Subscribe
Share
Share
Embed
AI is changing how we work, but the real breakthroughs come when organizations rethink their entire foundation.
This is AI-First, where Box Chief Customer Officer Jon Herstein talks with the CIOs and tech leaders building smarter, faster, more adaptive organizations. These aren’t surface-level conversations, and AI-first isn’t just hype. This is where customer success meets IT leadership, and where experience, culture, and value converge.
If you’re leading digital strategy, IT, or transformation efforts, this show will help you take meaningful steps from AI-aware to AI-first.
Jesse Henning (00:00):
We have probably a couple hundred pages worth of travel requirements and somebody might say, "Oh, hey, how do I get my meals reimbursed?" Well, instead of having to read all 200 pages of the travel manual, they can ask that question in natural language and you can really cut right down into what the answer for that was. And this is actually really helpful because on our side, we're very much in lab operations. Our goal here at the library and on the operations side is to make sure that we can do everything we can so that scientists can focus 100% of their time on doing research.
Jon Herstein (00:32):
This is the AI First Podcast, hosted by me, Jon Herstein, Chief Customer Officer at Box. Join me for real conversations with CIOs and tech leaders about re-imagining work with the power of content and intelligence and putting AI at the core of enterprise transformation. Welcome back everyone. This is episode 21 of the AI First Podcast, where we tackle the pressing questions facing CIOs today. I'm your host, Jon Herstein, Chief Customer Officer at Box. For over 80 years, from the birth of the atomic age through today's breakthroughs in clean energy, quantum computing, and artificial intelligence, Argonne National Lab has stood the absolute frontier of human discovery. And today, we're bringing one of the people who makes that happen right here to the show. I'm very excited to bring Jesse Henning, library and information manager to you. Welcome, Jesse.
Jesse Henning (01:28):
Hey, Jon. Nice to meet you. Happy to be here.
Jon Herstein (01:30):
Jesse, I'm going to start with a bit of a softball here. What is the real tension between accelerating discovery with AI and determining what is exposed to AI? How do you find that balance? What's the tension?
Jesse Henning (01:43):
Yeah, Jon, that's a really great question. On our side, so for library and information management, our side at Argonne, we deal a lot with scientific publishing and also assisting our researchers with finding journal articles and information that help them influence their research. Where we see some tension is, is making sure that we're not exposing either the information that scientists, they're working on these new projects. They want to make sure that they're first to publish, that they're working on novel things. We want to make sure when they're asking questions of AI or they're interacting with AI as they develop their scientific information products, that that information isn't leaking outside the enterprise, either for other researchers to scoop them, that's a big deal in scientific publishing. And also make sure that we're protecting privacy and all the other things dealing with subjects of experiments and anything that might be national security related.
Jon Herstein (02:32):
Of course. Yeah. It's funny. I've heard that term scoop in the journalism world, but hadn't thought about it so much in the research world, but that's a real thing.
Jesse Henning (02:39):
Yeah, absolutely. Scientists really want to closely guard the work that they're working on. It's a lot of time. It's a lot of money. A lot of their careers are spent doing this to make sure that their findings and their ideas aren't leaking out to another third party that would be scoop their research and write the article before them. That's a big deal.
Jon Herstein (02:57):
So I wonder if you could give a little bit of that background and talk specifically about the kind of work that you do at Argonne.
Jesse Henning (03:03):
We're a US Department of Energy National Lab. The US Department of Energy funds us to essentially do the research that may be impossible anywhere else. Here at Argonne, we're at the home of the Advanced Photon Source. It's one of the ... I believe it is the brightest coherent x-ray source on the planet. These other national lab, like Lawrence Silvermore has the National Ignition Facility, these just giant, massive scientific instruments that allow us to really explore the nature of matter and things that would almost be nearly impossible in private industry. And we work with a lot of collaborators across industry. We work with collaborators across academia, and then also too internally here in Argonne. So we do a lot of that work and turn it towards questions of working with energy security, national security, creating new materials, battery chemistry, all things that are important to our world today that might require a new material or a new method.
