Driving smarter business decision making with the right data insights solution
We're going to talk about. What. Are AI, in ml and how did it work and I think it's really important to start off that AI, in ml is really there. To to. Augment, the. Lives of humans it's, really to focus on high-value, activities. For them and it's not about replacing humans, which is sometimes something that people, seem. To be thinking and that. This is going towards there if. You look at artificial. Intelligence. And what. It actually and a machine learning of what it actually is. Then. Then artificial. Intelligence is really around. Using. Software, to think. And act like humans, and and then there is a subset, of that where you, could really, go. Into. Using. Machines, to, actually. Work. With algorithms, to, improve and, learn faster and the, more data you have the more accurate becomes. And then, there is another subset. In the, artificial intelligence, world where. You. Talk about deep learning and deep learning is, is really mimicking. The brains of humans, and using. Multi-layered, neural networks. So. Why, am L an AI if, you look at this statement from, 1982. We're. Drowning in data but star for knowledge so this is still true a lot of organizations. Have a lot of information where. They don't know exactly what they have where they have it they've got a hard time finding it and so. Although. This. Is still very true. A lot, of organizations. Only. Use 10% of their their data this is coming out of a lot of research that's been done by different organizations as, well and in. The end. This. Is from 1982, the sad part is is it, is from 1982, and still, the case we haven't really, sold it yet so. How, can we use artificial. Intelligence, and machine learning to, unlock inside, so there's, a lot of unstructured, data out there and in, order to really, work, with this kind of information. You should really, try to take. Out the key metadata, and put. It and classify, it and to put it into baskets, so to say and then, you. Make, unstructured. Data you, make it structured, and then in structured, format so you can then start, to analyze. It and work. With it of course and then, in the end you. Want to integrate it into your business processes, to drive strategic. Decisions, and the, good thing about, the. Current world is that we use. All. Kinds, of clouds out there and with the power and the computer, of those those, clouds, we. Are able to process, so. Much more data at a larger. Scale so, that you can really get. Make sense of your data of your of your document so to say. If. You look at the. Unstructured. Data set, on. One hand it really. Allows you to to. Come. Up with new ideas business. Ideas and to really get. The get. The value out of it but on the other hand there was also cost associated to, it keeps on growing and then, unless you have access to it. It's, it's. It's. Not worth much but it poses a big risk and this is something that we're going to discuss during. This this, session as well, because. A lot of organizations are, saying like ok I want to be, able to make my my, strategic decision I want to base them on actual. Facts and information that I that. I've got it over time and only, in 14%. Of the cases this is something, that's actually true, and in, many other cases a lot of organizations, are struggling with this huge, amounts. Of unstructured, data and, why. Is it so hard to work with this this, information well. Research shows that, the, vast, majority of, this, unstructured, data consists. Of various, formats you're, talking about business documents, different software, but also emails. And. And in, attachments, which makes it very very difficult to. Really. Structured, in such a way like, I showed before that, you can use it moving forward. So. The. Scary part about this as well is that the, data keeps growing. And a. Lot of organizations, are identifying. That and seventy percent of the time it's growing faster, than, they. Could ever imagine and, that, means the, pile is getting bigger and bigger and it's gonna be harder, and harder to really, get yeah use it the way you should. So. If you look at it value. Declines. Over time so a lot of customers, come to us and they said like yeah well actually that, data is not that important, anymore because it's two years old so. You, know what, do you want to do with it and that's that might be true but at the same time there is still a cost associated to, it it's it's still increasing, and and, storage. Is not for free so to say then, there is also the, risk if you don't know what you have it. Poses risks and, you. Need to know where, all the information is stored you need to know what's, out there and. If. You, look at all. Of these. These. Things and you put them all together then, poor, information, governance has, real, big consequences.
