Fully Automated PDF Accessibility for Financial Services

[♫ ♬ Music ♫ ♬] [Paul Campbell] Hello everyone and welcome   to our webinar for Equidox AI, a fully automated  PDF remediation solution. We're very excited about   this cutting-edge technology that is solving  for challenges within the financial services   sector. By way of introduction, my name is Paul  Campbell and I will be joined by Tim Needles   and Dan Tuleta for the next 30 minutes. Just  some logistics for today: the live audio Q&A,   as I mentioned a little earlier, is not available  for this webinar. However, you can drop questions   in the Q&A chat button at the bottom of your  screen and we will get back to you after the   webinar with any answers to questions that  you post. Additionally, this webinar will be   recorded and will be sent after the meeting, in  addition to the deck and a short survey. We're   happy to do a more direct interactive session  with you and other team members if they're not   available to join here today. And we would love to  further our conversations as to how Equidox AI may   be a fit for your organization specifically as a  follow-up. Overview of the agenda for today, first   Tim Needles our President and CEO has joined us  to talk a little bit about who is Equidox Software   Company, who we are where, we've been, and where  we are going. I will then take over to discuss   the challenges that we've seen in the financial  services sector specifically and our solution to   the problems we've seen. Then Dan will talk about  why do we make PDFs accessible and what's really   driving this, followed by an overview of Equidox  AI and how it works. And lastly the last half of   the meeting a demonstration of our solution. With  that said I'm going hand it over to Tim Needles   to give an introduction to Equidox AI. Tim? [Tim Needles] Oh, thank you, Paul, for the   introduction. Much appreciated and thank you  all for joining us here today. We're exceedingly   excited that you're here with us while we showcase  our AI solution. Now we know that many of you have   been feeling the pain of PDF remediation  for some time in terms of excessive cost,   compliance issues, and risk mitigation issues from  lawsuits. And you'll be pleased to know that there   is a better way… a better solution… And after the  demo, I think you'll agree that Equidox AI is a   superior solution and can solve your problems as  it relates to PDF remediation. Some of you might   think we're the new kid on the block, but Equidox  has actually been in existence for over a decade.   Way back when a Canadian citizen was trying  to apply for a government job posting on the   internet, but she was unable to do so because of  her visual disability. So she sued the government   and she won her case. The government sought  out a solution but they couldn't find one,   so they asked the marketplace to respond. Hence,  we started building a solution and haven't stopped   innovating since. Over a decade ago we created  a robust SaaS solution that now hundreds of   customers around the world are using. The solution  is truly world-class and is adding tremendous   value to the marketplace of digital accessibility  for enterprises, government, and educational   institutions, and certainly financial services.  Our customers love the product, evidenced by   the fact that nearly 100% of our customers renew  their subscriptions every year. Well, that's all   well and good for the legacy SaaS product, but  we started hearing years ago from organizations   like yours that had tens of thousands, or hundreds  of thousands, or even millions of documents that   needed to be remediated even on a monthly basis  that there really wasn't an automated solution   for that. The traditional service providers were  sending the documents to India and other countries   but there was really no efficient way to scale in  that model. So companies were forced to settle for   a solution that frankly is really time-consuming,  expensive, and doesn't truly mitigate the risk   of a lawsuit for your company. So our demo  today will prove that there is a better way   to remediate large amounts of of documents,  saving you a bunch of time, a lot of money,   and all the hassles of old-school remediation.  And, you know, it's great to accomplish these   benefits and we're really proud of what we've  done and achieved. But we're really most pleased   to provide a game-changing solution that assists  the visually disabled. We're really passionate   about helping these people and it's invigorating  to work with organizations like yours that share   our commitment and passion. So thank you again for  taking the time to be with us today and we look   forward to serving you. Paul back to you. [Paul Campbell] Thank you, Tim.  [Tim Needles} Welcome. [Paul Campbell]  So as Tim mentioned, Equidox AI is a fully  automated PDF remediation solution that   removes the traditional manual remediation and the  auto-tagging methods, while still increasing the   quality, accuracy, and compliance that we're  all looking for. And we've really found that   the financial services sector can get great value  out of Equidox AI. This is because of number one,   unfortunately, the rising lawsuits in the industry  under the ADA and AODA in Canada and the need for   end users with visual disabilities to access  important information from their financial   institutions as an equivalent experience as  sighted users. Number two, large quantities   and sheer volume of templated and reoccurring  documents that consistently need to be accessible   because of these compliance requirements. And  then number three, the challenges that these   organizations have with current processes quality,  speed, and overall vendor sprawl because of the   large volumes which we'll expand upon here in a  moment. So many customers are doing something but   unfortunately are still exposed or getting sued  because of the volumes that they have to manage.   It's a very challenging thing to overcome. Some  of the larger use cases we see in the financial   institutions may include, but are not limited  to, any type of reoccurring monthly quarterly   customer statements, fund offerings, portfolio  insights… These are all great candidates for   Equidox AI because of the high volume, repetitive,  and templated nature. Now the main challenges for   financial services organizations when it comes  to PDF remediation that we have found are:   number one, costs. Multiple vendors, outsourced  providers, and overall investment in internal   personnel is very costly. It can be a runaway  train of cost because of the industry standard.   Price per page is exponential and manual  work to this capacity and scale can be very   expensive. Quality, number two, because of  the cumbersome manual processes and corners   being cut to manipulate checkers the auto-tagging  mishaps and issues, and then multiplied by volumes   of pages in scope and complexity really leaves  organizations exposed to non-compliance lawsuits   because of these elements. And number three,  speed. Demanding legal requirements and quick   turnaround times to get accessible information  to your customers consistently is really not   realistic to accommodate with traditional manual  processes and the auto-tagging methods because of   the volumes to manage when coupled with quality.  And, as Tim mentioned, we wanted to create a   better way and we saw this gap in the market.  And Equidox AI solves for these challenges. Our   experts have found a way to truly automate the PDF  accessibility process for many of the financial   sector use cases involving high volumes.  Equidox AI automation allows good quality,   usability, and compliance every time because of  our unique model creation. We don't auto-tag,   or cut corners with checkers, or rely on the  human element to get the process done. Equidox AI   automation also accommodates aggressive timelines  because of relying on technology. We can dictate   how fast the solution runs and turn the dial up  or down so to speak to accommodate timelines that   that may be required for the customer. And lastly,  Equidox AI automation allows for lower costs,   process improvements, and vendor consolidation. So  with that, I'm going to turn over to my colleague   Dan to talk about why it's important to make  PDFs accessible, and how Equidox AI works, and   then lastly a demonstration of our solution. Dan? [Dan Tuleta] Great thanks, Paul. So hi everyone.   And I assume everyone on this call is probably at  least somewhat familiar with various accessibility   laws like the ADA, Section 508, and the Affordable  Care Act. Now I'm not a lawyer so I'm not going   to go into all the details of how these laws  work, but at a high level, just as there are   requirements for organizations to provide  physical access such as wheelchair ramps,   elevators, Braille signage… Organizations need to  ensure that their public-facing digital content,   including PDFs, is accessible to everyone,  including people with disabilities. Ignoring   the accessibility of your digital content opens  up your organization to legal risk. There have   been thousands of organizations who learned  that lesson the hard way when they were sued   for exactly that type of problem. And there are  thousands more who quietly pay large settlements   out of court and yet they still have to go  back and fix their accessibility issues after   the fact. So 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, it is not going anywhere, it's not going   away anytime soon. So it's always good to be  aware of it and address it in a proactive way.   For anyone who isn't sure of why we are making  why this digital accessibility stuff matters,   people with disabilities use various types of  assistive technologies to interact with digital   content like PDFs. A very common type of assistive  technology is called a screen reader, which is   capable of reading digital content like websites,  applications, and of course PDF documents. Screen   readers use digital tags to navigate documents and  tags need to be properly encoded into a document   to organize the content and make it compatible  with the screen reader. So think of tags as the   framework of the document which gives the screen  reader the ability to navigate and interact with   all of the various elements in 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 every day   to interact with PDFs. Based on this survey, we  found that at least two-thirds of the documents   that they interact with are inaccessible to  people people with disabilities. If you put   yourself in the shoes of a blind person, you can  quickly imagine how frustrated you would be if you   could not read two-thirds of the documents that  you came in contact with on a daily basis. And, on   top of that, imagine the potential privacy issues  if you have to go ask your neighbor to read to you   your monthly bank statement, or your credit card  charges, or your investment portfolio summary.   