[Music] [Hina Naqvi] Hello, everyone. My name is Hina Naqvi, and I'm Group Product Manager on GenAI for Acrobat.

Here is the agenda for today. We will start off by answering the uber question, why do users seek out AI PDF tools? Then, we will move to showcasing what Acrobat has just launched around AI.

Then, I will share our beliefs on what responsible AI means to Adobe. I will follow that by sharing the broader vision around PDF, as well as a road map related to AI.

And last, we'll round it off with a few questions and answers.

Some of us can relate to this image here. Too much information and we don't know where to find it. Today, PDFs are consumed the way they were authored. You have to make manually sense of the overall structure of the document. You have to comb through multiple pages of information specific to you, whether it's a single pass or sometimes you must make multiple passes to find out exactly what information you're looking for and is relevant to you. Sometimes, the document is so long and sometimes, it's just complex. Users don't want to have to read the whole thing to extract out exactly what they need and a little help can go a long way.

We can also categorize information into two buckets, structured and unstructured. Structured data like once in a database, which despite challenges, we know how to work with. There are tools to query that. But unstructured data is where a lot of time is spent at. 80% of information in any enterprise or organization is within unstructured data.

Did you know that there are three trillion PDFs out in the world and millions are being created every day? PDFs are gold mines of critical business data. They can be created from anything. That's the beauty and simplicity of it. You can create and distribute PDF to communicate information to others. But PDF itself has no structure. It has lots of text, tables, and images. So how do you make sense of it programmatically? I am thrilled to share the two new generative AI capabilities in Acrobat that will supercharge the way you work with PDFs. You've got what we call the generative AI summary on the right-hand side, and the AI Assistant on the left. Literally, before this, I would sit through and read a document from cover to cover online. I would highlight the bits that are relevant to me and then I would add comments to the PDF, and then I would take those and I would summarize it in the document. Whereas I now, I can say, "What relevance does this have to me and my role? What areas should I be focusing in terms of my role?" Often, the author knows about it, but the audience might want to consume the document differently, whether they want to skim or they want to go into details. These two features are there to help the reader get precisely to the information that they want.

And in many cases, a lot of times, people use a lot of content of the PDFs for reuse. That's a high-use case for Acrobat. Adobe is also looking at how to super charge repurposing of the PDF document. How does consumption and reading of the PDF help the creation of the other content is a critical part of the journey at Adobe.

Next, I will start us on how each function within the organization can save on hours by consuming content via these two features. I will kick start with the marketing function. Many of you can relate to some of the documents and business processes that are listed on this slide.

Extracting insights from corporate disclosures, marketers are often reading white papers, competitive landscape and other marketing reports.

Some of these can be really long and detail intensive. You're also creating and reading messaging documents, creative briefs, and product fact sheets. And sometimes, we're all short on time, and any help with consumption can go a long way. Meet Ryan Little. He's a director in a midsize global marketing organization and works on digital marketing solution. Ryan established himself as a thought leader within the company.

Research shows that the magic number for blog post frequency is about two to four posts per week.

Ryan needs to come up with the content ideas very quickly so he can start drafting the blog post due this week, and he needs to work with multiple collaborators like his product team and legal team.

Ryan is writing a blog on digital trends in the retail industry focusing on topics such as customer expectations, innovations, employee empowerment, and data utilization.

So let's see what Ryan can do here. Ryan opens up the PDF document in Acrobat, and he doesn't know what is included in that document.

He starts off with the GenAI summary to skim to see what's available in this PDF. Summary is a concept that is just synthesizing information from multiple different documents, starting with a single PDF. But eventually, our goal is to expand to multiple documents and generate and communicate the synthesis with a wide variety of stakeholders.