So, You've, got higher cost because of the storage which is increasing. Over time there. Is of course a loss in revenue because you don't know what per ticket for activities, are or what happened in the past or you cannot learn from it and this, also creates, a competitive, disadvantage, and. So. In. Order to really work. Well you have to increase your information. Governance because, it also exposes, you to risks. And the. Risks are associated to different, laws and regulations, that are out there, you. Know the most important, one of the most, known one is actually, the EU. GDP. Our regulation. Which, is you, know started off in, in. 2018. But, ever since there have been other laws out there, and they. Have one thing in common. Strong, information governance, is, really. Really important, because you need to know where you have your information and. You. Also need to know what. To do with it so, in order to really. Understand. What. We can do. And. How we can get you access, to your unstructured, data uncas. Gonna take over and she's, gonna talk to you about some use cases with. With inside, so anka, um please, go ahead. Thank. You. So. I think um you know, it's very clear, that the, machine learning and AI opportunity. Is really in a couple of different you know places one is how, do you actually drive value, out of your data out of that unstructured, data, but, it's also about risk, as as as you, just mentioned on tone on. On. Our side I want to talk briefly about our platform. And. Then really. Dig into some use cases that that, we see with customers, so, so, that it hopefully it ties. Together sort, of the theory, of you. Know machine learning in AI and the practice, of using this technology to gain, value, from your data so. Iron Mountain insight is a cloud native, intelligent. Content services platform, we. Use machine learning to automatically. Classify. And extract. Metadata, from physical, and digital data, I'll talk a little bit more about what that all means, but. It's it's really about. Providing. You, with the, ability to gain, insight, from your data and to, apply data governance to. It so. From a platform perspective, there. Are a few there are a few different, sort of levels, that. We're gonna talk through right first is the ingestion or, you know where can I what kind of information can I ingest, and. And where from right, so. We, believe that many organizations when. They look at their unstructured data you're. Faced with a, variety, of, repositories. And a. Variety, of. So, it could be physical, paper. Records. That you need to ingest. It could be data, that's on tapes. It could, be in the cloud it could be on a sharepoint. Server. It, could be a file. Share for. A certain line of business it could be. Original. Videos it could be images. It could be many many many things it could be office documents, obviously, right and so our platform allows, you to ingest any of, those types, of formats. From a wide. Variety of. Different. Repositories, so. Ingest. All of the stuff that you have that. Makes sense for you to ingest, in order to drive value out in. The, end so, the second step in order to get us there is of course the the enrichment right, so, the the as I said the platform, is a cloud native, platform. What, does that really mean that means we built it from the ground up, to take advantage, of all the benefits that you get from being in the cloud so things like having basically. Unlimited. Storage. Having. Basically, unlimited compute. Right it's really just a matter of you. Know scaling, to to to. Allow you to process, and, I'm the way that you want to and at the at the speed that you want to so. Building. It cloud native was important, to us so that we can take full advantage, of the benefits of cloud right, so. Platform. Is, essentially. Storing, the data in the cloud and then also processing. It and processing. That's where you know we really. Deal with the classification. As well as what what's referred to as extraction. So. That's where we pull metadata, out of, the documents, in order, to enrich your metadata that you can make, available to your users so they can find things and use things and analyze things. So. The enrichment, is first classification. That. Really is just what. Is it now. I know what I have right, it's, classified, and, and, that allows me to do a number of things right one from a processing, perspective once I know what it is I actually know, what's interesting about, that kind of a document right so if I'm looking at an invoice I probably want to know who the vendor is and what the total is and what I ordered, that I'm being charged for if.