(03:57):
That's what the national labs are here for. And then hopefully the science that we create is then picked up later by our business and other collaborators or folks just out in the open to be able to commercialize some of that and take those innovations and bring those to market and allow those to be solutions that help solve problems in the world.
Jon Herstein (04:15):
So at some point, the research that you do with the labs becomes public knowledge, and I assume that's not true across the board, but at least some of this stuff becomes public knowledge so that it can then be commercialized. Is that the idea?
Jesse Henning (04:27):
Well, so a really cool part. So we're on the publication side for scientific publications that are unlimited access. So we don't have any issues with national security concerns or things in between. Those are publicly available to everyone. Those are released to the US DOE Office of Scientific and Technical Information, and those are viewable online right now through their website. So that research can be picked up by anybody from a multinational all the way up through a hobby chemist that may be interested in what's going on in material science. And so our shop here at Argon handles a lot of that transfer to make sure that we're making that information and that data publicly available wherever we can.
Jon Herstein (05:05):
How is AI being used to explore within the research libraries and the sort of knowledge services that you provide?
Jesse Henning (05:12):
I can talk a little bit about how our research librarians are using AI here locally. And we can actually talk about an example where we're using Box. We had researchers that said, "Hey, I've got 300 of these and it's impossible for me to read all 10,000 pages of all these combined, but I'm looking for specific information I think that might be in there." What we were able to do is be able to take those technical reports publicly available, put them into a box hub and allow the BoxHub RAG to be able to pull out some of those insights so that the researchers could highlight specific sections and specific papers that they think may be useful to their research. And then what happens is this sort of feedback loop with research librarians. We provide that to the researcher. The researcher says, "Hey, it looks like there's this set of citations or this particular topic, give me more information here, allowing us to refine delivering those resources to them." And then it feeds back in, helping people refine, refine, refine all the way down until they get the particular insight that's useful for the research that they're working on.
(06:11):
It's essentially what librarians have been doing for years and years. It's the great thing about AI is it allows us to work with folks, researchers, library patrons everywhere, work with them along with AI to refine down to the exact thing we're looking for and then synthesize knowledge between all of those different pieces. And that's how we're using it here at Argonne, just in that small perspective there.
Jon Herstein (06:35):
Right. And you mentioned citations a couple of times. I would imagine in this world, it's incredibly important to be able to go right to the source so that people can actually check that. I don't know how much concern there is around things like hallucinations, but how do you think about that citation component of this kind of research?
Jesse Henning (06:52):
There's a ton of concern around that. Always being able to reference back to the actual publication or the actual source that it came from is critical for us as librarians and library staff. We really pride ourselves on making sure we have the right answer, not the right enough answer. And I think that's where we're using AI here locally is to be able to get us closer to how can we speed up the move between the ... We have no idea what this particular subject is because for most of us, I was a C+ chemistry student, but I work with chemists and material scientists on the regular, and so I have to translate a little bit of what's going on in their subject area. But how do we move from the, we have a question up to the right enough answer so that we can hand that over to our library staff to really dig in and find the true specific real piece of content and not have it be potentially AI generated or hallucinated.
(07:46):
So that's where we see AI being a partner there in this quest to find the right answer.
Jon Herstein (07:51):
Right. Are you finding that that ability is enabling you to deflect questions? And I don't know this for sure, but questions that might have otherwise gone directly to the researcher because people couldn't find it versus, well, now I can go research it myself, get right to the source, very quickly, zoom in on the thing that matters and answer my own question. Does that sort of dynamic play out?
