Sign Utah Claim Secure

Document type sign claim utah secure

you my name is Jared Ellis and I worked with an organization called my other disability I department the organization and I'm also our operations officer Milo disability is an organization that is asian white and we represent individuals before Social Security Administration in disability actions for Medicare actions to get reimbursement stuff for hospitals for medical expenses and they've occurred after to get a church is to Medicaid or Medicare and we have one central offices in Utah and then we have regional office long the country as her nationwide we handle claims in all 50 states and how kind of unique model in that our attorneys all work out of their houses and their staffs romantic sex release all the mail comes into the location the statute the various attorneys are all Amanda central in one location and so the attorneys can work off of our ski area and they can all interface with that directly in terms of the amount of mail we get we process really high plains of them ten thousand pieces of melody a two-vehicle process and as I indicated earlier would work in all 50 states and the program itself is administered through the states and so that would mean the one document would look different depending on the state of the game from though will be the same document and so it gave us a great deal of variety to our documents because the same document could conceivably have 50 different variations so our workflow is determined by document type so as a certain document idea whether that would be a document to denial over whether that's a retaining packet where we signed up a particular client or our services how they classify the document and determine where it goes which staff is gonna be working on the case and then what they should be doing on the case based on the type of document they received and so we had five hundred or so variations on the docks and subtle differences could you know make a big difference in terms will happen workflow and so being monogamous plus by appropriately was very important tourism we at the time would train a few classifiers so we would have people that we would these a multitude documents and it would take literally months to train somebody in all the different variations of documents and if they got that wrong of course of credential difficulty for us and then we would also have to concern and worry that this person might go somewhere maybe they move out of state or they find it in her job and so if we did this summer he was trained in the reading of the documents and classifying the documents those kind of a different business so about two thousand to two thousand documents process to month 500 different types of documents created the problem as I indicated with the training to some examples by way of complexity issues certain documents would be the same no vary based on works 9 for example and so we had a representation document that might be signing the primary representatives name that would be what variation to document and on-site document we might have a document that was signed by a secondary representative with one sign ago client or of the document website like the client could possibly in the wrong spot or we have a document the Sun like a client but wasn't date where it was dated in the wrong spot and so for each of these different document types we would have a different upload option that would create a different workflow based on the document and so against ACTA complexity was that the same document get help Oliver types of variations I was the same thing with our fee agreements and so we could have a quick signature we could have an in science document by a certain representative we get a woman was to correctly and then we would need different fee agreements at different times rate processes both the case became contested and we have to a judge well that's an entropy dream and if the jets that denies it in the next level again that's another favorite even the documents basically the same but you have seven or any different iterations based on the process and so these types of issues essentially added complexity or difficulty it was a long training time as I indicated to get them to speed on things and then it was also a lot of two with the work because we would have it assigned by one person blindly and because it was so critical we've got a ride it would go out with a staff on the case and we would all spoke them classify the document lightly and if we got match the new personnel proper classification if we got a mis manager person within any arbitrary determine property classification to that Nina so all scholars will be relative is a business how we are the automated entirely paperless and so we're constantly looking for solutions like which Koufax that products would offer a war aware of OCR endocr reading software's and so had looked a couple of years back into trying to automate some of this because obviously was the volumes of melville we have it's great cumbersome so contact a group that did go see our they came in and look at all our documents and essentially told us that our problem was too complex that we had too many different document types with the 50 iterations of each other in different states that it just wasn't anything that the appellant if you do and so we looked at everything and that was kind of the verdict that we just couldn't field so I was disappointed by that but we put that on and continue to what we're doing for a couple years until I was put in touch with the Michael who came in contact with me through a mutual acquaintance and he told us talking about what he does and said he thought maybe he felt with some of the issues we had in processing your documents and I told him that I'm already gone and I know that he's eager and excited thinking maybe we're at a client but that in the bottom analysis he would determine himself that our document for too complex to be done and so he said we'll give me a shot so we met up and spent some time together and I gave him a bunch of sample documents and it was nice I mean how would you read these and he said well you haven't even though God for it is I said well I will appear in theaters people us when we do the software in the same thing and so you know explains that it's more than just a template approach I've got this actually really continent so I think the solution what was crucial us was really partnered with the right people but I was personally skeptical we would be successful just because of how expensive are venomous for how how many variations we had so we had eating spoiled my colon and discussed this he came back with a