Introducing Acrobat Analyzer

Transcript

Welcome, everyone. Thank you for joining us. I can still see a couple of people coming on, but we’re going to kick it off with our webinar, Introducing Adobe Acrobat Analyzer. We have a pretty exciting agenda for today, but first of all, let me introduce our speakers. If we could move forward to the next slide. Thank you. So my name is Jonas. I’m your moderator for today, and I’m also going to be moderating our Q&A part after the introduction.

But our key speaker today is Pranav, and then we also have Hadlim here, who’ll be taking over some part in answering lots of questions. But let me hand over for Pranav here to introduce himself.

Yeah. Hi, everyone. Really excited to have the opportunity to share this innovation new product that we just made available for all of you. My name is Pranav Parikh. I’m the head of the Acrobat Analyzer team that’s building out this new product with Acrobat Sign. We’re really here to share more about this product, answer your questions, and see how we can solve some of the unique use cases and challenges that you might have faced before with this new solution. So with that, I’ll hand over to Hadlim.

Hi, everyone. My name is Hadlim, and I’m a product manager part of the Acrobat Analyzer team. Really excited to be here with you today. So let’s go ahead and get started with a great webinar. Back to you, Pranav. Thanks. Let me just quickly demonstrate the agenda here. As you can see, we have a pretty tightly scheduled topics. We want to tell you a little bit about why we’re focusing on Document Intelligence now, what Acrobat Analyzer is, how it works, and then we’ll also spend some time on a demo and go through a couple of key business use cases.

Before we then end with Q&A. For Q&A, and this is the last bit I wanted to make sure everyone here understands. Since we have a pretty broad attendance here, the chat for this meeting is disabled, but you have a Q&A button on top of the Teams webinar. This is where you can interact with us and ask questions, and we’ll do our best to keep up with all the questions that come along. And at the end of the webinar in our Q&A section, we will pick out the most upvoted questions first. So you have an active role here, not just in putting a question up, but also in upvoting questions that you might be interested to learning the answer about or hearing a bigger answer in depth. So there’s a little arrow bar under each question that you can vote for. And with that, you can utilize the Q&A bar and we’ll make sure we’ll lead a good webinar for you.

Lastly, polls. We will introduce just two polls here during the webinar. This is just to grab some feedback from you as we go along. I’ll launch the first one now just to get you used to it. And then we’ll have another one towards the end of the webinar to grab some feedback from you again. So here we go with the first one to learn a little bit about your role today. And Wei, from what kind of department you’re joining us from.

And the second one then focuses a bit more on actual document review and really you might be involved in. OK, wonderful. I see the first question answers coming in. And with that, I’ll hand over to Pranav to take us away. All right.

So let’s start with learning more about Acrobat Analyzer. It’s enterprise-wise document intelligence now available with Acrobat Sign. And what I would love to do today is set the context for what this new solution is, what type of problems and use cases we are trying to solve. We’ll talk a little bit about the actual construct of the product. And then we’ll dive right into the demo. We’d love to show you a demo. We’ll love to do a live demo of the product, show what it’s all about, what type of problems it solves. And then we’ll talk a little bit more about how to actually use it and how to integrate that in your workflows. So with that, let me get started with the problem that we are trying to solve with this particular solution. So with our investment in Acrobat Sign and the Acrobat ecosystem, over the past many years, we’ve heard from our customers and partners that one of the key challenges that they face today in their businesses across the enterprise is the document-centric business workflows that are really mission-critical for the business. So for example, in the finance and accounting organization, the revenue recognition workflow or the payment terms and billing schedules management workflow. In revenue ops and sales ops teams, accelerating the code to cash and driving contract operations and contract reviews for that, or extracting contract value and terms to be able to drive other sales operations processes. Similarly, on vendor management, having full visibility into the vendor spend and vendor license entitlement, auto renewal clauses and timeline, renewal date tracking, all of these things which require somebody to really go in and have visibility into, or manually kind of get that data from vendor agreements. Similarly, on procurement operations, having deep insights into purchase orders and invoices and understanding the statements of work and using that to be able to track vendor work and vendor compliances. And similarly on the risk and compliance side, tracking audit rights, SLAs, monitoring SLAs and reporting for compliances in a variety of different workflows. And one of the common things that is sort of unique and common across all of these different workflows is that they all depend on some data that is buried in these documents. And these documents are customer contracts, vendor contracts, purchase orders, invoices, audit reports, statements of work, where the information is written in an unstructured way in these documents. It doesn’t exist in any other system of record, like the CRM or CLM or ERP or the Span Management Platform. And the goal is to really review this data, get this data and use that information to either drive a workflow or make some decision. And today this requires somebody to go in and manually open these documents, review them, get this data and use that data to actually, or input that data in some other system to be able to drive these workflows forward or make some decision and record the decision.

