Name: Simplify, Streamline and Succeed - How a cost effective, innovative AI solution can enhance the processing of documents in Finance and Insurance
Uploaded: 2025-10-23T10:46:04.866Z
Duration: 29 min 52 s
Description: Simplify, Streamline and Succeed - How a cost effective, innovative AI solution can enhance the processing of documents in Finance and Insurance

Transcript for "Simplify, Streamline and Succeed - How a cost effective, innovative AI solution can enhance the processing of documents in Finance and Insurance": Good morning, everyone. We'll be starting in just a few minutes' time. Just give some time for people to join. Hello, everyone, and welcome to Simplify, Streamline, and Succeed. We're really excited to have you with us today as we explore how AI can transform the way that you handle your document processes. I'm Sean McDowell from Open Connections, and I'm joined today by my colleague Richard Rogers. Together, we at Open Connections help organizations to simplify and to automate their document driven workflows. Open Connections was one of the first document companies to enter the document solutions space around thirty years ago. We're absolutely committed to quality, expertise, and long term partnerships with our customers. We partnered with Clipper because we saw the need to offer our customers a modern cloud native document processing platform. Clipper has, to date, processed more than 500,000,000 documents across more than 1,000 brands around the world. The platform that we're going to show you today is called the Doc Horizon. And we believe that Doc Horizon will help our customers to take their document capture to the next to the next level. Many of the challenges that you might be facing today with your previous generation capture solution are now a thing of the past. For example, a given invoice may have been configured as a template, and now the date field has moved or it looks slightly different. So your data extraction fails. The data is all still there and it's still presented, but it just looks different. So the system no longer recognizes that that invoice. With Dockerizing, that problem has gone away. This slide shows the flow of documents and data into and through the platform from email, file stores, REST APIs, mobile devices, and scanners. In fact, almost any kind of document from any source. Platform then carries out the OCR, data extraction, classification, fraud detection as necessary, and it can show the results to a human being if necessary as well before passing the doc the documents and the data into downstream systems. We can also integrate with a range of third party AI tools to enrich the data that's been extracted, and you will see a few of these mentioned later in the presentation. For the more technical amongst you, all the capabilities that we'll show you today are available to use via APIs. So if you need to or if you want to integrate with your existing ecosystem, it's very, very easy to do. We're going to show you, some advanced OCR. We have examples of machine print and cursive handwriting. We're really excited about the handwriting because if you've ever tried to get a machine to read real world handwriting like this or even like this, you know what a challenge that that has been and even still is in some cases. I was with a customer a number of years ago, and he was a director of a he was a customer services director for a large utility company. His name was Barry. And Barry said to me, he said, Sean, I've got a problem. He said, I've got hundreds of letters being sent in every every week into our company. He said some are compliments, some are complaints, some are handwritten, some are machine printed. And he said that I've got staff who have to sit there and read them, triage them manually. And and he said I need help because it's a problem that's getting bigger and bigger. I said to him, Barry, I said, we've got the latest technology, but we don't have any technology that can help you with that. And he was very, very disappointed. Or Barry has since retired, so he's not on the call today, unfortunately. But I could tell you that if he was here today, he would be thrilled by what we're about to show you. I must emphasize that these results are possible with just just minimal very minimal setup. Audible documents are challenging in every industry. Every document can be tampered with, whether it's an invoice, whether it's a passport, or whether it's a a supporting photo for an insurance claim. And we can now build robust fraud detection into your document workflows. Imagine being able to interact with your documents in much the same way as you do with chat DPT or Google Gemini. This is what the Doc Horizon prompt builder brings. The the the ability to build AI prompts, to do data extraction, or summarization of documents. And this works on pretty much any type of document, whether it's a form or whether it's a a free form less structured documents as to the contract. If you do have questions, please put them in the chat. We're deliberately keeping this webinar short out of respect for your time. What we'll do is we will email every attendee the full set of questions and answers after the webinar. And if you don't want to ask a question in the public forum or or if you happen to be watching the replay, please email questions@openc.co.uk, and we'll definitely get back to you. So without further ado, I'm gonna hand over to Richard now for the demonstration. Thank you, Sean. If you stop sharing, I'll, pop my screen up. To share now. See that. Excellent. Good morning. As Sean said, my name is Richard Rogers. I've got fifteen or twenty minutes to try and show you some of the more exciting parts of the Clipper platform, but there is a lot more to it. So if if after the, today's demo you you want to know more, please feel free to send us a message at the questions @openc.co.uk, and we will get back to you, and we can always set up another short demo. So let's jump straight in. The first thing I'm gonna do is demonstrate the system. I'm gonna use the, testing facility that's built into it to try and do this quickly for you. As you'll see, I'm logged in. I'm in a demo account. And the first thing I'm gonna do is show you something under the financial model. So this is a prebuilt configuration that understands a wide range of financial documents, and that covers everything from invoices to receipts. I'm gonna take the try service. I'm gonna switch to a