working prototype we could run our document samples through yeah look as though on the pulmonary past it was successful and then he brought his team in and we got to work on the project well it turned out the project was in fact than we originally contemplated and his group working with our group really through the collaboration I was successful coming up with a lot of unique solutions design so kind of this agilent collaborative approach that they brought to our business to come in and work one-on-one with our staffers explain and get things as correctly as possible so slow look at the outset with time and work in collaboration from both sides we started staying with a successful applicant so some of the things that were tricky being traded for distant place where the dart group use different terms than the people Koufax we use things like manual processing batch glasses extraction online learning validation operations local terms we use internally that are quote Xtina used as well and those words meant different things together they've been to us let's go over some night any confusion that regard just going with I see returns another lesson to learn was that we had Tennessee at the outset to focus on the documents that were the most easily readable by Koufax and so of koufax they come from it easily will announce where we're going to start but it wasn't necessarily consistent with the documents that we have nice volumes and so doing it over again we would probably put this on high volume first to take the hit on getting softer be able to read those another thing that was helpful for us was the visualizations to be able to see a big picture so the utilization of flow charts for our staff so they can understand the process bullying and be able to see the process travel trying to finish and so those are some things that they brought the table lessons learned the business case for it you guys can read the slide so so far it's working out really well for us our goal was to try and save six date employees in our front office within the first two months we have saved five full-time employees and the production is still ramping up we have a lot of documents / currently processing 35 to 40 term the documents using Koufax in fact less check and it was at thirty-three percent and so so far we are happy with that another another positive result is that we would do the double-blind classification so we have somebody classifying the document initially this was all in my thinking looking at it and they were classified and the court of the second employee would also classify but one thing we found with Kovacs because then Koufax was so effective and classifying that on the second round of classification there was a disagreement it was in every instance the human that was classifying at home and so it's been nice for us because we've been able to get rid of a lot of the documents that second little classification so in terms of the business case so far it is looking very good for us we are reaching our goals in terms of amount of personnel we have working do we got for classification and our goal is to get to seventy percent of the docs and so we're continually refining and trying or them to go through go back south more in terms of future projects now that we've seen what it can do we've been thinking of other ways that we can utilize the software in our business we compile medical records files for administrative logic isn't it to decay these cases and one of our biggest difficulties is getting duplicate evidence out of these plants and so they are compiling these files as time goes on and we've got to collect amount of records on any any individual has a claim and so when we collect medical records the last criminal records within certain pages the medical records will come in will lend them to a judge's file and then nine months may pass and we might need to give updated medical records well we said in the medical records request we put a field day rate on request and say we just need the medic record territory of state but when the facility gets the request they never look at the day and they just drop everything in the middle that they have and we get a giant pile of butter crackers and if we submit the big pile onto the judge the judge is going to be ignorant or the same are you sending me all these records that already have already seeing these I do with records that I've seen before and soul of the things we actually win a movie about our preferences we have to go back to the physical file and we have to look and see if we already have these date ranges of the fire and it's in terms of paper and processing difficulty in crates for a fellow horse because many of the gunner Burgers coming electronically but we don't have the ability to side by side like an electronic file and look at the better progress on one screen look at the file on the other screen and so what they've done is governmental records electronically printed them on a printer looked at the purpose of the paper documents and compared them with an electronic file if they see duplicates and these will be horrified under page sets of documents sometimes but if they don't if I duplicates then they will pull those out and then once they pulled out everything i think is to look it then they've got a hoarder the evidence all my day very time it's been four people once they have organized in my day and Hamilton at remote we would then shred those documents after we've a little bit as kind of a clean fire and then those will be submitted so what the Koufax is anything has to do is for us to run the documents through Koufax and run the judges bathroom Kovacs and then you will freedom in our matches and will identify them so you might have a front pages it can say oh nice sixty or seven are are identical to the documents of the file and then you can open it up and look side by side and you know when they're duplicates you could first have them removed and then let's remember that we can go in and insert dates on the various visits and then export them to a file with no duplicates and out by day and so in the process of making this work right now really very excited about it and so we think that will happen to return on investment also reduced in a verte employees that were working on these cases so other things that were looking at on reading the medical records that we have to identify the types of cases of other claims that we could be making on these cases based on certain keywords that we can identify a mental crackers so it's been great frankly working with good partners in this and our experience or ours been tremendous it's a great product and what you feel as