We think there is a better way to solve this problem using technology. And that’s where we are introducing this new product called Adobe Acrobat Analyzer. It’s a new Gen AI application that is now available with Acrobat Sign that lets business users extract structured insights from tens of thousands of unstructured documents to automate document centric business processes. So how does Acrobat Analyzer work? The way it works is that you bring in your thousands of documents. You can bring in contracts, MSAs, order forms, SOWs, invoices, compliance documents, any type of document that you really want to analyze, you can bring them into this product.

Once you have those documents in, and you can bring these documents using APIs or a SharePoint integration or by manually uploading them. The most common ways to get these documents into the solution is through bulk ingest APIs or SharePoint integration. Once you get these documents into the product, the second step is to really define what you want to extract, what insights you want to extract in plain language.

So for example, if you want to extract whether all of these documents have certain clauses, whether all of these documents have certain type of financial terms, or just extract the party name and governing law and the license quantity from all contracts, whatever you want to extract, either it’s a data or some type of insight, you define that in plain language into the tool. And then it actually goes in and extracts the data or insight from all tens of thousands of documents that you uploaded. And then that extract that insight and data are available through APIs, as well as within the user interface to actually review and also available as a CSV output.

So I think if you take a step back and think about what it actually achieves is it takes thousands of unstructured PDF and Word documents and then it gives you structured data and insights in the schema that you outlined and the specific data and insights that you have specified. And that’s what we’re solving for with Acrobat Analyzer.

So let’s do a live demo.

I’ll stop sharing my slides and then we’ll get to the product and we’ll see what it looks like, how it works, and we’ll take some real examples to talk about the features and capabilities. So with that, I’m going to stop sharing and I’m going to share my screen here, my browser.

Okay, so this is what Acrobat Analyzer looks like.

You go to analyzer.adobe.com or you log in through Sign. There is a workflow to log in to Analyzer through your Sign account, where you go to your Sign homepage and on your Sign homepage, you see a tile to directly log into Acrobat Analyzer and that’s where you kind of get here. And once you’re logged in, you can revisit that page using analyzer.adobe.com. Once you are on the homepage, this is what it looks like. You see your overall view of all your recent activities. You can create new document collections. You can create new attributes and we’ll talk about what those are for analysis.

You see your document collections that you already have and then you see your attributes and we’ll talk about what attributes are. Attributes are effectively things that you have decided to extract from these documents.

We also have a few sample sales agreements and a few sample supplier agreements so that you can easily test and get started with the product without having to load your own documents.

Now, the way you bring your documents in is through APIs or there is a way to upload files here. And then there is also a way to link your SharePoint folders. So if you were to link your SharePoint folders, you can pick which specific folders it can get linked with and then it will actually mirror the state of those folders where once you sync with specific set of folders, if you start adding new documents to those files in those folders, you’ll ingest those files into Analyzer. If you delete files from those folders, those files will get deleted from Analyzer. So Analyzer maintains a live sync with the specific set of SharePoint folders that you specify.

Now let’s go to a collection. So I’m going to go to my all files collection.

I loaded a few demo agreements here. Now, you can load any type of documents Now, the setup that you see here is fully custom. So in this example, we have set up the date added, which is the date these documents were added to the system, effective date of these agreements, governing law, the title of the document, whether there is termination clause or not in the language of the termination clause. And this setup is fully custom. And we’ll talk about how to get to the setup. But the more important part here is that when you click on a document, you see the most important insights that you want to see from these documents extracted from all of these documents. So you click on these documents, and on the right hand side pane, you start seeing contract value, parties, document title, currency, effective date, payment terms, termination clause, and a variety of other things. Now what’s unique about this solution is that everything that you see here on this right hand side pane is fully custom defined by you, which is the user. So the user can decide what they want to extract and see on this right hand side pane.