I, Am, looking at a photograph I might want to know whether there is a particular, landmark, or person. In. That photograph, doing, a certain thing perhaps, someone, is hitting, a baseball and hitting a homerun right so once. I know what it is then, I can decide what, I want to know about that asset, and that's really where the enrichment starts, so. Enrichment, is about using machine, learning to. Identify certain. Parts of, of. Text. Or certain. Features in an image that. We are interested, in based on knowing, what kind of document, we're dealing with so. What does that really do for me right now. I know what I have and I can add metadata. To it and now. That, means that I can search, and access this content much, more easily because, I have metadata, so what, I've really done is I've, added structure. To, this unstructured data so. I can search it I can access it I can visualize, the data I can do content, analytics on, it I can. Also apply. My. Policies, right I know what it is I may. Know that it contains personal information. Because that's what I've extracted, right I'm, looking at you. Know a certain set, of documents. And I know there could be personal, information, so one of the enrichment, or extraction, activities could, be. Identifying. Whether or not there is personal, information so, now I can apply not just my retention. Policy, because I know what it is but I can also apply, privacy. Policies, or respond, to privacy requests. And then, the third. Value. Driver, here is automation. Right so if I have workflows. Or. Processes. That are still very manual where human beings have to you, know sort of go, and and and you, know pull out information or you. Know find the right sort, of information, for the next step to be able to occur. Those, are the types of opportunities, we're again, as Antoine mentioned, earlier we're. Looking to augment, and help the human do the things that they're actually looking to do as opposed to you know trying to. Manually. Do things that we can do, with, the machine. So. Let's look at some use, cases here and the first one, Antone. Actually already brought up right so, customers, we hear a lot of, customers. Especially, in Europe, obviously, who, are trying to comply with gdpr, and they're they're struggling, they're. Getting a good. Number, of requests, from employees, who now you know have the, right to ask, what. Personal, information if theirs is being stored by their employer, and. So of course you know why is this so difficult, it's difficult because first, of all that, data often resides, in lots of different places right, it could be, documents. That are on, a SharePoint, or or, a file, share it could be documents. That are in an. HR, system it could also be in physical, paper. Right I've been at Iron Mountain for a long time I filled out forms when I first started, that were physical, paper forms, and. I bet they're, still in a box somewhere. So. It's, really important, to make sure that you're able to produce all of the personal information that, you're storing, and. Doing so can take you know quite a bit of time if it's distributed, like that and you may not be you know completely, certain which.
Box Something, is in or. What other boxes, might have information, about that particular, employee. So. In this case from a processing, perspective, what we're doing is using Google. Document. AI, as, well as Iron. Mountain pre-trained, models. In order, to enrich, the metadata in in a couple of different ways right one, is of course to, look for personal, information does, the document contain, personal information yes. Or no if. It does then, we need to also know what employee this, personal, information, pertains. To right whose data is it right and. So what. We're doing here for, customers, is essentially, you know pulling out, perhaps, an employee ID if it's part of the document, but. Certainly a first, name last name of the employee which, then might allow us to pull. The employee ID if it's not to be found in a document, from. An HR system from a system of record, right so I know the first name last name I can look for that, person, and. Add, that employee ID so. Now I've got metadata, that allows an HR person to come into the application. Search for a certain employee ID or, first-name lastname, combination. And find. The personal into the documents, that contain personal information. In. Terms of the output I, just. Said you know you might have someone actually using, a search UI of our, UI is a very visual search UI so it really feels like you're seeing, your records, as. Opposed to a you, know just a table of metadata. And. So the employee can use that UI to find, the. Information. As well as export, that information, and provided, to that employee that requested, it we. Can also provide. Metadata. To, another, application, perhaps in this case you might want to take, some of this rich rich metadata, and actually push. It to your HR system, so that you have it available there, as well and. Then finally and very importantly. Gdpr. Also, requires, that you, apply. Retention. Policy, right so you, can't just keep stuff forever right, you, are obligated. To keep. Things, for as long as is required, from your business perspective. You. May. Need to get rid of it if the employee requests. It and. You, most certainly will need to get rid of it when it's retention. Period is up so, when, we know what documents, we're looking at we can apply the appropriate, retention. Policy, and. We can make sure that we process, destruction. Of these records in a timely fashion. Going. To a very very different, kind. Of use case this is for media and entertainment. So. This could be, companies. That hold archival. Information, such. As you know a large collection of images, perhaps. Of famous actors or dancers or, what-have-you could. Also be sports. Teams that have archival. Footage of, many many many games that. They store, and. These companies or actually, for that matter of music, right these. Companies just like everybody, else. Is, is. Excuse. Me struggling, to to. Be able to get to the right assets, when. They need them right so there are vast collections. Of very. Valuable. Images. Videos, and documents. But finding, the right, image. Or the right document, is incredibly, hard right so you have people sifting, through looking. At images you may have people watching you, know movies. To find that one spot where a certain scene or a certain phrase was. Uttered by an actor and. So what we do here is we, use, a variety of different. Approaches. To, transform. Anything. That's on the screen right, so speech to text for. Anything that's being said OCR. For anything that is is you. Know sort of written on the screen or visible, in the image so it could be the name on a jersey, or it could be, perhaps. A something. That's actually written on the screen we. Detect, objects, we can detect. Landscape. We can find. Landmarks. So, you know there's the Eiffel Tower we, can find that and label that automatically. So when you search you can actually search for Eiffel, Tower and find it magically, in in your collection. We, can detect, where there are faces, we can, look. At those faces and see whether these are happy people or maybe not so happy people, we. Can you know look for similar, images. So for example, if the director, is, looking for a certain, where, you know perhaps a, very famous player, is, is seen you know hitting a, home. Run in baseball you. Can actually look for images that are similar with just one click so. It's really rich, rich metadata, that allows, folks. To find. Exactly, what they're looking for and find it quickly. Here's. An example of what. Look like obviously. This is a baseball, example.
