← Lead With AI
Lead With AI
Former Apple AI Engineer Brings Privacy-First Automation to Document Management
Former Apple AI Engineer Brings Privacy-First Automation to Document Management
Loading summary
Transcript80 lines
- [00:00]
A
Today, we're diving into a company that's redefining what productivity really means in the age of AI. Nitro software started as a simple way to handle PDFs, but under the leadership of CTO John Fitzpatrick, it's evolved into something much bigger. This isn't just about reading or editing documents anymore. It's about AI that understands your work, automates the routine, and gives time back to the people behind the screens. Let's get into it.
- [00:34]
B
Welcome to lead with AI. I'm Dr. Tamara Nall. In each episode, we will take you behind the scenes with visionary leaders shaping the future of AI across public and private sectors. Join us as we explore groundbreaking projects and innovations that are transforming industries and making a real impact on people's lives. Let's dive in.
- [01:02]
A
Hi, everyone. Welcome back to lead with AI. I'm your host, Dr. T. And I always like to start by saying I am so thankful for all of you who listen. Because of you, we hit number one in technology on Apple podcasts as well as we're a W3 winner. It's because of you. It's because you tune in, you like you, follow you forward. And it's also because of wonderful guests like the one I have today, John Fitzpatrick, CTO of Nitro Software. John, welcome. How are you?
- [01:35]
C
Hey, Tamara, I'm great. Thank you so much for having me on today.
- [01:39]
A
Absolutely, absolutely. So let's get into it. But before we get into the software itself, the AI power software itself, tell me a little bit about who you are, what's your passion? Who are you at your core and how does that bring Nitro into the equation?
- [01:59]
C
So I'm John Fitzpatrick, I'm the CTO here at Nitro. I've been involved in the AI space for over 10 years. Prior to joining Nitro, I was at Apple where I was responsible for the tooling and pipelines to build a lot of the on device AI models. And I came to Apple through an acquisition of a startup called Voices, where we were building kind of full Stack Voice AI solutions.
- [02:23]
A
Okay.
- [02:23]
C
And before that I was at, you know, a varied number of other companies, from startups to larger businesses. So in both the cybersecurity and telecom space.
- [02:33]
A
Awesome. And so tell us a little bit about Nitro. What is it?
- [02:38]
C
So Nitro is. It's a well established business. It's a trusted provider of PDF document management solutions and E sign solutions. And our customer base ranges from individuals, small mom and pop shops, all the way up to extremely large enterprises across varied industries.
- [02:57]
A
Got it. And everybody knows and they tune in because I really like to focus on AI powered, AI driven products and that Nitro is. So tell us, how does the AI power Nitro and how is it different from any other document and PDF company?
- [03:15]
C
So actually, I guess what most excited me about coming to Nitro was the opportunity to essentially completely reimagine how people work with documents. Right. So traditionally it's been viewing, editing and signing all of them. But I really felt through my experience in the AI space and through the really careful application of AI into this industry in order to really understand the documents and then from that be able to automate a lot of the laborious and repetitive tasks that many of our users spend just huge amounts of time doing. And I guess when I joined we first had to develop a framework for, okay, what are the things where AI can really add value in this space? And so that requires us to have a really good understanding of what our customers are doing with our products. And, and so since then we created a framework that allows us to answer exactly that question because we didn't want to suffer from that condition that we see a lot of out there, which is just adding AI into your product for the sake of AI, it's just prevalent, right. And it's just so that you can say you're an AI business. And so the framework we put in place allows us to understand which things and workflows customers would most benefit from AI assistance. And so that ends up being what may sound boring to a lot, but excites me. It's things like data extraction from your documents, it's form creation, bulk document processing, redacting within documents, summarization, document question and answering. And so for us, like these are really proven to be the sweet spots, right. So typically, and our approach to it is what are the repetitive, time consuming work that slows people down and if we can help automate them and speed up those tasks, that allows them to just focus on higher value work and work on what really matters. And so that's, that's really how we have approached this. Yeah, so not really. I mean, we work across a large number of industries, right. So it's typically a B2B. Our biggest customers are certainly in the B2B space. So pharma, legal, you know, a lot of heavy engineering and manufacturing businesses, a lot of banking and financial services and so on as well. We also do have a B2C model as well. So, you know, individuals, we have an E comm motion, people can kind of purchase our product. And so it's just Everybody who needs to do any E signing and any kind of PDF based document management, they're really kind of the customer base we have. I would say that within each of those companies then they will have lots of different functions. And so it's more function specific rather than industry specific where we would automate workflows. So you know, our new Smart Redact product, for example, which we can talk about, would solve for the legal departments within those organizations. A lot of the, within those big enterprises, our buyers are typically like the IT buyers for the business. So they tend to be our customers, but our users are everybody across those businesses.