Jesse Henning (08:15):
Yeah. For some things, we do see that happen, especially when it's kind of step and fetch. So the cool part is that the scientific publishing and the way that specialized libraries work, there's a lot of fore knowledge you have to have to be able to dig through and understand them. We try the best we can to make them accessible to folks. But the great thing with AI is that AI is actually, it knows those processes. It knows how we've organized our scientific collections and how the publishing system works. So it allows people with natural language to zero in on what they're looking for there. So we see there's some case deflection there, but what we actually have seen is that by kind of eliminating some of the small, I need to find an article in a journal, easier questions, with those being answered partially by AI and advanced search, we now actually have a ton of time to really focus on the tough questions, the things where we deal a lot in what's called a gray literature.
(09:09):
So these are things that maybe the internet doesn't know. We have microfilm and microfiche and paper articles. And sometimes we end up, there was one case where we had to call a Coast Guard commandant to ask them questions about records that were not on the internet about a shipwreck. It gives us time to focus on the really tough finding instead of having to split time of those really, really difficult questions with some of the ones that are simpler that might be able to be handled by AI or search.
Jon Herstein (09:37):
Right. That is fascinating. And it's a very common theme that I've heard across pretty much all the conversations we've had on the podcast where AI, there's always this fear of, is AI going to replace the people doing this work? And the answer is, well, no, not really, because what it's doing is actually freeing up that human resource to do the kinds of things that you just talked about that AI probably wouldn't be able to do, right?
Jesse Henning (09:59):
Yeah, exactly. It also too gives us the ability to handle a lot of those things, even just with our own personal research. It allows us to get up to speed much more quickly when we're working, let's say we're working in a specific scientific domain that we don't know a ton about, or if it's a particular historical item that we need to know rather than spending maybe on a hard finding question where we would've spent an hour or two or maybe a day digging through technical reports to find context, we can use AI to cut to that context immediately, give us the brief that we need to be able to dig in on the tough question. And so it really minimizes that amount of prep time that we need. And it also helps us understand what scientists are asking for because now that we have the brief, we can speak their language, which is great.
Jon Herstein (10:43):
So that sounds like a great example. Are there other examples where you're seeing real true non-hype value from AI in the research environment?
Jesse Henning (10:51):
This might sound a little silly, but for researchers and everybody here at the lab, a lot of researchers end up presenting a lot of their talks at conferences and symposia across the world. And so we actually have a pretty brisk travel department that handles and arranges all of that travel for all those scientists everywhere they go. And because we're government affiliated, we have a lot of rules about how you're supposed to travel and not travel, not things that go in there. Where we've actually seen a lot of really cool non-hype value in sort of a small cases is that we have tons of documentation about how we're supposed to travel, rules and regs and requirements and data sources and all this. And what we've been doing is curating those sets together and then putting them into chatbots so that folks can ask those natural language questions.
(11:39):
So for example, we have probably a couple hundred pages worth of travel requirements and somebody might say, "Oh, hey, how do I get my meals reimbursed?" Well, instead of having to read all 200 pages of the travel manual, they can ask that question in natural language and you can really cut right down into what the answer for that was. And this is actually really helpful because on our side, we're very much in lab operations. Our goal here at the library and on the operation side is to make sure that we can do everything we can so that scientists can focus 100% of their time on doing research. That's what they're here for, not filling out receipts or having to deal with hotel reservations and all that and answering those questions. And so the more time that we can use to cut that away is the more time they can work towards actually answering these scientific questions that take the majority of their time.
(12:32):
So we've been seeing a lot of non-hype value kind of in those very, I don't say boring, but just very mundane business cases where you've got a process that's there and you can ask a question.
Jon Herstein (12:43):
And it's the sort of activity that those folks probably don't want to be engaged in. If I had to spend an hour sorting through my travel versus an hour doing more research, I think I know what the answer would be.
Jesse Henning (12:54):
Yeah. Unless you really like picking hotel rooms out, I mean, there's that.
Jon Herstein (12:59):
And it's also a very common use case that we see is just this idea of a policy hub or a SOP hub that can be queried. And that helps with two things. One is what you mentioned, which is not having to go find that document within the 200 or find the right page, the right paragraph in a 200 page document. But the other is if people don't want to do that, typically they're calling someone or emailing someone or Slacking someone to ask that question and it's then taking time away from that person. So it also frees up the team as well.