though you know within a certain amount of time a very short period of time received a good return on the investment and as we move forward of the future will pay us those dividends year and so good experience with the software spring which you guys are doing for businesses and I guess prevent us from telling lies my name is Michael miles i'm at doc's tech and the president CEO redbox tech where Koufax partner we've been working with Koufax for about 10 years Koufax of dining partner have had the opportunity to work on a pretty wide variety of projects but are excited to talk today a little bit about the project that would work together with Jared and of either team on creating so like like Jared talked about my disability is an advocate for those patients who are five from Social Security disability claims and they have a tremendous number of different document types because of the kind of business that they're in and so as we talked to Jared there were really two major projects that we identified the first was the question about whether we can automatic Lassa fication so could we scan in all these documents and automatically identify what kind of document they were among the five hundred or so document types that mighta had identified and the second piece of that project then was once we scanned in these documents for any of them that were part of the medical record or would be had in the medical exhibit could we identify whether or not the documents that we had received or duplicated medically the company already had from this particular patient the first phase of that project is the automated classification phase and like Jared mentioned we start on that project back in july of last year and went into production in January it was a six-month project for us and essentially was about two and a half months of development and about two and a half months of user acceptance testing Jared mentioned to you that we are essentially a forty three percent of their documents today or they run through the Koufax model the reason that that's where it is is because having those documents classified correctly is extremely critical to the miler process they've got an extensive set of workflow steps in their own line of business application that these documents are routed to based on these classifications and in the Koufax world we sometimes would call document classifications being sort of a general type of document and then within that classification criteria that would have metadata which would determine in some instances how these documents were routed around and where they were sent for processing and approval it goes back to the lesson learned that Jared mentioned earlier and that was again document types mylar called every single one of those variations a different document type and many of them had different writing routing rules so we've been based on a specific state form that was filled out as you c n imagine it might need to go to a different attorney that specializes in disability claims that particular state so even small nuances between what we might look at it considered to be the same document type were important to the my other business model and so from a classification standpoint then we were we were put in a position where we had to we had to be able to tell what were these and where do they need to go and have any need to be routed so what was important for us then or what was the lowest key to how we were able to do this successfully one of them certainly was integration whenever you talk about leveraging the power of koufax any information that we can gain contextually from a line of business knowledge about the process helps us do a better job of using the technology to automate the process and so in the mylar case we integrated tightly with their existing line of business application they had a program that they had written in house that they use as their CRM system as their process automation and document repository system and so we integrated tightly with that to understand what were the document types where was this client in the process of having their case heard and appealed and finally to decision Don and where did these be to go what information do we need to know from the document order property routed based on what kind of a document it was client was in the process we did do a some initial sorting during the document prep stage so before documents were even scan we did do some initial sorting we separated out as an example those that were part of the medical exhibit file they were scanned in two separate batch classes so that we could actually apply a separate set of KTM project rules to those medical records as opposed to the other claims documents that they're receiving that went through a separate set of KTM processes and we also work to be able to collect and capture all of all the different document types that mother was receiving some of their documents came in via fax some of them came in from an email some of them were downloaded from a pearl someone came in through a traditional scanner it was important for us as we implemented the Koufax solution that it was truly a multi multi captured solution forum so how did we do it then well classification was key and we did it essentially by leveraging and I say we just to be clear by the way it was my team so I had like Jared mentioned I'm kind of on the sales side so we joke about this a little bit with that within dock tech but aren't a sort of promises things that an implementation team has to do it so I don't want to take any critic if I did not do this and feel free to ask any questions you want to try to accurately communicate what our team built but they did a great job in understanding how Koufax could be used the power of the tool and then leveraging that to provide the business value that we on the sales side had promised to the customer we started with what we call layout based classification and I recognize that we have members of the audience they probably are very familiar with Koufax and some of you that may be relatively new so I'm going to just briefly describe what each of the different classification methods are and how we use them in the case of miler so layout based classification is essentially what you think about in what is the document look like you look at a document and based on how its laid out or what it looks at looks like you can tell from a high level is this an invoice is this a contract this is an application is this a form it's it's the easiest way to be classifications the quickest way to classification in the Koufax world and it does a great job initially if you have pretty distinct differences in the way