- [06:52]
A
Right. And I would think, you know, different functions like, like you said, legal compliance, maybe even in. What is it in? Like you mentioned something about electrical where just where there's a lot of heavy manufacturing. Yes, absolutely. Okay, so tell us about a holy smokes moment where, you know, someone, a customer used nitro and they, it changed everything for them. They, they were. Wow. Tell us about that. Take us through.
- [07:20]
C
I mean, I think I've got a couple. Right. So it's a difficult question because you want one? I'll give you two. Right. So I think one of the holy smoke moments for me was last year we launched our document Assistant. And so what we had focused on doing was with a privacy preserving version of AI using private large language models to provide conversational support to any document. So, you know, I think that's a game changer. Being able to simply ask about a document, even when it's in other languages and get instant accurate answers is still, you know, it's still incredible to me. And you know, that's particularly useful for very large documents or for complex documents. Right. So I personally, you know, I'm not a lawyer, I'm in the tech space, but I have to, you know, work on and review a lot of contracts, employment contracts, vendor contracts and so. So all written in very heavy legalese.
- [08:21]
A
Yes.
- [08:21]
C
And so being able to ask questions of those and just get it in plain English back to me is really incredible. And despite me knowing exactly how it works, it still amazes me.
- [08:32]
A
Will we ever really know how it works? But yes, we just know. So can a customer. Oh, go ahead, go ahead.
- [08:38]
C
So I was going to say actually the second one of those, and this was, it was our redaction, our new redaction product. So this is something we know a huge amount of our customers do within our PDF editing software. Right. They have to redact vast quantities of documents, often in the hundreds or thousands of pages to find and remove Any potential PII data. And so we created the Smart Redact product from that. We had this aha moment when we looked at all of this data. And so what this massively accelerates customers to do is to take those documents and what would have taken them, you know, many hours to find all of the PII data in IT and redact it. Now our AI will automatically identify all the names, addresses, phone numbers and account numbers. And then it leaves the user in control. They can select what they want redacted and it removes that from the document. So transforming something that takes hours into something that we've done in minutes or even seconds. Right. So huge, huge game changing.
- [09:38]
A
Yes. And when it redacts the pii, do I have to like put in the name and it looks for it or it just says find all pii?
- [09:45]
C
It will automatically. And this is the real part of it. Traditionally, that's exactly how customers would have used our product. Right. They would manually review the document. They would know the specific names and things like that that they're looking for. So they'd search for it in the document and then redact it. This will automatically identify all personal names, account numbers, Social Security numbers, license plate numbers. Anything that is PII will automatically be detected, highlighted in the document and then you can just redact it. So really, really amazing application of AI.
- [10:14]
A
Got it. Now, when you hear these stories, like, what's your victory dance? How do you celebrate? It's like, wow, another home run here.
- [10:25]
C
No, what we typically do is go, okay, the positive feedback is great, but where are they still having pain points? Where can we add more value? Where are they still? What's an additional feature that would benefit them more? Right. So that's my perspective on, I think any good product LED business is that you think that way. Right. So, yeah, you have a quick win. We shipped it. Okay, move on. Now, what's the next problem we're going to go solve?
- [10:48]
A
Okay, well, John, that's amazing, but y' all gotta celebrate too.
- [10:51]
C
Of course, of course, yeah, yeah. A few nights with the team.
- [10:55]
A
Yeah, exactly, exactly. So how does it work? I have very curious listeners. And so if we were to open up the hood and really look at the brand of nitro, how does it all work?