configuration that includes the forward detection just so that we can show a little bit of that later on. And then I'm gonna select a sample file. So I'm gonna put this invoice and I'm gonna click extract. Within a few seconds, the system will have processed the document. First thing to notice, this is a a photo taken on a phone with the invoice on somebody's desk. It's not come through straight through a nice, high value scanner. This was received by email and has been processed by the system. The kind of information it can pull out very easily, you'll see see it all here, the country code, the date of the document, the invoice number, this, the supplier. It can then drill down into tax details and into the line items showing you each each of the individual rows on the invoice. Now there's plenty of systems out there that will also see our an invoice for you, and this is doing a very nice job of it. Let's just nip over to the full detection. And this is showing that there are several levels of full detection within the system. It's spotting that the line items have actually been tampered with. We haven't got time to drill into all of that. If anybody's interested, ask for a demo later on. But it's also noting that the even noting what product last edited the document. Now invoice is obviously a common document within a lot of organizations, but in the finance and pensions and other industries, similar industries, we we process a wide range of documents. One of the other ones I'm gonna show you now is the identity documents. So a lot of organizations in these industries have a requirement to know their customer to validate the person they're dealing with, and it's not unusual to ask for driving licenses or passports as part of that process, and we've got a requirement to read those documents. So I'm gonna pull in a a passport. Now this is a nice clean scan of a of a passport, but I can show you others where in in similar to the the, invoice, they were scanned. Somebody's taken a photo on their phone, sent it in. And as you'll see, again, it's pulling out as as I hope you'd expect by now, the all the exact details. And this works not just for UK passports, but for a wide, range of passport formats from around the world. So even if our customers are not working with UK passports, it reads them very nicely. There are actually some standards for passports that make that a little easier than you might think. But also notice it's pulling out a SNP. It's taking the photo and giving it back to us as a an image that we can then pass on to other systems as well. If we go to fraud detection, it's actually spotted that this document's been tampered with. It's, alerting us to the fact that some splicing was taking place. So in fact, if I pull up the original passport, you'll see this person actually was from British overseas territories. The passport's been edited to make them a British citizen. And in this country, that would grant them rights, for example, to work that they might not have had, with their original passport. Now all of this is very nice. These are all well known, documents that you can have pre trained systems for. But lots of organizations deal with the type of documents that are either internal or documents, and forms created by them for which there is no standard that would allow us to preconfigure the system. And for that, one of the options we've got is prompt builder. Prompt builder is starting to use AI to read the document, But today, we're gonna actually create a new configuration for a document that we've not, seen before. And in this case, we're gonna use I'm gonna get set up. We're gonna use a document that is quite common, which is a pension transfer form. Now there are lots of pension companies out there. They all have their own forms. There's very little standards or conventions to them. And in fact, this form, although it's been edited slightly to the the company's name, was taken straight from the website of a well known pension company. Now you'll notice on the right, we've got a lot of preconfigured, templates. In fact, there's 24 of them currently for all sorts of things from shipping labels and menus and vehicle registrations, but we're gonna use a double dose of AI today. The first thing we're gonna do is ask AI to analyze this document, work out what's useful on it, and configure some questions or prompts to ask AI to extract that information. That will take a few seconds, and we'll get that going. Now it's worth bearing on that this is a three page form, so there are some things that it's gonna pull out that you won't see on this page. And now it's configured, and and including put them in groups. So we've got a group called personal details. It's, like, identified that there's a section for that. Whereas if we go to page two, there's other sections. So the name of the scheme that we're transferring from, so it's created these in groups, created the fields, and this piece of text that you see here is the actual prompt that we're asking AI to use to extract that piece of information. And the beauty of this is that if the layout of this form changes over time, the system doesn't rely on a position or anything like that. It it just knows how to go and find that information on the document. So to speed things up, I'm gonna scroll down and delete a few of these sections that are actually pretty standard. So the declaration will take out the return address, and I'll go over this skip past it. Once we've done that, I can preview. And this will show me that the data that would be extracted from this sample form. And you'll see that already, it's pulling out a reasonably good amount of the information very reliably. Something I meant to point out on the prompt fields, if we go back, is it's actually worked out what type of data it is. So we've got text fields, date fields, numbers, but also Boolean fields. So where we've got tick boxes, it's identified those and built the data appropriately. Now that's taken me a minute, maybe two minutes. Another ten minutes, and I could maybe tweak or configure this, and it would be ready to use. Now how we use that is entirely up to us. We could build that into a system that monitors an email box, interacts with Microsoft Teams. It could be putting the data into a database at the back end when it's finished, or it may be using the human in the loop to present documents where it's less confident to an operator for for some manual intervention. Well, that's