Leveraging AI for SMBs: Turning Buzz into Business Impact

Leveraging AI for SMBs: Turning Buzz into Business Impact

Show Video

[Music] Thank you everyone!, Thank you Rod for the invitation   My name is Johnny Saffar, I am the CEO and  founder of OrNsoft So OrNsoft is an AI Software House We have been around since 2009  working in many different countries and helping   many different types of companies The objective of this presentation is to make you understand what is this technology We've been hearing  about AI since the beginning of the year, a lot   and uh some stuff were true, some were not really  but there's a big buzz and uh and the the girl   my goal today is to make you understand  what is AI and how it can help you "really". So first let's talk about demystifying AI, so the  the the first thing I would like to uh to say   is about a the survey that was done by the leading  scientists, leading AI scientists, and according   to those AI scientists, 50 of them are saying  that there is one chance out of 10 that we are   actually all going to die with this AI.  That's uh that's quite a statement and   just so you understand now, it's like  saying there is 50 percent of the engineers   who did build the plane saying that if you ball  the plane you have one chance out of 10 to die   so I I don't know if you if you I will  not take the plane to be honest so we we did start really to speak  about AI since 2020 people did..   the hype let's say started but yeah I started  a long time ago, it started in 1950s when the   basic concept of where I was created uh the the  first person who fantasized about AI was called   Alan Turing and he said that there is a  basic set of questions and if the machine   can answer those questions the same way  human being will answer those questions   it means that the reasoning and the..  the reasoning of the machine   is matching the one of the human being, so  that's great, a very very nice Concept in 1956   the real birth of AI happened with the first  machine learning, and I will I will get soon   into details of what machine learning means,  and then the first neural networks in 1957.

So Since the 50s all the way up to about 2017  there was progress but not that much, right   each field of AI was, was really individually  developed and the progress was really segmented   per type of AI, it was AI for audio, there  was AI for images, there was AI for voice...   many different types of AIs and when you were  studying and learning about one type of AI   it was a not the same books, not the same way to  learn as the other types of AIs. Those types are used in many different fields right sometimes  they are combined and what happened in 2017   is called, there is a paper that was published by  a researchers at Google regarding a technology   called Transformers okay, Transformers are very  important and that's what did made the big change   because that allowed the convergence of  all the AIs into one main AI, so I'm gonna gonna formulate it a different way, imagine  you have images, and you have text   and you have video, the text for us it's a  simple to understand there is a text that can be   understood by AI, and then and then processed by AI,  now if we talk about video, video is a completely   different type I mean different type of processing  Etc... so what does the the Transformer technology do   is convert this file into its basic element,  it transform this file into its basic element   basic element is binary file I'm not sure if  you're familiar with binary but it's like a   basic computer language and the Transformer  technology will do this for all the formats   so when it does this to all the formats then the  principle of of AI to learn and understand and uh   it will not be different from one type of  AI to another, they are all the same now   so the the I will get more into exactly how it's  working soon but now you have to understand all   those incremental uh advancements for each type  of AI that was very slow since the 1950 are now   all converging to one and when one scientist does  an advancement in one one field, it's benefiting   all the fields, so that made the first real  stimulation of AI, so let's go a little more   this is actually the graph you can see  from 1950s to 17 was not that big right   2017 the Transformer model, Transformer  model, you can see this as an engine    a new engine to process the information was born,  the scientific Community took this technology   and used it across all the types of AIs, 2017 to 2018   you can see the curve goes up  and then something happened in 2018 a new type of engine again called generative  pre-train Transformer, and that was released   by the openai community, because back then it  was an open source and an actual community   and this uh pre-train Transformer was a  very very uh.. I'm going to explain because   this this is very important to understand, this  is not a New Concept "pre-trained Transformer"   You've been using most of you I'm sure,  cell phones seems at least the 90s the text   messaging, you know where starting to  type a word and it was finishing the world, or   suggesting the next one, this is the technology  that's used for the generative pre-trained transformer Just at a much higher scale  right, because there's billions of informations. 