- [11:08]
C
Sure. I mean, I think if you, if you open up the hood, I mean, I think the first thing I would say is you'd be reassured and confident in our approach to applying AI. Unlike as I mentioned earlier, many of the companies out there, I think, are just adding AI into their products, but they're using third party services and models. The first thing we do is we look at what's the exact problem we need to go solve. And then from that we'll choose the best models out there, we'll tailor them to our own use cases, and then we operate those in dedicated private instances. So we have different models for different use cases and they operate all within a nitro environment. So none of your data leaves our environment. We also never store any customer data or documents. And I think that's really, really important, particularly for enterprise businesses that are very sensitive to the adoption of AI. So this means we don't have your data, we process it and then it's gone. So we can never use it to train a model and it can never be exposed. And so under the hood, it's a selection of various AI models that we've carefully selected and tailored to our use cases.
- [12:24]
A
And when a customer uses nitro, can they start with like a blank document and build from there, or do they need to kind of have the document completed to do all of the redaction and all of the different capabilities?
- [12:38]
C
Yeah, so they can, they can, they can start within our PDF products and they can start with a blank document and just work on it. The workflows tend to be that the document was generated elsewhere and they're making edits to it, they're combining documents, you know, they're doing redaction and those types of operations on PDFs that have already been created. And then, you know, as you know, for things like signing, they will have created the contract or whatever else it may be, and then they will put it into our signed product to be able to send it out for signature.
- [13:10]
A
Okay, got it. All right, John, that's amazing. That is really amazing. All of the benefits, AI power benefits of nitro. Let's go to real world magic. Give us that jaw dropping moment where a customer experience. You already gave us two examples, but we are so nosy. We like examples and stories.
- [13:30]
C
Sure. I've done more, I've done planting.
- [13:32]
A
Okay, good, good, good. Let's hear it.
- [13:35]
C
I guess so for me, one of the most jaw dropping moments was it came from a customer in the pharma space. Right. So they have to process these huge quantities of patient intake forms in the clinical space. And so again, this is really boring to many people, but excites me. And I think, you know, in a lot of cases, the boring application of AI is actually the best use cases for it. So these are documents that the pharma business would have to send out the endpatients then would fill them in either manually on a tablet or on their laptop. The company would then get all of those forms back. Often they'd be generated from Word documents, these flat documents, and they then have to manually go through them, extract all of the data into a structured representation, and then do their analysis. So obviously massively time consuming. And so this actually came out of. We did a hack week last year and this feature came out of that and then made its way into our products. And so it was taking that example, use case for the customer. And our autoform solution essentially transformed that workflow for them. Right. So what it can automatically do is it'll take any flat document they have, a Word document, it'll convert it into a form that has all of the metadata, so it knows which field is your date of birth, your name, your address, all of that information. And then they send those forms, users fill them in again. They can fill them in manually and we use OC or then to understand those documents, or they can fill it in on their laptop. But when they get those documents back, they could have thousands of them. And our autoform processor then will take those and within seconds extract all the data from them, put it into a nice CSV or Excel file, and give you that structured representation. So like you can imagine the time saving that is rather than have to manually troll through all of these documents. So for me, that was a big kind of wow moment.
- [15:31]
A
Yeah, no, that does sound amazing. And I'm just thinking about even my going into my doctor's office and that's just one of like literally really have a thousand. And all the patients that they see and, and all of that and the ability to be able to aggregate that data, you know, because different people have different conditions and the, the medicine, you know, has to make sure that it can work. I don't forget what the medical term is, but has to work together and, and this is definitely game changing.
- [15:58]
C
Yeah, exactly. You know, those forms don't even need to be identical. They can be kind of disparate forms, but as long as they have similar information, our system will group that data together. And so, yeah, huge, huge time saver.
- [16:09]
A
That is amazing. That's amazing. Now you just mentioned this with this example, in terms of the volumes of data, how does nitro think about ethics? Like the ethics of having all this data and using that data to basically help the customer? I know you said that you don't keep the data, it disappears or it's erased or what have you talk to us about how you think about ethics and those guardrails.
