Senzing CEO Jeff Jonas on AI, Data Analytics and Entity Resolution

foreign speaks we're talking to a longtime Ministry expert about data analytics how to get more future companies analytics practice we'll look at artificial intelligence and we'll look at a technology called entity resolution with me today is Jeff Jonas chief executive officer and founder of senzing Jeff happy Friday to you hey thanks happy Friday to you too and I know we talked before we started you you've been you've had a busy morning I appreciate you being able to get on board here this morning you've got a lot going on so thank you for being here this morning well thanks for having me a few minutes late but I have arrived sometimes the Rival is worth it um I've got some hopefully insightful questions to ask you but let's do the kind of the elevator pitch you're stuck in an elevator with a VC you want to get funding which you don't need that I realize but the what what is the elevator pitch what does senzing do for those people who don't know about it organizations struggle with figuring out who's who it's about you know deduplication it's about record linkage it's people call Patient record matching has a lot of different names people struggle with that we've made it easy we make that an easy task for developers so we don't sell cars we sell Transmissions people building software plug Us in and it solves that as easily as it would solve basically adding in a spell checker or grammar checker so your your users really are are professional developers yes yeah or I just ultimately of course the companies that hire those developers are your ultimate clients yeah the buyers often somebody that's a business unit exec yeah yeah all right data analytics in the year 2024 I think there's a rule that says I'm supposed to ask about artificial intelligence it's the only thing people are talking about this year so Jeff data analytics 2024 you know around the corner do you see AI as a savior of those companies that are struggling with data analytics is that going to cure all the problems well it is it is not going to cure the problems it's going to definitely change the landscape though about how people are spending their resources and I I think a lot of companies are just taking a pause to go you know how are we going to align our Capital spend with these new um technologies that are really really epic changes in what's possible right yeah generational changes do you think people are in a position actually understand the possibility I think people are sort of like you know fogging about so to speak I think um I think it's being over imagined what's possible now but even if it's 50 of the imagination it's phenomenal and and you know if we're talking about llm specifically I see them as more qualitative than quantitative and some are looking for them for answers and complaining about hallucinations but you've missed the point they're they're designed to be qualitative and so if you if you really think about harnessing them that way I think it's going to accelerate you know successful use cases the large language models are expected to be quantitative or or this might be qualitative not quantitative I mean meaning what exactly well an example would be one of the things you might want to know from an entity resolution engine is why did it put those records together well the sensing API sends you a a string of machine readable stuff called Json guess what you pass that through an LOM it flowers it up turns it into language it's and it'll say it different every single time but you know what it'll tell you it'll tell you something like hey the inbound record had this address one two three main street it turned out to match with a score of 88 to that address but the next time you run it it'll be more beautiful language but it'll be stated very differently you know just I can't resist a short detour into the world of large language models and that it it seems like it's a Hot Topic and some companies are thinking you know what do we want to build our own do we want to buy one off the shelf and try to customize it or you know what what's your thoughts about it seems like it's one of those big competitive advantages it's a little beneath the surface of the AI Hysteria let's just talk for a sec what's your sense of the large language model at this Mark at this moment in time well I think it starts with where does the data flow I think a lot of organizations don't want to present their data out into the wild and so therefore they're you know the the the the move I think in a lot of cases is going to be to bring uh to bring that in-house and have the language models run in-house and there's quite a few efforts going on where people are training it against their own internal data so it'll be very it's going to be very interesting to see how that evolves but it's going to evolve really really fast and I you know I liken this to when um IBM's Watson you know won the game Jeopardy it captured the world's imagination and then tens of billions float in well what's happening now it's hundreds of billions is flowing in like this is whatever that was that led to now what's happening now is orders of magnitude bigger yes right no I think it's it is it is a Quantum Leap from from those those Jeopardy win days absolutely um all right so here's here's the key question so companies are struggling with the data analytics practice there's they're sort of doing it they know they're missing a lot what's your advice to them you know how how to get some more advice more and more results from their data analytics practice well I think it starts with architecture and you know I feel like I've been beaten on this drone forever but people would just keep building batch processes and if you're trying to be competitive you're trying to make decisions you're trying to make sense of what your organization's observing fast enough to do something about it and fast and smarter and faster than your competition so when people are creating processes that they re-run a batch run every weekend or once a night it's going to be hard to be competitive and those architectures aren't going to grow up and be real time and so I I think a place to start is real-time architecture so that as the data is arriving as you're having the call with the customer as that information is being collected