Ryan wants to get the key insights and summarize them and create his blog post. Typically, in the interest of time, there's so much content for Ryan to consume that he just picks five or six points that are important to him and they would resonate strongly with his audience. As you can see, then, we have created a summary of the document right here, and we are using a proprietary technology to create this summary, as well as the different table of contents. So here's a very nice outline of the document. And as you can see, we're creating section headings within the outline. Clicking on any of these section headings takes you directly to that part of the document so that you have a 25-pager where the options today are really to basically read and crawl and read everything from top to bottom. But here, this allows you to quickly open and close each of the sections, jump to those sections, and see what the section is talking about.

What I really like with these results is that Acrobat is not just giving me headlines, but it goes into a deeper and provides a little bit more framing of what is included in that particular section. So as you can see, that I have the table of contents of the documents summarized here, the executive summary, I can see this talking about how retailers are striving for personalization, for communication. And then we can go to the next section. It talks about how the importance of the integrated systems help the retailer employees to be able to transition from omnichannel to a more unified commerce system.

So these are some of the points that I want to bring out in the document, and then it also, kind of, details where the conclusion of the trend report is. As you can see, that it is basically taking this document and nicely synthesizing it down to a single pager, with all of these headings that are nicely tagged, and you can see that each of them has a Copy button here. So if I wanted to copy some of these ideas in my blog post, I would be able to do that.

So now that Ryan has nicely oriented himself in terms of all of the different sections of the document, has a fair idea what it's all about, he will go delve deeper into the document by using the AI assistant.

This is where, basically, summary plays a very important role, and it plays as a springboard for Q&A. To get you started, we have kick-started your process with three suggested questions and these get generated for every document. So you can see that it talks about what is mentioned for the Walmart study in Section Eight.

So it gives me a quick paragraph of what that's all about. And one of the things that I wanted to bring to your attention is these numbers, and these are what we call the attribution. And because Adobe's custom attribution engine is basically a proprietary technology which helps us with generating citations. So basically, you can easily verify the source of your answer. So as you can see, I can jump through different parts of the document and I can see where this comes from. This additionally helps build trust in the reading experience within Acrobat.

And just with the rest of the tools, I'm still able to basically use my highlighter, I can continue to add any comments, for any other person that I'm sharing this report to as I'm consuming this. And then you can see that if I had this question around, what percentage of marketing professionals are struggling with workflow issues? It's starting to generate response, and in no time, I should be able to get...

whether the information exists in the document or it doesn't.

So one of the things that is a new pattern here is that we're not all used to writing these questions within the document itself, but this is a behavioral change that is needed of us that how we interact with documents, how we ask them questions, et cetera.

So now you guys will see the magic of this AI Assistant. So I will basically copy paste the request that I have, which basically asks the assistant to write a blog post that's and it tells exactly where it would be posted. I've even specified the length of the post so it doesn't overwrite in terms of number of pages. I've specified the tone, and then, it doesn't need to have a section header. So I'm very, very clear in terms of exactly how I want this. And given my experience with writing a blog post, I'm able to write this prompt.

So as you can see, that it provides me with a very nice, overview of the document. And this is something I'd still do our light view edits on it. I will definitely review what it started, but this is a great starting point.

And the best part about it, I can leverage these attributions to different parts of the document just to verify the information. So I'm going to go ahead and copy this...

Blog post...

And copy paste all of the content here within LinkedIn. This is where I post all my blogs. So I can delete this title since I've copy-pasted it. I've set this up. And yeah, as you can see that this is a very neatly written article that highlights all of the different parts about all of the trends that are there. And I'm ready to post this very quickly. It even generated the hashtags that are needed for this report. So I hit click Next, and then I'm ready to publish it.

So basically, you can see that I have another document that I need to do this week as well. And I've been doing something that would take me generally somewhere between an hour to hour and a half to consume the trend report and write this blog has basically taken me less than 10 minutes to get accustomed to the different sections of the document and being able to write the blog posts. And generally, I usually have my TLDR moment, which is too long, didn't read, kind of moment, but now I don't have to have that. And I can consume the document very quickly.

And the next critical function in the organization, which is your HR function, we are seeing a lot of use cases as well. And these are around policy consumption and drafting, creating of them, and also consuming the onboarding and offboarding packages.