- [16:39]
C
So I guess there's a few things there, right? So I mean, one of the things I've tried to bring to Nitro is a maniacal focus on privacy. So obviously when I was at Apple, they're known for that maniacal focus on privacy and I've brought that into Nitro. So, you know, every time we're making engineering decisions and building these products, we need to make sure we're doing it from critically, from a user preserving and a privacy preserving perspective in terms of, I mean, one of the things I said earlier is like we don't, when you send your documents in, they're processed ephemera, right? So if they say within our contained environment, they're processed ephemerally, which basically means we process it and then it's gone. Right. We don't store that data or anything like that. That obviously can make our job a little bit difficult if there's a bug or there's an issue in the system. Right. But we're willing to take that because the privacy of our users and the data protection so much more important. So I think, you know, the other part of ethics on AI for me is the whole, you know, will it take all of our jobs? And I, I don't think we're there yet. I think, you know, for me, the real power of AI and this is something I think about, which is it's not replacing humans, but it's augmenting them. Right. So that we can focus on what matters more. And so again, automating a lot of those laborious, time consuming tasks for our users and letting them work on those higher value tasks and amplify their productivity is what's really important.
- [18:09]
A
Oh, go ahead.
- [18:09]
C
No, no, go on. Sorry.
- [18:11]
A
No, I was going to say now to the whole not keeping data. I mean, you know, AI of course works with training data to get better and better. So are you saying that Nitro doesn't need that? Like I'm thinking about the pharma example, right? Yeah. Where it takes the data, it pulls it together, knows what is related to the same patient. How does it improve if I come back and visit? Does it not train on that same data related to me? Or is that really not. Okay, got it.
- [18:45]
C
No, we don't train on your data. We do things like for, you know, in, in Smart Redaction. You know, at the moment it's not released yet, but we're adding a feature where people can add in custom dictionaries. So just Particular terms they would like removed, they will be removed. So there's those types of customizations that a user can do themselves. But we don't take your data and train on it. We benefit from, you know, there's a lot of models already out there, so we run multiple different, you know, of the large language models and we just use those and run them inside our own environment. We do prompt engineering as well to get the most value from those models. And then there's open source data sets for a lot of this type of data. And we will use that to train things like our own smaller models and classifiers and so on. But we absolutely do not take any customer data store it or so we would have no ability to train on customer data. It's too important.
- [19:42]
A
Perfect. Yep. No, I understand that. Let's talk about the big future. What does a future powered by Nitro software look like?
- [19:53]
C
I mean, it depends on your time frame, I think.
- [19:57]
A
I know, I know I used to say 20, 30 or five years from now, but with AI, things move so quickly. I mean, we can hang up the phone and that could be the future.
- [20:04]
C
Exactly. I think with the rate of change, it's incredibly difficult to predict very long into the future. Right. What I would say is, I think as we see more and more dynamic AI generated content out there, I think the need for like trusted, immutable sources of truth and storage, perpetual storage of that and auditing of that data is going to become increasingly important. And I think PDF as a format actually is a huge part of that future, I believe. Right. Because it is the format that you know, when you share something, when you're finished working on something, it's the format that you share with others. There's audit logging within it so that you can make sure, you know, if it's modified in any way, you know that it has been modified. And so I see, you know, PDF evolving over time and becoming increasingly important in this world. I think within our products specifically, I think what will change is how people consume and interact with these documents. Right. So, you know, traditionally you would have opened up a PDF and if it was a really long document and you needed a particular piece of information, you might scroll through, you might search for certain keywords and so on, you know, so now we're already seeing that change. Right. People will just ask the question of where is the pertinent information in the document that I'm looking for. And I think over time our tools will just handle more and more of that administration for you and that automation. So it will present you with the pertinent information that it knows you need. It will give you the hey, here are the actions that I think we should take on this document for you to review. And, you know, I think with the solutions we have today and our trajectory and how these are evolving, that's really how I see kind of the future iteratively evolve within our tools.
- [21:54]
A
Got it. Amazing. Okay, so I've enjoyed the conversation and people want to get in touch. So what's the best way to. If people want to try nitro Today, whether you're B2B, B2C, what's the best way for them to do that?