Jesse Henning (13:26):
We're seeing that in scientific collaborations and how we're able to ... If you have a group of researchers working across the globe, you have meeting notes, you have all of your interstitial research products. And again, making sure that people are all caught up on the same page rather than having to meet again when somebody was gone for a little bit, you have all the information there. You can ask, what are the changes that happen between X and Y date or who's working on this or what are the new, what happened at last week's meeting rather than having to reconvene and start your next meeting with 30 minutes of catch up. And we see our researchers using that a ton using these collaboration platforms and AI enabled tools to help facilitate that, getting everybody up to speed quickly.
Jon Herstein (14:07):
Right. And you probably have the perfect population for that in the sense that I would imagine if you're a scientist, you're very used to writing things down, taking notes, documenting everything, and you now have all of that as a source for answering those kinds of questions.
Jesse Henning (14:22):
Yeah. I think what was it in mythbusters? The difference between science and screwing around is that we write it down. Maybe don't quote me, as not a scientist, that's at least how it looks like to me.
Jon Herstein (14:35):
We got to get that into our tagline somehow. I'm going to work on that with our marketing. So you gave a couple of great examples, but are there other specific workflows where you've found that AI is proving even more practical than you expected it to?
Jesse Henning (14:46):
One of the processes that I mentioned was, and it's probably familiar to a lot of other folks, but it's where you just need document review and document approval. So for us in our shop, what we get is scientists submit their documents to us, whether it's a journal article or a conference abstract or a slide deck. And it needs to be reviewed to make sure that it matches our scientific standards, matches standards of rigor, but that also too, that we're making sure that all of the national security, export control, anything that may be related to their research, that all those rules are followed and we're making sure we're not exposing the lab to any additional risk. And what happens is that sometimes these submissions have to happen really quickly. You get a notice that abstract of yours has been accepted and the conference is in a week and you've got to get your slides together.
(15:35):
And for us, we have to try to find a way to speed that process where how can we make sure that all of the right folks review it, make sure that when they review that those reviews are quick, they're able to focus on the sections that matter to them. So what we've been doing is experimenting with looking at agentic workflows, both with our behind the fence secure AI gateway, and then also too with some work that we're doing in Box that's helping us speed that process along so that it's assigning tasks to folks for a particular, let's say you're an export control analyst, you get handed a 200 page technical report that needs to be presented next week. You're not going to be able to get through all 200 pages in time, but what you can use is you can use AI assisted review to be able to focus on the key areas and hopefully speed that review along.
(16:18):
Still keeping the human in the loop, still making sure that we're the ones that are making decisions, but to help us kind of wade through some of this, especially in our space or if it's controlled space or national security space, there could be single sentences that could pose pretty substantial risk. Similar to, I'm sure that a lot of other folks using AI for contract reviews, all those things, that's serious stuff and a little bit of language can make a big difference. And so making sure folks can focus in. So we're looking at building those into our publication review and release process here.
Jon Herstein (16:47):
I think what I understand is that you're saying a single sentence could be problematic and it's not so much that you're relying on AI to take it out, but you're relying on AI to more quickly get to those things so a human can look at it and say, yes, this is problematic, or actually that's fine.
Jesse Henning (17:02):
Yeah, exactly. Amplifying that signal to noise. And we super care about the signal that anytime that we can reduce that noise and amplify that signal, because we've got our technical reviewers, they know a ton of stuff. They are extreme experts at what they do. And so by making sure that we're not spending that expertise on things that may not be problematic or high risk and really allowing them to focus on the things that pose actual risk, we're making sure we capture those and we manage them before they become issues.
Jon Herstein (17:31):
I'm sort of curious how you think about the role of AI more generally in relation to the content governance world, given how important it is and given the volume of research, output, documentation and collaboration that you have, that's got to be incredibly important. So can you give us a sense of how you're thinking about AI's role in this area?