Whatever that document might be. Obviously,  you want to keep that information confidential   and private. So just to to further emphasize the  points that I was just making a couple of slides   back, there's some additional information here  about the volume and the types of lawsuits that   the financial sector has faced and will continue  to face moving forward unless they start doing   something about it. So just to reiterate,  the digital accessibility requirements that   organizations must adhere to are not going away  and they will continue to increase in attention   will be paid by state and federal mandates, the  Department of Justice, disability advocacy groups,   and individuals who simply want to be able to  access their critical information. So let's talk   a little bit about the human element of AI. One  of the main challenges around PDF accessibility   is that each PDF document is unique. We've 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 the specific elements within the  content. I've been working with PDF in the PDF   accessibility market for about seven years and I  have seen a lot of organizations and I've talked   to a lot of them that have assumed that they are  accessible… they have accessible documents because   their documents have some tags in them. But  they quickly learn that those tag tags and those   documents are not usable nor are they compliant  and their organization is still wide open for   litigation. So just be aware of any organizations  out there talking about auto-tagging technology.   Auto-tagging is really just masking itself as a  solution to fully automate PDF accessibility. But   auto-taggers are capable of simply putting tags on  a page. There will always be accuracy issues and   the inaccuracy of those tags will lead to a lot of  confusion and frustration for that screen reader   user. Additionally, auto-taggers can and will  leave that organization open to further litigation   because there is no guarantee of compliance with  any of the web accessibility standards. So if and   if you pay to outsource your huge batches  of documents to auto-taggers, you are not   mitigating your risk of litigation because these  auto-taggers again fall short of true compliance   with the standards. And then the alternative of  outsourcing the remediation work to third parties   who are almost exclusively located overseas  introduces a mountain of data privacy issues.   And even if you can work around that, the volume  of documents is impossible to keep up with. These   outsourced remediation providers will cut corners  and do the bare minimum amount of work to make a   document pass an accessibility checker, but not  actually make the document compliant because it   simply takes too long to meet customer deadlines  at this type of value. So incorporating artificial   intelligence, more specifically, computer  vision and machine learning into high-volume PDF   remediation allows our accessibility experts to  train an AI model to accurately identify and tag   all of the elements in the document template. The  use of AI developed by our data scientists, paired   with the human element of trained accessibility  experts that we have here on staff allows for   incredibly accurate, usable, and compliant PDFs  to be returned to the customer in a fraction of   the time because the AI works exponentially faster  than humans manually tagging each page. AI doesn't   need to take vacations, AI can work 24/7, 365  on demand without any breaks. And, of course,   AI doesn't need to cut corners to meet deadlines.  It can do it the right way the first time. So how   does all of this work? So our accessibility  experts use example documents of customer   templates to properly identify the various  elements on the page. These elements might include   text and paragraph structure, various levels  of headings, lists, tables, graphs and images,   and of course, the very important reading order of  the content. This training data is then fed to the   AI models to apply what it has learned en masse  to many thousands, tens of thousands, millions   of pages that have simpler similar templates and  formatting. So although the mechanics of how AI   technology works is rather abstract and a lot more  complex than what I'm capable of showing you on a   simple PowerPoint slide, there's a few examples  here of how we can visualize AI at work in this   scatter plot that you see on the slide. Each of  the green dots represents a page within a PDF.   They are grouped together based on similarities  that the computer vision finds. So this cluster   will contain all of the pages that contain pie  charts. So you can keep an eye on the pages   here that are all going to contain pie charts.  