So for example, if I am in sales and all I want to see is basic general information, then I can just set up this general information and I can just set up these five things and that’s all I will see. Now let’s say I am in finance and my specific role is in FPNA.

So I’m actually, in this demo, I will set up this section to mimic an FPNA user. Now FPNA users, to be able to track booking, they have very specific clauses in the agreements that they look at. So for example, proof of concept, terminate and replace, termination for convenience, development licenses, a variety of different scenarios that they look at. So they can set up just these clauses and terms and nothing else. So again, everything that you see here on the right hand side is fully custom set up by each user.

Now the other thing that, and we’ll talk about how to set that up. Now the other thing that this does is, by the way, as you click through these documents, for each document, the setup that you have created, the same insights appear.

Now the other thing this does is, now if you want to really drill down into an insight, you can just click on the citation and it actually opens the document. And once the document is open, you can actually see where the insights are coming from. So now once the document is open, you can actually quickly see where each of these things are coming from and you can navigate through the most important clauses and terms and things that you really care about just using the attributions.

So now what this has given you is the ability to take your documents, define the insights that you want to extract and then use those insights to quickly review the document and quickly navigate through the document.

Now let’s see how these insights are set up.

So we call these insights attributes. So when you go to the attributes pane, we ship some of these insights out of the box. So where you see the system insights, these are some of the insights that we have shipped out of the box. You have the option to turn them off if you don’t really need them.

Now what you can also do is you can create custom insights. So here are all the different custom insights that we have created. Now the way to create this custom insight very simple, you go and say, create a new attribute and here you give this attribute a name. Then you select whether it’s a text type insight that you want to extract whether it’s a number, it’s a date, it’s a table, what is it that you are looking to extract? You can define whether it’s strict extraction or interpretive extraction.

You can define what type of document this is. So you can select contracts and agreements, and forecast expense reports, financial reports and audits, financial statements, form and authorization, invoices.

And you can also say other. And if you want to create your own type, you can just type your own document type.

You can give it a custom attribute group. So here I’ve created these groups, you can create your own group here. And then all you have to do is give it a description. So you define what this thing is in natural language, and you then give a few examples. So let me show you how we have done this. So I’ll give you an example that we’ve already set up. So we’ve set up this clause called terminate and replace. So I’ve given it a name, I’ve said, okay, I want to find a clause called terminate and replace. It’s a text type attribute. I want to perform a strict extraction, which means I’m not interpreting some broader answer. I’m looking for the existence of this clause. And if it exists, I want to extract the language of the clause as a strict extraction. I’ll define my document type as contracts and agreements.

I’ve given it a group because I’m mimicking an FP and a user. And then I’ve given it an English description. And this is what a subject matter expert would define. So again, we’re not looking at some industry standard definition of what this thing could be. We are asking the user to just say what they’re looking to find. In this case, the user has defined this as, it’s a clause that allows all or part of a prior commitment by a customer to be terminated and replaced by an amended contract. So this is the definition of this clause. And then the user has provided a few examples of how this appears in their contracts.

Now, what we do in the backend is we take these examples, we take this description, and then we create a semantic meaning of what the user is trying to achieve.

So there is no prompt writing here. There is no complexity of the user defining it in some complex prompts.

All the user has to do is explain what they’re looking to extract, provide a few examples, and then that’s it. Then we take that and create the semantic meaning under the hood. And the user can save that attribute.

Optionally, before saving, the user can evaluate the accuracy of that attribute and there is an evaluate workflow. We’re not going to go through that entire workflow, but this workflow allows the users to quickly build confidence in the accuracy of the attribute that they are just creating. And they can kind of get to higher accuracy by providing a very precise definition of what they are looking to find. Now, once you save this attribute, as soon as you save it, it becomes available for all files as part of this right-hand side pane. So I’ll show the terminate and replace clause here. So for example, in this document, it was determined that it doesn’t have it. Let’s find a document that has it. Okay, so now we’re not able to find any document. So that is an interesting problem. So I click through a few documents. So we set up the clause. Now we’re not able to find it in any of these documents. That means I have a large pool of documents and only some of those documents have this clause. So now how do I find these documents from a very large, so imagine there are 10,000 agreements loaded here. Now I will set up that terminate and replace clause as an attribute. Out of 10,000 documents, maybe there are only 10 documents that have that particular clause. Now, how do I go and find those documents? So that’s what we have done is we have taken these clauses and this set of these attributes that the user has set up and we have turned them into filters.

