a final step before they are publicly   distributed. The advantage of tagging PDFs at  the post-processing stage is that we do not have   to disrupt or completely rebuild your document  creation process. Your designers and producers   of mass documentation can continue their process,  and we will handle the accessibility component at   the very end of the creation stage, but right  before that document reaches your customer. switch over to the Equidox batch  interface. We're going to do a   little demonstration of how this technology works. So, as I said, we do have this small interface  built for demonstration purposes. But in theory,   this type of technology could be deployed  to users within your own organization,   who would be able to use an interface like  this, perhaps slightly customized for you,   to manage this workflow internally. We can, of  course, also do this as a managed service for you. Now, what I'm going to do to get started is I'm  going to go to the 'Upload Documents' tab. This   will take me to this view here, where I can  then open up the folders on my hard drive.   I'm going to just grab this document that  we want to run through the process today.   When I drag and drop that document into this  rectangle, a blue 'Upload' button will appear.   I'm just going to give this document a  few seconds to upload. When it uploads,   it'll then be ready. It's going to be in the  system, ready to go for the batch process. My next step is I'm going to go to the  'Create and Run Batch' tab. From this view,   we have this dropdown menu where I can  select a model that I want to apply for   this batch. I'm going to select this model  right here. Once I select that model,   another set of options appears, and this  is the document that I just uploaded. So,   I'm going to select this document, and  I'm simply going to hit 'Run Batch.' Now,   when I hit 'Run Batch,' it's going to populate  in this list, and we see some different dots   here to kind of indicate which stage it is in  during the process. We have some other details   about the document right up here, showing  us like how many pages, how many documents,   how much time it's taking, and so on. Down  here, the different stages of the remediation. Right now, Equidox is analyzing the  document and applying the ML zones.   The computer vision and machine learning  are working in the background to analyze   all of these various pages and accurately tag  the content. Once it finishes getting tagged,   it will then go to the export engine. We're  basically going to rebuild the PDF and create   an exact replica version of that document with  all of the tags properly organized and applied   to the content. Once it finishes exporting,  I'll be able to download that completed file,   and that document will be fully tagged, fully  compliant, fully usable for screen reader users. Now, while that is running, I'm going to  open up the document that we just uploaded   just so we can take a look at it. If I look at  this document in my default PDF viewer here,   Adobe Acrobat, this document is completely  untagged. There's no tag structure at all.   If we were to run accessibility  checkers across this document,   virtually everything would light up as an error,  because there are no tags for an accessibility   checker to check for, so it assumes that  everything is wrong because it is wrong.   You cannot have an untagged document. This is  completely unusable to a screen reader user. If we just take a quick look at some  of the content in this document,   this document is a 100-page  excerpt of a much larger document,   but it is a physician's directory. This  would be for a healthcare insurance type   of company that is going to provide a PDF  document with all of the hospitals, doctors,   physicians, physical therapists, dentists,  and everyone in their network that you can   use as part of your healthcare coverage. These  healthcare providers must provide this in a PDF   format—it's the law—but the challenge is that  these documents are constantly being updated,   often being updated every single month. They are  typically upwards of 1,000 to 5,000 pages long.   Another challenge is that many providers are in  markets where they have to provide this document   in multiple languages, not just English but  also Spanish, Mandarin Chinese, or Vietnamese. Given the complexity and frequency of updates,  any team of humans would struggle to keep up   with the volume demand and the constant flow  of these documents as they get updated monthly.   Auto-tagging technology is not reliable  for these complex documents, as there is   a very specific heading structure, reading  order, and variances across these pages.   This makes it impossible to rely solely on  auto-tagging technology to do this accurately. So, when we go back to the batch interface and  see that the process has completed, we'll open   up that new document and see a dramatic change—the  document will visually look the same, but all of   the content will be tagged accurately. This new  document is completed and ready to download. If I just hit the 'Completed File Download'  button, I'll download this one, put it on my   desktop, and then open it in our app. Here's  the new version of the document. If I go to   my tab for the accessibility tags, you can  see the big difference. All of this content   is now tagged. If you start to tab through it,  you'll see that we have the heading level one,   then the heading level two, heading  level three, heading level four,   and heading level five, with the subsequent  information about that specific area of Iowa. As you tab through, you can see that  all of the content is grouped together   appropriately and read in the way  it should be. This is critical for   the usability of this document. I could, of  course, tab through all 100 pages of this,   but what you'll see is more of the same  accurate, consistent tag structure being   applied to many pages without a single human  being having to be involved in the process. This would allow this particular healthcare  company to manage literally hundreds of thousands,   if not millions, of pages of documents just  like this throughout the course of the year,   without having to outsource it or  rely on auto-tagging technology. This technology can be deployed in a number  of ways: It can be inserted into your document   creation process, we can run it for you as  a managed service, or potentially you could   deploy an application like this using this type  of interface to manage the service yourself. Just keep that in mind. Now I'm going  to jump back to the slide deck here,   just to wrap things up, because we're  just about out of time. As Paul said,   this is being recorded, and we will insert  all of the information about that demo,   as well as the slide deck to be shared with  you. Here are some links if you'd like to   learn more about digital accessibility  and how it relates to your PDF documents. In conclusion, I just want to thank everyone for  joining us here today. We hope that you can see   the value and capabilities of this new technology.  Please do not hesitate to reach out to us for more   of a one-on-one type of consultation, so that  we can discuss your organization's unique use   case and how Equidox AI can be applied to  them. Also, we will be sending a recording,   so feel free to share this with anyone in  your organization who might benefit from it,   along with a link to the deck. If anyone  asked a question using the Q&A feature,   we will get back to them with  a response as soon as possible. Lastly, there's going to be a short survey  sent out, so if you don't mind just taking   a moment to fill it out, we would greatly  appreciate it. Thank you again for joining,   and have a great rest of your day. For more  information about how Equidox Software Company   can help you with PDF accessibility,  email us at EquidoxSales@Equidox.co   or give us a call at 216-529-3030, or  visit our website at www.Equidox.co.