Jesse Henning (17:49):
I think that AI is really important for being able to manage content and combine and move and identify that stuff. But I think about the human element for content governances is that it still requires folks to be able to provide context. AI doesn't thrive without context. And so where I see AI is we set the context, we set the framework that it's going to work within. We create metadata and we create our workflows that are very human, and then we set AI towards those and have it work within that bubble. We've been doing that a lot over here with how we work with just defining different types of scholarly documents and how the things we want to find and the things that are important to our lab leadership, things that are important to researchers, we are enriching those documents with that context and we're helping for telling AI, these are the things we care about.
(18:41):
So explaining that, setting the ground rules, and then allowing the AI to manage the governance according to the rules that we've set.
Jon Herstein (18:49):
Have you started work using AI at sort of tagging content with metadata, for example, to help structure the context that you're referring to, or is it all sort of unstructured context today?
Jesse Henning (19:00):
In some cases, we're using it to extract metadata according to templates and things that we've built. I think about this as a professional library, and I always think about it as if you go to a library, it's our job to get the shelf, to put the things on the shelf, to put the spine ... Not all librarians work with books, but this is something that is common for most folks. We set the rules for how things are findable and how you can locate them and what they're about. We don't write the inside of the book, but we help get it all categorized and put into place. And AI is kind of the same way that if you just came into a library and nothing has spines, the books are just thrown on the floor, there's no labels, they're not even the same color, AI is going to approach that.
(19:43):
And even though it can read really, really fast, it's not going to have a good idea of that context, how these things are related. What's the difference between the fiction section and the nonfiction section? Because the two of them sound alike if you don't have that context. And so what we've been working a lot on is try and provide that context. We build those frameworks, we build that metadata, we instruct AI on how this is important. Not too long ago, it was a question about looking at procedural documents for records to find out what record documents are generated from a particular procedure or policy. And we said, "Go and look for the records." Just the first pass. And not knowing anything, it pulled out every ... We would've had to keep 300 different records when we asked AI to say like, "Hey, what's this? " And then when we told them, "No, we're actually looking for a very specific set of records defined by these particular criteria." It was like, "Oh yeah, I know what those are.
(20:31):
" And we were able to pull out the metadata. So I feel like that's the tangle here in governance is that AI is very, very good at executing on rules. It's just developing those rules and designing them so that they make human sense and communicating them to AI. Then AI can execute really well there. And I think that's really important for governance. You have to be very clear about what you're trying to govern and what is meaningful to you.
Jon Herstein (20:53):
How do you think about, as you look across all the things that are happening with AI at the lab, what do you bucket into this is working, it's repeatable, it's scalable, and we consider that true production versus we're still experimenting, we're piloting, we're figuring out what the capabilities are and haven't yet committed to moving into a true production mode. How do you think about that?
Jesse Henning (21:13):
Oh, well, I'm not a cybersecurity expert, but I work alongside cybersecurity experts. And that is the thing that keeps them up at night all the time about those different things. I think where we've seen things here at the lab that are AI solutions that go to production is when they are scoped and they are contained and they work within those sort of governance frameworks that we've already set. We've had a lot of success using Box here because it's all inside. It's kept inside the box pun intended, working with BoxHubs and Box AI that all stays within the fence. So that's where we found a lot of those solutions becoming durable when they're well governed from the beginning. I'm probably a bad case, especially too as we have access to some of these new code tools. I've been kicked off a few APIs because I'm not a software developer, so I'm just telling Claudco to invent things and it does, but it doesn't know to be polite all the time.
(22:11):
And so for me, I think that those sorts of things is where it gets out on the edge where you're working with AI that can develop nearly anything for you, but it may not have human sensibilities about like, "Hey, how does software design work? What does a shipable product actually look like? " That's where you need to ... It's actually, I found it to be really great because now I can at least begin to speak kindergarten level software to our developers and I can actually now learn something. We're kind of speaking a similar language and we can learn from each other. I think that's the main difference. If you can implement AI within a framework that you really can't do a ton of damage, those things tend to stick. And sort of the experimentation side, those maybe become experiments that you turn and say, "Hey, this is a great proof of concept.