document looks so layout took place layout based classification for the miler project was the way we started and it provided a great framework for us but it did provide enough detail differentiation between the various document types that we had so we went on to content based classification content based classification is where you're going to mostly are the entire page and then based on the OCR content of the page you're now going to make a distinction between what the document type is the downside the content based classification is that you both start the entire page and you're comparing the entire page to basically an entire OCR page that you have in your sample set so it works well until you get again some of the very small nuances that we found in the minor model one of those examples is we have something called a 20-day notice and a 75 day notice well the layout is exactly saying it's a notice and ninety-nine percent of all the words on the page are exactly the same the only difference is it says 20 instead of 75 so in order to classify between documents that are this similar you essentially have to de Lawd what we call instruction or context based classification and that means that you have to give extra weight to certain types of words that fall within certain context on the document so obviously the words that fall just before the word notice so 20-day notice that the set of words that come just before notice as an example were more important to us than any other word on the entire page and so by using the understanding the context of this document then we were able to accurately classify the difference to a 20-day notification and a 75 day notification so the Koufax tools i guess my my kind of overarching method about classification are extremely powerful the more you can understand about the doctors that you're working with whether that's a bar code that's on a form number that's on a separator sheet that you use any information that you can glean as you capture those documents can help you in building your classification model and all those things are really important to us as we work with Jared and his team on the minor project we also used the Koufax online learning extensively as we moved into the user acceptance testing like I say that for us was fairly long it was two and a half months to initially go into production and his team is still certifying a different doctor and additional document types so as we essentially continue to roll through document types the number of documents that are being violetta system automatically grew up to 43 today our goal is to be at seventy percent automated classification within the next month or two as the model continues to get better and as his team continues to gain confidence that classification is being done correctly one other aspect of classification I think that would be important as you consider doing a similar project within your organization often in the Koufax model when you use a classification of any type that's a step that happens before extraction so as you think about a typical Koufax workflow you're going to capture the documents that you need you knew all the image clean up that you can to give you the very best image as possible and then you're going to go through a try and classifying and based on the classification and then you're going to go through and do your OCR and do your extraction however in our case because we were using that OCR contd in a way that it was critical for doing the classification we actually created a data field for each of our document types as part of the bass class that was classification we have hypnotized the classification types in that drop down so that as the users review these documents and if there were Corrections that they needed to make the various classification tags were in the drop-down they were alphabetized they could start to type a classification type and it would fill in the rest of that form it would auto fill it forum but it also allowed for us to change the classification tired as we continued through the model so we would use layout based classification but then as we got into content and instruction and context based classification we can change the value in that drop down in order to accurately classify the documents before we presented it to a validation operator instances with that needed to happen so having the classification type as a document field for each document was it was an important decision that the design team made as part of the project after classification then we needed to know who these documents were for and so we went on to what's traditionally called extraction this would be reading or mine in the documents for the key pieces of data that we needed to determine which miler client was documents for and we did that essentially based on the document type as you can imagine some document types have a Social Security number and some don't in some cases we were reading social in some cases we were reading the name we were able to use the Koufax match emerged server so there was a fuzzy match that we were doing we didn't have to have an exact name we had done a look up through our integration with their in-house line-of-business applications so we knew which clients they had we knew what social security numbers for example were balance of security numbers within our model in which customer names were valid names and so we were able to compare the results that we were reading off of the document with this dictionary that we were creating based on the book up that we did with their in-house line of business application and that allow for us accurately that identify who the client was that these documents were fourth as we received it so what did we learn from the project the collaboration is key I think we often go into a project like this and recognize that many of many of us as solution providers have worked with Koufax for many years and so sometimes there's a temptation to say you know we know Koufax and we how to build this product for you so tell us what you want and we'll build it for us when we work with clients like Maya and especially as we look at more and more complex projects it's very important that that interacts would be collaborative and that it not only been collaborative but a bit but that it be iterative so it was key for us that we have access to both the IT side of the miler group as well as the business side and we work very closely with them we had regular meetings we presented on regular basis about the progress we were making around the model and sometimes we would take two or three steps forward and sometimes we would take two steps back in trying to