- [22:07]
C
The best way is to go to our website, gonitro.com and so there you can. You can actually start a free trial of nitro for 14 days. You can try out all of the tooling we have available. And look, we've got some incredible tools available that can save huge amounts of time, particularly our new smart tools and AI features. So, you know, we talked about some of them, but there's a few more in there as well.
- [22:31]
A
Did you want to mention those? You can. If you want to mention them, you can. Sure, yeah.
- [22:33]
C
Yeah. I mean, I think we have, you know, so we talked about document reduction, document summarization. Within our sign product, we have a new feature which is again, another one of those wow moments where auto form field placement is what we call it. But essentially, if someone is sending a contract out for signature, they often have to go through a huge amount of work to drag and drop all of the initial fields, the sign fields.
- [22:59]
A
Oh, it's awful.
- [23:01]
C
It is, it is. And so actually this came out of discussions we had with, with some, some of our legal customers, right, where it was taking them hours to do this. And we went, oh, we can solve this with AI. And so now one click of a button and will identify all the recipients, place the right signature fields, dates, everything else in the right place. You just quickly review it and then you can send it. So again, another. Another game changer. So that's another thing. You know, if you sign up for the trial, you can try it out. So something for everybody.
- [23:26]
A
I would say, yeah, no, that's amazing because I've done that before and I'm. And then you miss some. And then they come back and say, wait a minute, I know I need to sign, but it's not there. Just a mess. So you solved it?
- [23:38]
C
Yes.
- [23:39]
A
That's absolutely amazing. So, John, of course, you are our genius that we're focusing on this week. And Last week, our genius has a question and it is, do you think AI will make us more connected or more isolated? It.
- [23:59]
C
So that's a very deep question, isn't it? I'm honestly not sure. Right. I, I think, I think probably both. Right. And I guess what comes to my mind when you ask that question is I compare it to the Internet. Right. And in many ways that's brought us much closer together. Right. It allows you to find and discover communities that you might be interested in, stay connected with friends. It allows us to do these types of things that wouldn't have been possible previously. But at the same time, there's the other side of this, which is people are spending more and more times doom scrolling and more and more times on the screens not speaking to their friends and family. And so I can kind of imagine a similar pattern with AI in many ways. Yeah, it'll take over admin tasks for you and maybe some of the higher cognitive tasks. But then I worry about things that are in the media recently, about people spending more time using LLMs for therapy instead of talking to properly trained therapists and their friends and families. And so I think like most technologies, it's going to depend on how we as individuals, businesses and society hopefully choose to apply AI and integrate into our lives.
- [25:11]
A
Yeah, no, that's true. While you were talking, there were two things I was thinking about. You said, you know, people spend time doom scrolling, which I found to be funny, which is true. Question is around whether or not we'll be more connected or more isolated. And I'm thinking, huh, I do talk less to my friends, but it's not because of, you know, the Internet or AI. It's because I'm probably calling and saying, or texting and saying, did you see this? Did you see that? And they're probably over it. They're probably over it and say, you know what? I've had my dosage of Tamara today.
- [25:43]
C
Well, I mean, I mean look at things like, you know, Instagram and so on. I mean originally they started as, you know, share, you know, you could keep up with your friends lives, but now if you're scrolling through them, how often are you seeing things from your actual friends versus, you know, what?
- [25:55]
A
Right, versus what?
- [25:57]
C
And I. Influencers and.
- [25:58]
A
Yeah, right. And I look at one thing and then the next thing I know, the next 20 videos is related to that and I'm like, wait a minute, I didn't even really look at that. So then I have to change the algorithm to get back to taking you.
- [26:09]
C
Down a rabbit Hole.
- [26:10]
A
Yeah, exactly. Okay. I love that. You're right. So basically your answer is it depends on us. And we have to, you know, be at the forefront of that to determine whether that we're connected.
- [26:21]
C
Human control is, is, is the most important aspect of these.
- [26:26]
A
Got it. So that's a, that's a great answer. So let's move into our bonus rapid fire. I'll ask you a question and then you give me the first thought that comes to your mind, starting with the most overrated tech or AI trend.