Then, of course, there is a consumption of a lot of training material, employee agreements that are also very critical.

And I'll just share a use case that we've had within our company. And so here I meet Elif, who is the HR, Head of Diversity and Inclusion, and as a diligent HR lead, partners a lot with our legal and compliance department. She is responsible for navigating through complex regulation updates that happen that can impact the organization in various geos.

And she routinely involves scouting of policy and regulation updates. So one of the things about this is, this requires her to meticulously read extensive regulatory documents from start to finish, and she needs to make summary notes, and she needs to be very, very detailed oriented about these documents. So after comprehending regulatory changes, she must formulate a series of actionable next steps to communicate to various departments within the organization. And at this point, this whole process is very manual and effort-intensive. And so with the assistant, as well as the GenAI summary, as I showed you previously, Elif can just bid farewell to exhaustive reading top to bottom of extensive documentation and manual creation of actions and all of the email communications that she's responsible for. I mean, she basically asks the assistant to swiftly highlight key changes that are articulated in the document and she just streamlines the whole comprehension process. She can also ask assistant specific questions around which policies and employment changes impact her department specifically.

You can easily see how this will be beneficial in the finance department, where people are reading financial reports, analyzing company earning call transcripts, audit reports, and generating combined reports. Your classic 10-Q earning call transcripts that perhaps an investor relations person on the team has to go multiple times, this becomes a very useful tool.

We're also seeing use cases pop up in banks where compliance is very critical for them and credit and loan documents are being evaluated against a very specific criteria that they have actually altered via set of questions or using them against the PDF.

In terms of adoption, AI plays a critical part even within the IT department. We find them that they have moved past the piloting and limited adoption, and IT is one department that is leading the charter.

Here are some of the examples of documents that our IT uses these features to learn from. Some of them include: learning from previous and unresolved support cases, providing intelligent scripts for agent interactions, whether it is reading manuals and enabling natural language troubleshooting. These are some of the examples that we've heard from our customers within the IT department and Adobe's IT department itself is using it for this purpose.

Whether it is looking up incident reports and combing through vendor agreements, being able to read PDFs at scale can really, really empower some of the members of this department.

Procurement function is involved in executing business processes based on information that is received from a set of documents.

Here are some of the examples that the procurement function uses this for. NDAs, finance agreements, whether it's your purchase order, vendor onboarding, supplier compliance, and statement of work. We can totally see this being applicable because some of these documents are long, and again, the system can really help them dig through the questions within these documents to get to the information that is specific to them.

Adobe has been investing in AI for over a decade and we are committed to ensuring the safety of our AI models with accountability, responsibility, and transparency. You can read about our content authenticity initiative where we have formed a coalition with several tech companies and are working with White House to bring it to our flagship products in Photoshop, Firefly, Acrobat, and others.

For Acrobat, the idea of taking an agnostic approach to LLM integrations is very critical for us. We are curating the best technologies, partners, and models to deliver the right output for your needs. For the two features that I have demoed here, we enhanced Microsoft's Azure OpenAI Service with proprietary AI and a custom-built attribution engine, and also includes 30 years of PDF expertise at Adobe.

This enabled Acrobat to deliver useful results to a wide variety of questions with accurate attributions and response structure. And this is the conversational experience that we've built in these two features. Adobe leverages Azure OpenAI's content filtering service that allows us to moderate hate speech, sexual, violent, and self-harm content that may be present in these documents. For both these features, the Acrobat GenAI service extracts all of the content from the PDF and caches it to a corresponding session key for 12 hours, and this allows you to use both the features in a seamless way.

We adhere to enterprise-grade standards for data handling, security, access control, privacy, compliance, and governance. At the heart of this technology is Liquid Mode, which is a proprietary AI and ML-powered models that allow us to identify the different elements within the PDF. Whether it's looking at the headings of the PDF, the paragraphs, the images, lists, tables, and so on, it allows us to understand the hierarchy, ordering, and makes it possible for us to reformat basically, a static PDF into a more dynamic, customizable experience. This is what we had launched previously and it allowed this is the foundation of some of the work that we've done across these two features.