Obviously, these are all going to be kind of   in the same neighborhood that the AI is capable of  locating and identifying. In this example, you can   see there are different multicolumn text layouts  that the AI will use to recognize different pages   and then group them together appropriately. The  AI will also pick up on things like font styles,   and size, and color, and background colors to help  it establish the tags on a page. We can even train   the AI to identify many potential variations in  tables such as the numbers of columns and rows,   table headers versus table data, and even tables  of different sizes that might span across multiple   pages. The result of all of this extensive  document analysis in feeding the training   data to the AI is creating a fully compliant PDF  without any human remediators who are expensive to   employ onshore or outsource offshore. And they are  of course liable to make errors or be forced to   cut corners just to meet an unattainable deadline  due to the volume demands of your organization. We   are also reaching full compliance because this is  not the auto-tagging method where we're throwing   sloppy tags on a page and saying good enough. So  beyond compliance and passing automated checkers,   the bonus of using AI for high volume and  hyper-fast remediation is that it will   produce incredibly accurate and very much usable  documents for people with disabilities. So your   customers who rely on assistive technology  will not be filing complaints or lawsuits,   they won't be calling your headquarters to  complain about a document that they can't   navigate or understand. So we're getting ready  here to jump into the demonstration of how the   technology works, but 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 I'll show you during the demo, that can  allow your employees to run the process from start   to finish with just a few clicks by uploading  the document or documents, running the batch,   and then downloading the finished PDFs. We can  also embed AI models into an existing document   creation and delivery system through the use of  APIs. This would be critical for customers needing   to download private documents like a monthly  statement, or an explanation of their portfolio   summary, or a credit card bill, whatever the use  case. A document that is supposed to be private   and confidential can be directly delivered  to that user. Lastly, another method is that   Equidox could operate the process as a managed  service. So we can take care of the remediation   and validation and get everything back to you in  a fully compliant way to be posted and distributed   publicly if that's the use case that you have in  mind. So there's different methods to implement   this technology. So just keep that in mind during  the demonstration. So just one more thing to note:   Equidox AI is applying the PDF tags at the  post-processing stage which you will see during   the demonstration. So 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 are not having to  disrupt, or more than likely completely rebuild,   your document creation process. So your designers  and your producers and whatever systems you have   in place of mass document creation can continue  and we will handle the accessibility component   at the very end right before the document reaches  your customer. Okay so I'm going to jump out of   the slide deck, and for anyone that wants to share  this webinar later there will be a video of course   with our demo attached for this slide. But I'm  going to jump into our actual Equidox AI batch   interface. So what I'm going to do is I'm going to  demonstrate the fully automated remediation of a   set of bank statements. So bank statements are a  good candidate for Equidox AI technology because   in general, they will follow a very similar  formatting and structure. But the data within   that structure will vary from version to version.  So for example, if I just open up an example of   what we're going to be talking about, people's  names, and account numbers, and mailing addresses   will always be different. But the location on  the page will be very consistent. Line items   on the statement might vary. So for example,  one customer might just have a single charge   on their credit card for a month whereas another  customer might have 500 charges requiring their   statement to merge onto four or five different  pages of line items. So there will always be   enough variants from one version to the next that  will make old-fashioned auto-tagging inaccurate   and unreliable. And of course, you cannot  outsource the remediation of confidential data   to third-party companies located offshore. Even  if you could the volume of documents like this,   that you're producing, would be impossible for  them to handle. So as I was mentioning before,   for a use case like this, a recurring monthly  statement, or an invoice, or a portfolio summary,   it is more than likely that your organization  would want to implement this accessibility of   this at the final step in the process so that  your customer is automatically receiving their   accessible document without any human involvement.  