great for forms, but what if the documents that we're processing are more generic, office documents, reports, contracts, etcetera? If we go back, we're gonna use this service agreement as an example here. Now this is one that was configured earlier. Now there's two two possible ways of going about this. The first one might be the more conventional, which is to have fields for who's the client, who's the supplier, contract start date, and various other pieces of information. Or we could just go for a more generic approach and I'll get to summarize the document for us to pull out any, concerning clauses, pull out the key terms. So let's give it a contract. So this is our service agreement contract. And as I explained, so one option would have been to go for this very structured format, if that's the kind of information that we wanted with start dates and suppliers and what the products are, or a more generic summary that might be pasted into SharePoint or other other document management systems as a top level field that would allow us to understand the contents of the document. Bear in mind, this is six pages long without having to read the entire contract. Some of you might have noticed in my samples folder a a picture of a car. I'm just gonna pop that one in while we're here. Now I'm gonna do it as if it was a financial document just for speed. Fairly convincing picture of a car with damage to the front wing And a different, type of fraud analysis here, it's not looking at the the rows or the columns, but it is spotting that it's it's reasonably confident that that was produced by AI, which it was. So we've got an amazing set of powers here to deal with both structured documents, such as invoices and receipts, identity documents, which covers a a wide range of passports, but also identity cards from various countries, driving licenses, and other identity documents. There are some others here that we're not focusing on today such as salaries and bank statements. But the most powerful is this prompt builder. It allows us to deal with any type of document that we've got in our organization, whether it's electronic, currently on paper, or the same type of documents stored in both formats, and extract really useful information from it and build that into our business processes. Now we don't have time to demonstrate it all today, but we do have this option here, which is the flow builder. And this is what allows us to integrate all of these facilities. And as Sean mentioned, there's APIs, rest APIs available that allow us to take all of this and build it into our existing platforms or processes. But sometimes we want to do that without having to get a a bunch of developers in and and create a project for it, and that's where Flow Builder comes in and it allows us to I'll very quickly open one up. For example, in the in the longer version of this demo, we can send one of the documents to an agent in Teams, have it process the document, do some of the work, and then send the massive message back in a Teams channel. But we can do the same based on email. We can do it based on, web chats and all sorts of technologies. And in fact, if Sean will reshare his screen for a second. Which means I need to stop. Yep. There we go. As I say, we it's way too big of a part of the product to demonstrate in in a few minutes, but these are just some of the products that it allows us to plug into and plug the technology that Clipper provides into the kind of tools we're using every day from our back end databases and our mail systems and our storage systems, and just our general Office documents, whether, Office applications, whether that be Microsoft or Google, and allow us to bring these technologies to our users in easy to use, formats. Now I sped through that at a reasonable speed. I don't know whether there's any questions to deal with yet. There are some questions in the chat. Yeah. And I've I've been trying to do my best to answer them. See if we can go back. So there's a question about any language limitations with prompt builder, and the answer is that prompt builder works with all Latin based languages. The second question was how easily does Dockerize and integrate with systems like CRP, ERP, or DMS? I think you've just answered that one, Rich. Then Sorry. I'll just expand on that one. The Yeah. Obviously, there's several of the major CRMs. So for example, Dynamics or Salesforce listed there. But even if you're using your own in house developed products, if there's an API or an integration, at the bottom here, maybe I should have just put REST. The the fact that there's a set of REST APIs for this is exceedingly powerful, and we've we've actually used it to feed into systems that Clipper have never heard of. But as long as there's an API there, we can integrate with it. There's a question around where is data processed and stored. And following on from that, can sensitive information be redacted or restricted based on user roles? So the the system is cloud based. And, again, I'll put it in when we prepare the response to this that CLIP will have several high level, accreditations for the security of the system. The cloud based nature gives us the ability to develop this product. Ever increasing speed it would fail, adding new technologies and new facilities to it. But you can rest assured that the using the APIs correctly, the all all your all your documents are are highly secure. There's a question also about adding prebuilt models. I'm not really sure about the context of that one. It says, can you add the prebuilt models? I've probably talking about large language models or Well, I I think we're we're probably talking about adding so if if if could you build your own prebuilt model? You certainly could. Yeah. Again, maybe I'll I'll put a a more detailed answer in the, set of answers that we're gonna send out afterwards. Okay. That's taking us to the bottom of the questions. So thank you all very much for your attendance. And as we said, we will circulate all the questions and answers, to everybody that's joined the call. If there are any follow ups, if you'd like to get in touch with us, please use the questions at Open Connections dot co dot u k email address. Maybe I'll just pop that back on the screen, briefly. Questions@openc.co.uk. That's the forward, that's the point of contact going forward if you if you'd like to follow-up with us. Thank you all very much, and have a good rest of the day. Thank you.