(22:57):
Let's refine that. Let's work on that with a developer." And then those things tend to turn into real products, especially here in the labs. There's not a lot of entirely ... I would say there are no AI only coded things. We always try to make sure we're reviewing and checking.
Jon Herstein (23:14):
So there's a few steps along the way between, "Hey, Ivycode is something really cool to let's get that into production." Is the experimentation happening more in a top-down model where someone in a senior role saying, "Hey, there's an area where we think AI can be helpful, Team X, can you go look at this using some of these AI tools we've made available?" Or is it the opposite where teams come up with ideas on their own, develop some things, and then they sort of bubble them up to leadership or maybe both? Is it a hybrid model?
Jesse Henning (23:43):
Yeah, I think it's both. And it's really funny because we have usage statistics and we see for some of our business AI tools, senior leadership is our heaviest adopters. Folks that need to take a lot of textual information and kind of combine it down and get the bullet points, get the things so they can make their decisions. So they're heavy users at the top. So they do find those sorts of, "Can't we use AI to do this particular thing or speed this along?" But also too, because of the wide availability of AI tools that we have here at the lab, we actually end up with a lot of folks that have cool use cases that have come out. So when I was mentioning about the travel guide that we were talking about earlier in the podcast, that came up because we had somebody that said, "Well, can't we just make a chatbot to look at this?
(24:27):
" They were introduced to box hubs and they say, "Can't we make a chatbot that has all of our documents in it? " And then they just started working on it, started refining it, curating the documents, curating the responses, and now it's in production. We have that chat, the embedding of the BoxHubs chat exists on our internet. And so when people ask questions, it's right there. And that didn't come from the top. That came from our travel group that said, "Hey, we're answering a lot of questions over and we spent a ton of time on this manual and it's really, really good and we're answering a ton of questions that could be easily answered by sections in the manual. So let's take our good work, feed it to something that can read our good work and help communicate with our end users." We see that all over.
(25:12):
That happened even with our side on the library, we're interested in how can we interface our library search with natural language search. So this is the behind the fence search of the material that we have here that may not be publicly available. How can we use that internal search interface and interface it with our AI and build connector, build a skill, build an MCP, build whatever it is to connect those two pieces together. And we just kind of try stuff out. And if we think, "Hey, we've got something here." The best part is that we have support. We can hand this over to developers. I always say, call an adult to go take a look at this thing, see if it actually works and then move on from there. So we've kind of seen a lot of use cases across the enterprise. And always, I think for everybody has this moment where when you're working with AI, maybe you're skeptical or you don't think it's useful or it's a toy or it may not matter for your work.
(26:04):
But we've encountered a lot of times where you see a use case where the key just turns in the lock and then it inspires people. For me, it was writing code. I don't know anything about writing code. I know a little bit about how the internet works, but I don't know anything about writing code. But when I found that I could write Python scripts and I can get the computer to do things that aren't specifically coded that I didn't have to buy, that turned the lock for me. I was like, "Oh my gosh, this is amazing. I can simplify this. I can compare Excel spreadsheets. I can do all that stuff that used to be super manual." And I think that happens with everybody across the enterprise. And it happens when you allow people to have those tools in an environment that's fairly controlled, fairly safe.
(26:42):
One of the big things when we were talking about rolling box hubs out to folks is you can't break anything.
Jon Herstein (26:46):
If you could wave your magic wand and you were asked to shut down half of the pilots, the AI pilots specifically that are going on in the institution, I'm not going to ask you which ones you shut down, but I will ask you, which ones survive and why? What would your thought process be behind saying that one stays, these other ones maybe not so much?