So what does it do, it's trying to guess the best  way possible, what will be the next word, and this   is the way for it to generate the the end of the  sentences for example, now if you take this into   its basic element, like we said earlier, you have  the beginning of a text it can finish it for you..   you have a beginning of an image, it can finish  it for you... same thing for a video, same thing   for an audio, we know today that with a  few seconds of audio of your voice, AI can   generate much more, and do whatever it  wants, same thing for images right, so   that's really the technology where we are at today  2018 that was released that stimulated even more   the community the adoption was very high and you  can see 2018 to 2022 it went even faster, now 2022   November 2022 that was not so long ago,  ChatGPT was released, chargpt was released   and something even greater happened, and we are  not talking about adoption by the scientific   Community or, or really only technical people, we're  talking about human adoption of the technology November 2022, ChatGPT was released. January, two months  later, they already reached 100 million user, in two months  so to give you an idea, the fastest  one before that we reached 100 million was Instagram and it took them two years, so in human  history it never happened something like this This was a big signal to all those  smart people doing business, saying oh!  maybe it's time right, and I'm sure you  you've had the same reasoning   now let's break it down a little bit,  talk about the basic concept of AI   Obviously there is much more, but I'm gonna try to  make you understand the logic behind it.  

we have four different elements that will compose the AI,  the first thing is called the data set, so the data set, it's very simple and those are the information  that we will use as a training, so we want to learn   we need information right. The second one is  called a neural network, so the neural network   I like to see it as the neurons and, actually that  was the idea behind it, neural networks is the   neurons that are in the brain where you store the  information and connect those information together   The third one is called machine learning, this is  very common World also, you've been hearing this   one most likely a lot. Machine learning is compared  to the synapse, the synapse meaning the the things   that will make the information work  together, that's also what mimic the reasoning   of the human being, I know this, I know that, maybe  and then I can extrapolate, now by combining those   things, that will give birth to what we call the  model, and the model is the actual thing that will   be used to do the job, so the intellect,  now you know things you can do things.

I would like to do an analogy, with the  first human being, and to make sure that's uh   very clear for for everyone, I'm giving  this analogy a lot to some clients, a lot of   clients actually, and they seems to like it  so, this first human being, imagine it's night   and he's seeing fire, first thing  that come to his mind is "something shiny"   or first information that stored in his neurons  in his brain is that there is something shiny.   This human being is getting closer, it's getting  closer and he realize that "fire is warm".   Second information, fire is warm, now  what happened with the "machine learning" ?  A third information is born saying "fire is good".  Who gave him this information that fire is good ?   Nobody! He did end up realizing that this is  good because... and this is the the what  

we try to mimic with the machine learning.  Now you can go even further, he's getting   even closer trying to touch the fire, now what  happened? "fire burn", "fire can cook"   "fire is dangerous", "fire this", "fire that"... a lot of other  information are going to populate the brain.   Those information have not been given by  anyone, they have been produced by the brain   and where it's interesting is, you remember at the  beginning, we said fire is good... now fire is not   so good anymore, so fire is not so good, so it can  also correct the information it had in the brain.  

That's the main difference with other technology.  Before that we had a great deal of automation,   We had great deal of of trying to do things that  looks like human being were doing it, but the goal   of AI is really to mimic human processing,  without the constraint of course, because   we know that AI don't sleep, AI don't take  vacation, it's not sick, and of course, work way   much more faster than human being. We'll get back to this information later.   Now! There's different type of AIs, and all those types are basically  combining different Technologies   always in AI, and by doing this you can achieve  different things. The first one is called Curative   AI, so the curative is the one that will help you  find a solution and do things to fix things, so   that's the most common one in business, we have a  problem here, we need an AI to do this, and that's   what it does. The second one, and this one is very  popular now is the generative AI. The generative AI is capable of creating new content, ideas, images  that's the one we talked about a little earlier.  

The third one is called predictive AI. So this  one is capable of seeing through a lot of   information, things that happened in the past,  not just basic events, but also go into details   thousands, hundreds of thousands of parameters,  compute all those information and give you an idea   of what will happen next. And finally the third one,  and that's my opinion, but I think   this is the most powerful one, is the prescriptive AI. So, in addition to being able to tell you   this what may happen, or this will happen, it  will tell you also, how you can avoid it, how   you can take advantage of it, what's the best way  and suggest what you should do next.  

Those things could be used in many different  areas of business. As you can see   there is no an industry that that will benefit  more than another form AI it can be used   in any department, it can be marketing, accounting  operation, administrative of course, legal, sales   support, etc.. everyone can benefit from AI. Now, the idea, the most important thing is to   identify really if you need it, and where you need it, because it's not just to say oh I want to do   this. No there is an actual logical way to identify  and say okay this can be done by AI, these can be   fixed by Ai and we have an AI implementation action plan if we can say that you   can download, after a presentation I will  make those documents available to everyone, you can   download it, just go through it and see with your  own company, you will see that it can help a lot.