get the model to accurately identify the document types that they had I think and not to toot the horn too much I guess of our team but it was very rewarding for me as we met with Jared his staff just a couple of weeks ago and said hey you know we're going to be giving this presentation what would you like us to talk about what kinds of things did you learn as you went through this process and they said that interacts with jobs tech team that time that they spent actually watching us work not just in the conference from asking questions but sitting down and watching us do our job was extremely important they said in fact we actually offered one of your guys a job we want to know if he'll come work for us because we're pretty sure that he knows our documents better than we do so luckily he didn't take the job Craig is actually here so I could joke about it because he spent a lot of time with that client and it was it was important that we learn their world that we sit in their shoes that we want them work otherwise we never would have been able to build this model in a way that it would work for him we did know the technology and that was important but they knew their business and they knew their documents and so really having the humility maybe having the willingness to listen and work together was critical for us to achieve the success that we did other lessons learned so discovery is off as always something for us its importance we always talked about from a project manager standpoint and yet this was another example as we look back at the project where we could have done even more oftentimes an engineering team is anxious to jump in and start building the solution they want to demonstrate the value of the product you know they've got enough but understand that they think they can jump in and start building the model we could have done a better job in this case in spending some more time documenting the process doing the detail discovery and I believe it would have even shorten the implementation time frame for this particular customer if we had done a little bit better job of that now we still bet the time frame that was sad we still were able to implement within the six months that we had told him as an estimated project time frame but for us making that investment in discovery was a lesson that we learned and one that we were again reminded is extremely important when you talk about designing complex solutions like this an example of that was around the documents that we choose to folk to focus on and Jared mentioned this before so I will spend a lot of time on it but we had chose to focus on those that we felt it would be the easiest to automatically classify we talked about this concept around quick wins sometimes in a strategy that you take around a project however because roi was so important to the customer because they wanted to understand what the value of the solution was going to be we would have provided more value for them quicker if we had spent the extra time and focused on the documents based on the volumes that they receive so you know you can look at a list of 500 documents they know how many they receive every month we we finally did go to that model but we could have gone there quicker and provided value to customer a little bit more quicker a little more quickly in the project if we could if we had focused in that way so I want to just to touch on briefly the next project that Jared talked about somebody asked me about it earlier in we weren't be clear about the difference between what we've done of us in production and what we're working on in planning so just to be clear the project that we've talked about up to this point is is in production and Jared talk about the business value that they're receiving from it we did identify upfront as part of the sales process that there was a second phase it was a second project and it was this idea that we could well the question was could we look at to scan medical records and tell you whether or not they were duplicated and so the second project working on with them now involves taking the X dot information if you're familiar with the Koufax model that you receive when you run the OCR and comparing those the X dot data that you've gathered from two different Koufax documents and being able to determine what percentage likely that there is the two documents are duplicated when you start to talk about medical records and medical exhibit files this becomes fairly complicated because you may have one set of medical records that were street was received from the doctor those are the original those medical records that may have been faxed to a hospital and so when we receive am at Milan they come in as a faxed copy from the hospital they're actually the same record they're covering the same time frame in the same kind of doctors visits and if you present both of those to a judge he gets pretty upset in fact jaron talked about cases where a judge would actually throw the case out because you have so much duplication within your fire and so we wanted to know whether or not these two were the same document but traditional OCR and some of the standard methods didn't necessarily work for us we have been able to build a model we're still in development of this solution for them today around this idea that we can accurately look at not only X doc data which essentially is the sum of all those see our data that Koufax has collected but also start to capture information about how many lines of text are on the page and how many actual physical lines are on the page and based on a collection of that data then tell you the blanket of the two medical records duplicated and instances where we think the likelihood is high that they're duplicated we then prevent those to an operato for them to make a decision about whether i was actually duplicate document if it's duplicated they can then remove they can then delete the duplicate from the file and then accurately order the exhibit file before they have put it and prepared for presentation before the judge so that's the next exciting project we're working on and it's been a lot of fun it's always challenging for us when we come in and they tell us that it can't be done but if somebody else is met with them and total that it was too complex but that's a challenge that we enjoy taking on we like those instances where somebody says hey we tried and couldn't do it because we've had a lot of success over the years working with Koufax and the Koufax tools and implementing projects successfully by investing in understanding the business problem and investing in knowing that Koufax tools and then trying to marry those two you