So every time a user sets up an attribute, it gets turned into a filter. So now you can do things like, okay, show me all the agreements where the contract value is greater than $100,000 and where the terminate and replace clause that we just set up is found. And when you apply these filters, in real time, those filters are applied. And now you narrow down those contracts. Now, when you click on them, you can immediately find that particular clause. You can click on the clause and then you can get to where that is in the document as part of, through the citation link.

Looks like we’re running into some slowness here on the network, but it should come up fairly quickly. Okay, so now we have found out where that clause is and where it is coming from. Similarly, you can, again, everything that the user has set up in a custom way, the user can go and find those things on those documents. So again, to recap, now what we have done is we have allowed the users to define what they would like to extract, insights or specific things from all their documents and set them up as attributes.

Once they are set up, they become available for all documents.

Using those attributes, users can quickly review each of the documents.

And when quickly filtered thousands of documents using the same attributes as filters.

So now you can imagine the power of this tool, which is you start from thousands of documents, you can bring those documents in, then you define how you want to structurize the unstructured data. And once you define the unstructured data, you can use the same, sorry, once you define the structured data, you can use that same structured data to really accelerate your document reviews and you can quickly find those documents.

There is also a full text search, so you can also perform a text search.

And when you perform a text search for, let’s say, agreement, you can, it’s a wildcard search, so whatever you search, you can search it in file names or file content, so it can perform text search on the entire content of all the files. But again, the real power is to bring the combination of filters using the schema that you created, being able to search within those. And then once you do that, so let’s again create that scenario, which is let’s say the contract value is greater than $100,000 and your contracts, let’s create a different scenario this time where limitation of liability is not found. So let’s imagine you are doing risk analysis and you want to find all high value contracts with uncapped liability. So now very quickly, you have used the contract value filter and limitation of liability filter to find all your contracts with uncapped liability. Now you can open AI Assistant. So this is your chat experience where you can say, okay, now I want to just have a very specific conversation with these contracts and you can ask deeper questions. Alternatively, you can say that, okay, these are my high risk contracts. I need to actually perform some deeper analysis on this over the next few weeks. So then you can quickly put that into a collection and you can say, okay, let’s create a collection called risky contracts and then that becomes a new document. So now when you come back here, you can actually see your collection that you just created. So now you can get back to the agreements that you kind of narrowed down based on the specific structured data that you created and that you sort of set up. Now that you’ve created these collections, now you feel like, okay, now you want to involve your team members on the analysis and have their help. So you can share this collection with your team members who have access to analyzer. So you cannot share it with anyone else. It only gets you, you can only share it with specific team members who have access to this particular solution. And then you can type that email address and you can share the collection with them.

Now let’s say you’re done your analysis. You have performed everything in terms of the specific review, specific risk analysis. Now you want to just get the data that you have here and get the data out to power your BI systems or to get this data into a system of record, like for example, enriching a CLM or enriching a CRM or an ERP.

So everything that you see here is available via API. So once you extract these things from all your thousands of documents, you can use the APIs, the Acrobat Analyzer APIs to get this data out of the agreements and into your system.

The other way to get this data out is to extract this as a CSV file. So you can select what you want to extract and then you can click on export.

And that really gives you the option to download all the data that you have that you just extracted from these documents as a CSV file. So you can use that CSV file for your BI systems or any other manual data exploration. So that’s the heart of the product. There are other things that are, there is governance, admin control, a couple of other things that we are happy to dive deeper into if there is interest. We’d love to connect and share more about this if you are interested. This is available today with Acrobat Sign.

And if you have access to Acrobat Sign today, and if you would like to enable this in your accounts, we’ll share information on how to reach out to us. But please reach out to us and we can set this up for your organization as part of your Acrobat Sign account. Now I’m going to stop sharing here and get back to my webinar slides so that we can talk about, we can talk about some of the use cases. So hopefully the demo was helpful. If you found it helpful, we’d love to see some reactions here on Teams. Feel free to share your reactions if the demo was helpful.