And we can of course accommodate this through the   APIs. So visualize a scenario where that is not  me pressing the buttons to operate the system,   but rather this is directly integrated into  your document creation process to achieve   true automation. So what I'm going to do is I'm  going to start by uploading the documents of   the the bank statements that I just showed you.  So I'm going to go to the Upload Documents tab,   I will open up my folders, and I will just grab  this .zip folder which contains the documents.   Once I press the Upload button it will just take  a few seconds to upload to the cloud. And I will   then go to the Create and Run Batch tab here on  the interface. From this tab, I have a drop-down   menu where I can select the appropriate model that  I want to apply to this batch of documents. Now we   have other demonstrations that we can provide for  different types of use cases depending on the type   of use case you have in mind. In this example,  I'm going to be doing that bank statement. So   I'm going to choose the Statement model. I'm then  going to choose the .Zip radio button here to then   select the .Zip folder that I just uploaded. Now  once I've gone through those three quick steps,   I simply press Run Batch. Now that the batch is  running, I can see sort of the steps that are   of the document in this process. So take note of  these various dots that represent the stages of   the process. So Equidox is analyzing each document  and automatically applying the correct zones or   tags around the content. Once the document is  finished at the AI step, they will go through the   export engine and become available for download or  automatic direct delivery to your customer. So you   can see these green dots kind of processing across  the page. Now just going back quickly to the   original document, I just want to make note just  really quickly here that this document coming from   the original folder is completely untagged. So  there's no tags at all in this document. It would   render this entire PDF essentially useless. A  screen reader user would have all of this content   just being read in a totally random order. The  tables themselves would be impossible to navigate   or understand. So this entire bank statement is  completely inaccessible. And of course very much   out of compliance with all relevant accessibility  standards. So we'll go back to the batch interface   here and the process is nearly done running. In  this current version of Equidox, in our little   demonstration environment, this batch takes  roughly 80 or 85 seconds to finish running. And   when it when it does finish, you can see that all  of these documents have gone through the process   and there were 20 of them. So 20 documents have  been fully remediated and exported. And if you   were to download any of these and then open them  up in the Adobe Acrobat view we can take a look   at the tag structure… if Acrobat will wake up… I  guess I'll save this and we'll open this one up.   And now when we take a look at it you can see that  these documents are fully tagged and furthermore,   they are accurately tagged. So you'll notice that  set the reading order, as I start to tab through,   is in a logical logical order. This image, the  figure which is the bank logo, is tagged with   alt text which we have built into the AI model.  The heading structure and the tables are fully   tagged. So all of this is 100% accurate, 100%  compliant, and 100% usable for your customer. So   they can freely navigate all of this information  all of these various charges and line items and   debits credits on their statement. And they are  free to interact with all of this information.   So without anyone having to manually go through  this document and set up all of these table tags,   and heading structure, and reading order, which  can take a very long time in the old-fashioned   manual way, we're able to deliver a fully  usable, fully accessible document to your   customer. So with that said, that is a quick  crash course through the actual demonstration.   Now I'm going to jump back into the slide  deck here just to wrap things up because   we are quickly running out of time. So as Paul  mentioned in the beginning, this slide deck will   be shared with you and there are some links here  to learn a bit more about digital accessibility   and how it relates to PDF documents. And then  in conclusion, I just want to thank everyone   for joining us here today and we hope that you  all see some value in the capabilities of this   technology. So please don't hesitate to reach out  to us for a more one-on-one consultation so that   we can discuss your organization's use cases that  you have in mind and talk about how Equidox AI can   be applied to solve for those challenges  that you have around PDF remediation. Also we   will be sending the recording of this webinar  so feel free to share this with anyone in your   organization who you feel would benefit from it.  And we will include a link to the slide deck. And,   for anyone who asked a question during the Q&A  through the chat we will get back to you as soon   as we can with a with a response to that question,  And lastly, there will be a short survey so if you   don't mind just taking a quick moment to fill  out that survey we would greatly appreciate it.   So thank you again, everyone, 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.