Jesse Henning (27:05):
I'm glad you didn't make me the executioner for some of those. The ones that I think that I would encourage are ones that have that context and that curation baked in where it's you don't have the rough edges, maybe if you didn't have a really well curated dataset, you would get inconsistent answers or things would come out strange. And then for people who are just on the edge of AI adoption, that sort of it doesn't work, blows the whole thing up. So you want to make sure that if you've got an AI solution that you're building it on top of something that has really tight curation, really good metadata, really great description and a very, very clear set of rules around what the expected output is. For example, I wouldn't say, "Oh, we have a thousand technical reports across all of Argonne dealing with all kinds of different subjects.
(27:57):
They're all from different time periods. They're all from different authors, all about different scientific technologies. We're going to throw them all into one place and now we're going to ask questions about them." That's not going to work. I would not advocate for that because you're going to get inconsistent answers. You have no context. Encouraging AI pilots that are really based on well-described data sets, I think is really great where I would encourage these sort of small scale AI interventions. It's like, how can you help somebody save an hour of their day every day? And that's everyone. So things like, they may not require this gigantic graph database or all these data pipelines, these just like quick tricks that can help everybody solve some of those daily problems. I would see those pilots would be the ones that I would encourage. And especially if you go out to the subject matter expert, the person who's doing the work every day, ask them, "What if you had a really, really smart, super obedient, but very literal intern?
(28:55):
What would you offload to that intern?" And then they'll be like, "Oh yeah, I'd rather not have to do this. " And then working on those small use cases that save maybe an hour of time, but that everybody has to do like handling, filling in time on your timecard. I know that sounds so silly as just sounded like business cases, but it saves everybody a ton of time. It removes kind of that drudgery that allows us now to have, not to say, I mean, I love my timecard, sorry, HR, but to remove that drudgery so that we can spend our time doing real human creative effort or human relational effort with each other, get rid of sort of the bookkeeping. Also, sorry, bookkeepers, I love my financial people, but get rid of the daily bookkeeping and really focus it on what humans do best, which is relate to each other and be creative.
(29:41):
And so I would encourage things that help us save time in those small ways for everybody.
Jon Herstein (29:46):
Are librarians becoming AI governance stewards or unstructured data stores or both?
Jesse Henning (29:51):
Always been. We take the best that humanity has to offer, our written history, the thing that allows us to look forward is the thing that allows us to look Backwards. And we take all of that and we get it into a way where people can find it and understand it and retrieve it and preserve it. We've been doing that, whether it's in physical space, whether it's in digital space, where we fit in with AI is again, communicating that value and almost working with AI as if they're a consumer or a library patron or a user that's coming to us. And in libraries, it's all about how do we lower barriers to entry, whether that's at your public library, making sure that you can get a library card without a bunch of fuss, all the way up to researchers that say, "Hey, I have a new LLM that I'm building in- house and I need to feed it technical reports.
(30:41):
And how do we get them in? It needs to be through an API or some other connector. How do we build that? " So it's both sides, stewarding the usability of the information and then also how we can deliver that information to the user. I know there's it for people who are listening, look up the five laws of librarianship. We
Jon Herstein (31:00):
Should put that in the show notes.
Jesse Henning (31:02):
Yeah, for sure. Yeah. It's like every book it's reader, every book it's user, it's really interesting. And it applies now more than ever.
Jon Herstein (31:12):
Jesse, I just want to say thank you so much. This has been a great conversation. I've learned a lot. I think our viewers and listeners will as well. And I really appreciate your time with me and your partnership as a customer of Box and looking forward to continue to do many, many great things together. Thank you.
Jesse Henning (31:29):
Yeah, this is awesome, Jon. Thank you so much for the opportunity.
Jon Herstein (31:33):
Thanks for tuning into the AI First Podcast, where we go beyond the buzz and into the real conversations shaping the future of work. If today's discussion helped you rethink how your organization could lead with AI, be sure to subscribe and share this episode with fellow tech leaders. Until next time, keep challenging assumptions, stay curious, and lead boldly into the AI first era.