So far, I've shared with you what we've delivered in the beta.

Just to give you a flavor for what else is cooking around PDF consumption, I'll start to share with some of the things that we're thinking across three different pillars. One is around PDF consumption, the middle pillar around actual creation, and the last one is around document collaboration.

So around the pillar of consumption, we would want to support scanned files. We've heard a lot from our customers that a lot of PDFs are actually scanned documents, so we want to be able to support those. The next thing that we're looking at is the ability to support multiple document where you can have a conversation across multiple files. You can actually open them in Acrobat and start asking questions across a set of files.

You will also have the additional ability to save these conversations via persistent history. The idea is that you should be able to replay the questions that you'd asked or just for the sake of keeping track of which questions and answers you received, you want to be able to refer to them later.

Next, like I said, in the creation process, people do a lot of view as of content of PDFs. That's one of the high-use cases for Acrobat, as I've mentioned before. So how do we supercharge that repurposing of the document? We are looking at ability to select text and transform it for reuse. We're also looking at ways on how GenAI can help with collaboration. As you're making comments within PDFs, you should be able to take action items, key takeaways from the feedback that's received in the PDF.

Our vision is that users should be able to tell the application exactly what they want to do, whether it's consuming the PDF, whether they're putting signature on it and giving deadlines, we want to make sure that this whole conversation with Acrobat becomes more and more conversational.

What I'm about to show you is not part of the beta, but this gives you a, sort of, flavor of our vision for the AI assistant as we're working through some of these exciting different pieces.

Writing a marketing brief involves distilling a lot of documents into a simple actionable point of view to help you seize new opportunities. With generative AI in Acrobat, individuals and businesses can engage with their documents using natural language to significantly accelerate their time for knowledge. You've opened a report in Acrobat, and your trusted assistant is here to help you. It starts by providing a summary and key insights upfront. You can either ask your own questions or use suggested questions to help you quickly start understanding this document. With just one click, you've got the answer to your question. And with sources cited, you can easily see where this answer came from in the document, like, this table on page four. Now you want to learn about a topic from different sources. Instead of having to go through many individual documents, your assistant can do it for you. For example, you ask, "What are best practices for CTV advertising?" It responds with information across multiple documents and document types, like PDFs and web articles. You can even click to view and verify sources. You or your organization can define permissions or data access controls. It can also help you identify patterns, draw connections between documents, and highlight other details that may be relevant to the report you're trying to write. Instead of taking multiple hours or days to review and comprehend these long documents, you've now synthesized, brainstormed, and developed a point of view in much less time. After synthesizing information across multiple documents and coming up with a point of view, you are ready to create a report. You can simply ask the assistant in Acrobat to create a report based on your conversation, and within seconds, a richly formatted document is generated. It suggests a white paper format, but you can choose to change it to a presentation, summary email, or web article, and then further personalize it. Next, you're prompted to apply a style using design templates to help advance the visual appeal of your draft. Once you select an option, it's automatically applied, instantly transforming the document. And your linked brand library, which includes your company logo, is also incorporated. If you want to replace the image in the center, you can select an image, enter any prompt into Firefly, and choose from multiple image options. As you review the text, you see that the overview is quite long. With Acrobat, you can rephrase and shorten the copy effortlessly. The document now looks perfect, and you are ready to share it.

So this is all of the exciting stuff that we have going on. We launched our public beta for Acrobat Standard and Pro customers with Individual and Teams subscriptions, and also Acrobat Pro trial users last month. And you can visit this URL to get access to it. We're also running a Private Enterprise Beta, as well for Enterprise customers. And if you'd like to participate in the program, please contact your account executive for access. Again, this is available in English, and this site is also available on both desktop and web as well.

I'm really, really excited to share these two features with you and see if there are any questions in the audience. [Music]