And And you can also see the visual search UI that I mentioned, earlier so these are all video. Clips, and. And, you, can see the metadata, is being, used as facets, or filters, that. You can use to find exactly, the right. Images. Right so here, I, could I could be looking for a number of different things right I might be looking for a famous player who is new, happens to be pitching. Here, I, may also want to use this to, perhaps. Talk to my, sponsors. When. The contract renewal. Time comes around so. That I can you know say hey. This. Famous, player was in front of your logo this. Many times when they were pitching. And oh by the way I can pull in the. Information. From my database on how, well. Attended, the game was so I might. Also be able to say and you know seven. Out of ten times we had you know a sold-out, audience, so you were exposed your logo was exposed, both. Within the, stadium as well as within, any kind of you know broadcasting. To, X amount of people, which. Might help me with my contract, negotiations. Right in, this case this is actually archival. Footage so I might also be looking at enhancing, my fan experience, and I may want to have you, know perhaps a little. You. Know some tidbits, of very. You know interesting. Footage, for, a certain player if it's their birthday or or, what, whatever that the occasion. May be so, the, visual search UI the, similarity. Search that we can do and. The ability, to look, for you know. Text. If you will with. A variety, like is someone saying a player name is there a Jersey with the player name is the player name displayed, on the screen at the bottom I can I can, search on all of that content. Going. Back to a bit more of a document-centric. Use, case this. This one focused, on contracts. This, happens, to just. Like HR be another area where many of our customers are struggling, they. May have you know hundreds, thousands, or even tens, of thousands, of contracts, or. Hundreds of thousands of contracts, and, unfortunately. You, know they're they're oftentimes, very, distributed, in a number, of different repositories so, it could be different products. It could be different regions, of the world and. In things are sort of you know stove-piped. There, is also in, many cases if, you if you grow. Through mergers, and acquisitions, you may have a variety. Of repositories. As well as contract. Formats, that you're dealing with that. That. You know simply are, distributed. And very hard to search right. So. What. Do we do here right from a classification. And enrichment perspective. This. Is a very very good example to give, you a better idea of you. Know a what the process, looks like but, also what the benefit, is right so when you talk to lawyers and many of our customers, companies. They'd. Love to be able to you, know answer questions, like you, know which, contracts. Have. Non-standard. Payment terms right or. Which, contracts. Have unlimited. Or. Uncapped, limit, of liability right, that exposes, the company to risk and. We you know we really want to understand, what that risk is. So. As I said this is a great example to sort of you, know, talk.
Through How. This process really works and, and the way it works it starts at the end right, you really want to start with what are the questions, that that you, need to answer with this particular, set of data that that you're looking to to. Process right so in this contract, use case it. Could be you know how many contracts, do I have for a certain customer across. Different. Products, and across different, regions it could also be, which. One of those is the most current one. What. You, know kinds of contracts, do I have either for, a certain customer or, maybe even for a certain region right, and so, you, know I won't read through all of these but the idea is that you know it's, really important, for you to think through what, are the questions that I have that I wish I could find answers to quickly, and easily now. What that allows us to do is is to, really you know take this big you, know pile of unstructured stuff, and start. Thinking. About what, structure, we want to have right, so, the. Questions, that we have require. Us to do you know two things one I need to know what kind of contract, I'm dealing with right so in this case here in this example we have four different kinds, of contracts, so. That's what we want to be able to classify right, so. Step one classify. Here. The contract, types that that I want to classify into. And then, within each of those contracts. There are different things that are interesting about them and that I want to be able to search on right so. So, now I can literally, look at the questions, and go OK in order to answer that question for, this kind of a contract what are the what, are the what, is the information, I want to pull out of my contract, right and so, you get this list of of entities. That need to be extracted, or metadata, that you want to add to. Your contracts. And. In. The end it's it's all. About you know adding, that structure, that actually allows. You to, search. And, access. And analyze, your. Data with, a couple, of clicks as opposed to you, know essentially you know locking, people in a room for a long time to read stuff right so, here's another example of what the UI looks.