it's available to the rest of the organization so if that same person happens to connect through a different Channel a hot second later you have the right information so real time I'll call them streaming or even sense making architectures does that mean it has to be built on flash or is that beside the point uh well there's lots of like our system for example you know you need really high speed i o some kinds of processes need super high speed i o other processes um don't but you need low latency end to end you know exactly yeah other thoughts about improving the results from the data analytics practice what about those companies are floundering what would you tell them more context for better decisions often it's no longer the algorithms in many cases I would look at the algorman you could spend a lot more money and make it a tiny tiny bit better but what you really want to do is kind of widen the observation space and use more data so when I say context I mean better understanding something by taking into account the things around it because he's bad in a sense you look at the words around it to know if it's a baseball bat or a bat that flies in a cave it's the same with organization's information I think of it like puzzle pieces really you get a red puzzle piece you're best understanding your best ability to understand is going to be to see what other pieces sit around it and this is where does that shift take place is that down on the developer level how would companies make that shift well as the data is arriving what you're trying to do is assemble it if it has to and I'll go to the energy resolution since it's kind of where I'm living these days is a new data arrives you're trying to figure out is this a new person a known person are they related to somebody that we're already doing business with and having that puzzle piece say land in the neighborhood of the same you know profiler resume as the customer or same household gives you more information so you can make better decisions so it's about Landing information in context and then using information and context to do decisioning let's go ahead and drill down into what sensing does and and you know how is sensing addressing the analytics needs of its clients and what's the sensing advantage well here's the thing is I've been building entity resolution systems for decades and they've been expensive and only for the elite if you want the really good stuff and they're hard to build they take it's a phenomenal amount of money you can get 70 of a of a go-to list of things features you want in any resolution like explainable in real time and accurate and easy to use that'll cost you a million to get 70. so I'm clear and you may have already talked about it but can you give us like the 101 definition what is entity resolution ah okay yeah yeah yeah yeah it's recognizing when two entities are the same even if they're described different this one says that this says Beth the dates of birth the month and day are transposed the addresses are messy or she's changed her last name is it the same person or not and that's entity resolution it it comes by a lot of different terms you know it's like deduplication is a term patient record matching is a term um fuzzy matching data matching isn't it 30 or 40 terms this is fundamental in every know your customer every AML system every CRM every Master data management every Insider threat every vendor supply chain every sanction of screening every fraud system all of them really if you don't do a good job harnessing who's who you can't produce great results and particularly if you're trying to catch bad guys they don't use the same name and address on every record what about them and we've spent years working on that all right so that that is entity resolution so so put that in the context of what sensing is doing how does sensing help companies so we've just made it easy for companies to do this and many companies as I was saying try to go and build this themselves you get a bunch of phds around you spend tens of millions of dollars and it's not real time and it's not that accurate it's not that good and what we've done is we turned it into a plug-in it's like it's a link library or a dll or a shared object as most of our our partners are on uh Linux and you just plug it in so we've just made it easy for it's the first time it's really accessible and easy for the masses so our purpose is to kind of commoditize energy resolution not just for the elite so it's delivered via VIA an API yeah we ship it to them and they run it local no data flows to us who wants to send all their customer data to me I don't want it so we send people code they compile it in as a function call and voila they got to interview resolution they can spend their money on other stuff well what would the there's probably not a direct question I'm asking you anyway because you can handle it what would the other some other big data analytics vendors we know the household names I won't mention their names but you know someone asked them how's your entity resolution they go we're good at entity resolution look at ABC so I mean a lot of the companies would say hey we already our large brand name analytics package already does entity resolution yeah yeah actually and and boy it's in quite a large Market a billion dollars even been funded towards entity resolution center companies in the last five years it's phenomenal what makes us different other than being easy to use which by the way is maybe the hardest thing of all right it's the fact that we're real time and scaled to billions of Records uh there's no moving Parts it's literally it's all this years years and years 300 years of my team's been working on this in this you know as in the history and we compressed it down into one Dynamic link library that talks to a SQL database so the Simplicity thing is is not to be overlooked but beyond that it's real time it's explainable um and it's sequence neutral which means if the data arrives out of order you get the same answer uh-huh and that's the hard part is part of the real-time learning we've built in this AI I'll explain that another way is you've made a bill you've seen a billion records and you've made a billion decisions now you get record billion and one our out our real-time learning algorithm goes wow now that I know that had I known