And so let’s now talk about a few use cases. So one use case, we actually covered this as part of the demo, which is extract semantic text at scale, and a very common use case is to extract terms and clauses from contracts, very common use case in finance, accounting, procurement, where you want to start with a pool of contracts and then narrow down and identify a specific set of contracts that have very specific clauses, very specific terms, very specific financial terms, governing law, any of those things. So that’s a common use case and we just talked about that. The other interesting implication of the ability to extract semantic insights is that now you can answer questions. So this is a common use case that we see today, which is in infosec and compliance, often security analysts have to look at SOC 2, SOC 1, SOC 2 compliance reports, other compliance documents, and then take those documents and the data in those documents and actually fill a form or fill a report to be able to complete an audit.

So today what happens is somebody actually takes one question in this questionnaire, understands the question, then goes to these documents, then finds the response to those documents, sorry, finds the response to that question, goes back to that question and then types a response. Now, in this case, what Analyzer allows you to do is each question can be set up as an attribute. So attribute that we talked about as a thing that you want to extract from these documents, instead of extracting something as a strict extraction, you can actually set up a question and perform an interpretive extraction. So each question in the questionnaire can now become an attribute in your environment, in the Analyzer tool, and then you can set that as an interpretive extraction.

And when you load these compliance reports, from each report, each question will be answered. So now you can automate the process of answering questions, a set questionnaire using reports that can come through in the quantity of hundreds or thousands through the course of the year.

Those are the two key use cases today. Now let’s talk about the key use cases. We see the most common use cases in finance, specifically in FP&A and accounting, and specifically within accounting in revenue recognition and in technical accounting. Then we see a lot of use cases in revenue operations and sales operations, especially when it comes to code-to-cash operations and variety of different contract operations teams involved in code-to-cash operations.

We also see a lot of use cases related to risk and compliance, mainly in the corporate risk team or vendor compliance teams. We also see quite a bit of use cases in vendor management teams when there is a need to track licenses or spend or renewal dates by the vendor management team for specific vendors. And then we see good use cases in the procurement operations teams where the procurement ops team is looking to understand the overall spend, they’re looking to understand the spend profile and categorize the spend profile by different vendors. They’re looking to understand their obligations. They’re looking to understand the overall rate variations in different vendors for the same category of spend. That’s where we see a lot of use cases.

So again, these are some of the scenarios that we know where Analyzer adds a lot of value in the processes that are currently manual and really drives that efficiency and cost saving and unlocks new opportunity for risk reduction and cost reduction. But there are many more use cases across the enterprise.

This is just some of the, these are just some of the key use cases that we have seen, but then there are many more across IT, across InfoSec, across even marketing where this can be useful.

So with that, let me talk about how to actually use this. So while Acrobat Analyzer is available with Acrobat Sign today, it comes with its own APIs and SharePoint integration. And it also has the ability to kind of allow exporting data through APIs. So our goal is to really make sure that this works with your existing investments. You’re invested in automating end-to-end processes. There is a chance that other than the manual document review all the steps before that manual document review and all the steps after the documentary review are already automated.

And what we want to enable is the ability to accelerate these document review and search and discovery as part of that broader automation that you might have already built. So we want to make sure that the way we sort of offer these capability through APIs and user interface actually fits well into your current processes and your current enterprise stack. So that’s where now with APIs, you can bring PDF, scanned PDF and Word documents from your CRM, ERP, CLM, even your ECMs or spend management platforms to improve using APIs. And at the same time, you can also get the extracted data out to drive your workflows and your BI systems and even data warehousing, if that’s where you need this data.

So that’s our goal. That’s how we have sort of envisioned and built this product, which is we really don’t want you all, our users, customers to really drive a massive amount of change management. We want this to be seamlessly available within your existing investment, existing enterprise stack and existing workflows. We just want to kind of help you accelerate your document review workflows.

So with that, I shared a lot today. We would love to share a lot more about the use cases, how to make this available as part of your Acrobat Sign environment, how we can solve your specific use cases and challenges.

If you’re interested in getting more information, please scan this QR code. It’s a quick three or four question form.

And if you can just fill that form, it will send us an email and we can reach out to you, find time with you, or we can just share specific information that you’re interested in getting. And if you want to just directly email us, please email us at analyzer-infoatadobe.com. And we would love to sort of engage, share more information and explore use cases and see how we can solve your problems.