Like For this right so show. Me all of the contracts, that have. Non-standard. Limits of liability, and. Oh by the way I'd like to have that by customer, and by contract. Type right, so. All. Of the filters, that. Are available. Allow. Me to essentially. Use, that metadata, that I've extracted, from the documents. Right and I can select, the non-standard. Contracts. You can see that at the top of the screen it might be a little bit small but there is essentially, standard, and non-standard and, that. Can be based on a number of different, rules. That might be somewhat. Specific. To a given customer, and. Then, you know by customer, and by type again. I'm pulling out customer, name I'm making that a metadata, field that is searchable and. Within the UI you, can you, can sort of pin these so that you have them right and readily available if, you have common, ones like customer, name or like document, type that you want to search on you, can just pin these so that you have them available in the UI and you, can get to what, you need in a click or two. So. To sum. It all up. Anton. Mentioned, this many. Many customers, are realizing. That they're not using the data that they have and that in order to be competitive they, need to be able to do that right they want to be data-driven about their, decisions. There's their, strategy, and in, order to do that they have to be able to use. Use. Their data right and get to it and and have. It as rich. As possible so. That users, can actually, benefit. From all. Of the metadata data in order to find the right stuff very, quickly to, make the right decisions. The. Second is you. Know you really. Want to embrace. That unstructured, data right, if you're just you know sort of managing, it and and trying to you know kind of keep from doing. Anything you. Know bad ie you're trying to manage your risk as best as you can but you really have trouble applying, policy. Or you're just hiring more and more people to respond, to these privacy, requests. In a timely fashion. In, order to avoid fines, and, frankly, reputational. Damage, but. You, know if you actually process. This data using. Machine learning in, AI you. Might not only be able to written to respond, to these requests. Much more quickly but, you could also help. Make. Use of that personnel. That you have to drive your business instead, of you. Know finding, documents, and sending them to people. So. You, know the struggle of course is there's too much of the unstructured data for, us to get our arms around it in a in a manual or or human driven way and that's, where machine, learning in AI and, the, power of cloud really. Help you because we can understand. Documents. And Street and add structure, to them at scale. Right we can process millions. Of documents images. Videos, and. And. Do that at scale and add, that rich metadata, so that people, can now use it and so. It's. Really about using. The Machine and. Using machine learning to, drive, new processes, and use, cases that, you may be doing manually, today you may wish you could do but. You really just can't and. That allow, you to identify opportunities, and, manage your risk very. Efficiently. So. The, best way to think about it is you know sort of one of one of two places would be you know the most natural, to start right either where, do you have the most opportunity. In terms of driving net new. Revenue. For your company, or. Where, do you have the highest risk, right, so, if you are in Europe perhaps. Being, able to comply, with with, GDP, are is a really, important, thing because the fees are just you, know very high right it's two to four percent percent, of revenue that's a lot and. You know generally speaking you know we find that customers want to be able to comply right they want to make sure that they treat personal, information, of their employees, and customers the right way so. Wherever. You have the highest risk or the highest, you. Know opportunity. That's a good place to start. Ideally, if you can find a place, where you can combine the two that's. Really, you, know the sweet spot right so if you know that you have data that could be immensely. Useful, if. You can actually leverage. It to drive your decisions, and that, data also happens, to you know potentially pose some risk that, would be a good place to start I mean it's really about where, could you benefit, the most in order, in in, order to you know drive the business case for. Implementing. A solution.