that in the beginning over the billions of decisions I've made should I made any differently that's very different it fixes the past fixed is the past and and the future is smarter than the future smarter see what happens too basically every other algorithm out there is as you get more data kind of maybe you learned a new piece of data and it tells you that that's not a junior and a senior or it isn't let's say it is a junior right that's a junior and a senior I shouldn't have put them together most systems that are batch in the month whatever once a quarter you rerun it then you learn that well you just went a long time without having the right information so using new observations to reverse earlier assertions in real time is non-trivial and it's right the accuracy up what about the question of artificial intelligence is I mean you of course data is always a foundation of artificial intelligence so and so the sentencing uh product can can you know tie into improving the AI it the sensing itself use artificial intelligence yeah we we kind of we built something purpose built just for just for entity resolution it uses statistics as it's ingesting to self-tune and it uses local knowledge as well about an entity like you when you learn a nickname and then it goes backwards in time and applies that nickname to see if it would make any decisions differently it's a yeah we cool well let's look at the I think the big question is the future the future of data analytics Ai and entity resolution say we're having the same conversation and oh 2027 or they're about just just a ways out what sort of things do you see and probably most important how can companies be getting ready for it now I think there's going to be less keyboards I think the way you know we're less one I'm sorry boards keyboards okay they're gonna go crazy because things will be Voice driven yeah or Voice or near a link or some you know eyeball tracking having a type I think you know now does that happen by 2027 I don't know but you know with these large language models I don't know if you've ever like you want to change your default directory Microsoft Word it says a bit of work where is it huh where are you going to go to do that natural language llms you're just going to be hey change my default directory you know just change it and I in in the space of energy resolution here and and by the way not even any resolution but just in in search Jeff tab tab Jonas tab tab June 22nd tab tab he's kind of saying hey would you find me the Jeff Jonas record it's going to be conversational search right not going to be clicking around looking for why the records come together why did those records come together and you're gonna it's gonna become very more it's going to be the way natural language is currently going to change a lot of the way people interface with computers I think that's going to be one of the one of the bigger changes and and if you talk to somebody today you're you're getting ready to maybe work with somebody you're like hey you're more of a Mac or Windows person they go I don't use computers you're like you don't use computers yeah no no I put in the work it's gonna feel like I'm telling you way before 2027 it's gonna be hey which which uh which AI do you use really Bard Bing uhbt they'd be like oh no I don't use I don't use an assistant you're gonna be like you mean to tell me you re-listen to the trans the recording of it of the one hour session and then you summarize it and then you do it all by hand that's crazy that's where we're going well it seems like I already asked my phone you know Thai restaurant near me in San Francisco and and my phone kicks out the answers that that almost feels like it's very close of course it's not a true analytics program but it's it's 52 data so to speak so that almost feels like it's here in many ways yeah I just I think it's gonna we're gonna see it integrated into more of the products that we use and it's going to become um it's it's exciting this is this era right now this is just really breakthrough in in Back to the end of the resolution there's been no giant breakthroughs for quite a while but what llms are going to do for conversational and any resolution where you can talk to an entity engine holy moly It's Gonna Change it's so exciting the uh llms are just going to so radically change the way people interface with systems and the the future is very exciting it's this so again a Renaissance era is is there one thing that you think might actually surprise people like oh we had no idea this particular idea was coming like gosh we're all we're really surprised shocked that this is happening now you put on your deep futurist cap that's a good see no keyboards right I'll tell you you know we're watching we in our lifetimes we saw a 100 Year pandemic in our lifetimes we're gonna we're watching us uh one of our own scripts become archaic they don't teach cursive in school anymore um okay so we're watching the you know and a Latin happen right in front of our eyes but these keyboards is uh you know the fact you got a tickety tickety type it out I that I don't know I think that's going to be one of the things that goes away and which how it gets replaced will be interesting you're right what's the speed and efficiency of that interface hmm interesting yeah and I I definitely see the the fade of keyboards you know we we interrupt we integrate more and more seamlessly with our computer system so the idea of a keyboard becomes archaic do you think it goes to eyes or do you think it goes to some kind of brain read your mind kind of thing yeah I guess it really I could see it actually would be a mind computer interface if we go you know way at an Olympia I think there would be actually a a mind computer and you know that has to be coming well it's going to be the same thing you're going to go see who you want to work with you go hey what mine computer interface you use and they go why don't use one they're like you you literally be the same conversation it's a lot of good stuff um I learned a ton as always like the entity resolution piece is fascinating uh thank you for sharing your Insight and please come back and talk with us again sometime uh thanks for having me it's great singing again thank you

This transcript was generated automatically from the video's captions and may contain errors.