- [26:40]
C
I think what we talked about at the start, which is the superficial addition of AI features into every product. I think the other one just related to our last conversation, which is the dreadful AI generated ads you see on social media platforms, I think is a very much overrated trend. I don't know whether it's rated, but a poor trend.
- [27:02]
A
Got it. What's the most underrated AI or tech trend?
- [27:08]
C
AI trend I think is it's not probably not under hyped, but it's certainly very early stages, which is mcp, so model Context Protocol. So I think this is a super important technology that allows large language models to connect to other systems, to pull information in and interact with other systems. And so actually we're getting ready to. We're building our MCP service at the moment that will allow users to automate really all of those tasks right from within any agent they choose. So ChatGPT or Claude, but they'll be able to use Nitro's platform to automate all of those tasks by just asking. So really excited about that.
- [27:46]
A
Love it. Mcp. What about a book we should all read on the future?
- [27:52]
C
I'll cheat here. It's not a book. It's not a movie either. So it's a science fiction short story by Isaac Asimov. It's called the Last Question. So it's pretty short. I can't remember how long it takes for. It's probably like 30 minutes. But it's about very long term future prediction. So it's about humans building smarter and smarter computers over time and asking the AI the same question that it can never answer because there's insufficient data. I won't spoil it by saying anymore, but I just think it's a fascinating story about how humans and technology might evolve over time into the long future.
- [28:30]
A
Got it. And that's the last question. I'm writing that in.
- [28:32]
C
The last question. Yeah.
- [28:34]
A
Do you get it like on Amazon or Kindle or.
- [28:37]
C
I think, I think it's actually out of copyright, so I think you can actually just Google it and read it. Google it, yeah. Yeah.
- [28:45]
A
Awesome. Okay, and then my last question is scare us. Wow us. What is your biggest, boldest prediction?
- [28:55]
C
I'll give a scary one. So. Well, so it's both a prediction and something I guess I hope I'm wrong about. Right. But okay. You know, as we rely, I think more and more on AI to do more and more of the cognitive tasks for us, I think it's not quite there yet. Right. It's still when people are coding, it still requires lots of code review and things like that. But as they get better and we do less and less of that cognitive work for ourselves, both individually and as a society, I think instead of accelerating innovation in the medium term, it can actually slow us down. Right. Because we'll have less and less people in our society in the detail needed to really innovate in an area. And so that kind of worries me. To be clear, I think the net benefit will be positive. But that worries me particularly for the younger generation who are grown up now with these models available to them and so don't have to go through the pain of trawling through a book and learning it for themselves.
- [29:52]
A
Right, that quick learning.
- [29:55]
C
Exactly. Yeah.
- [29:57]
A
Got it. Amazing. Well, John, this has been such a relevant topic. I mean, we're all using documents and we've all gone through the hardship of trying to make sure that we're searching and we're putting the right blocks for signatures, dates, et cetera. So what Nitro is doing, and I love, throughout the conversation you were telling us about all these latest developments that can help improve our life and make us more efficient. So I've thoroughly enjoyed this conversation. Thank you for what you are doing. Thank you for what the nycho team is doing. What's the best way for us to get in contact with you? I mean, you mentioned the website. If you could mention that again, spell it out. Talk to us about your social media handles. The social media handles for the company and then also how people can connect with you.
- [30:44]
C
Sure. So gonitro.com is Nitro's website. So again you can go there and learn lots about our products and how we apply AI, download a trial and get started really, really quickly. I don't do a huge amount of social media, so LinkedIn is by far the best way to get me. So I'm just John Fitzpat on LinkedIn or you can just Google LinkedIn John Fitzpatric Nitro and you will find me.
- [31:08]
A
Yes. Perfect. Awesome. Well, again, thank you so much for being here. Absolutely. Yes. Yes. It was a good time, a good conversation. And everyone tune in next week for our next guest. But until then, lead with AI.
- [31:26]
B
Thanks for tuning in to lead with AI I'll see you next time as we continue exploring the cutting edge innovations shaping AI across the public and private sectors. Until then, keep leading with AI.