Now we've done a benchmark with our own AI,  obviously we have our own softwares also .  And this is the latest Benchmark we  did, it was at the beginning of the year,   and to give you an idea where we stand right now,  our AI can execute a week of work, done by   50 highly trained people, in less than two  hours, with the five time better quality   so, just realize the amount of work we are  talking about, the AI can do it in about two hours   the benefit that you can get from this is not  just, okay obviously the main one is   money, right, but also, time, the client satisfaction. You know when the client doesn't need   to wait a week, or a few days to get his answer,  and he can have it right away within two hour,   that can be huge, also Imagine, when you have to approve a loan or   process a project or whatever, this can be  done in few hours, usually it's taking half   a day, two days, so that we are talking a lot  of benefits from this kind of technology. Now If you're not convinced yet that you should board the plane    I'm gonna give some examples of first  AI implementation that went wrong, and then I'm   going to give you examples of AI implementation  that went very well, with a little more details so first the implementation that went wrong,  there is quite a lot, I did select a few   some from big companies some from smaller one, and  in I try to find things in different Industries.   Mainly when it goes wrong, there is a reason.  The first reason could be because the AI used  

the wrong data to learn. So imagine you have a new  employee, and you tell him you know what I'm going   to teach you what you have to do, and you teach him  the wrong way, this employee is going to do a bad   job, are you going to tell him : "hey, you did a bad job!"  You don't train me well, how do you want me to do a   good job" you know it's a logical thing right, so it's the same thing with AI   first thing and most of those problems are coming  from bad data set, the second reason is   that limitation of the technology, so sometime  and that happened a lot, we I've seen it   mainly during the hype of AI on 2021, 2022, people  saw ChatGPT, or Bard, or some other generative AI and they   said oh great that's it, I'm firing everyone  from my Support Department, and I put the chatbot   oh, okay, you can try but that may be a problem.  The AI ended up talking to clients, or you have   AI : "ho, you have a problem with your car? Guarantee we replace the car!!" The client comes in, okay where is my new car ?  No sir, we don't replace the car, you know!  The agent said you replace my car!  I don't care it's your problem!  But it is an AI ! I don't care... and and you end up with a lawsuit..  

So you see, the risk the liability it's not because  you use an AI that you're not liable right,   you have a business, you have to be very careful what  you do with this technology, I'm not saying it   it cannot be done, I'm saying it has to be done  the right way. You see in these examples, we   have the Covid-19 that was a very famous one.   The technology was, I really believe the technology   was working, and the main problem was the data set,  like we just said, we have some chat that were   telling patient "kill yourself, it's better, you will feel good",  that's uh.. that's a problem also   some were biased, because there is also a problem  with the data, the the AI sees that men is always   at this position, has been trained with  hundreds of thousands of information,   it's gonna think that, okay maybe men  is better for the job you know, while it's not true..   today we can say this with certitude,  and the latest one I saw is from a lawyer in   New York, who used chatGPT to prepare a filling  and the AI made up information, it was using a   information that for the appealing everything that  went after that that did not exist, and when the   judge saw this, he said wait a minute let me check,  he checked, and this information didn't exist   so the his career finished on the spot right,  but you realize that it could be used,   but the right way, generate AI is very powerful,  we believe it can be used in a certain manner I'm  going to give the next example, and we'll  see how it can be used the right way so The real world application of AI "the right way" I'm going to give an example, first how   we did implement, to be honest it was difficult  to get the approval from clients on sharing   information that may be confidential, the first one  is regarding our company how we did it, because I   have no problem sharing this information, and the  second one is from a local business here in Miami   and we'll go over it, and I will explain  what type of AI was implemented and how.

So, the first example is regarding customer support So in our company,   we have clients, and those clients, when they  have a contract, can raise a ticket when there   is an anomaly, or whatever question they may have.  Now at the beginning they were sending email   to the company, and to handle those emails  we had two human agents, those were technical   support people, Junior, two years experience,  paid between $20 and $25 an hour, and they were   to be honest very efficient, but the amounts of  tickets grows as you acquire new clients right   So, what did we do, we did try to centralize  those requests into an Extranet, so the extranet   is like a client portal, where clients can go,  open a ticket, until now, nothing special right clients where uh okay with that of course, but they  didn't want to go there and open the ticket, and   they were still sending emails, so we did add an  extra layer, and that was not AI, it was automation   a lot of people out there are trying to sell you  things or tell you about things that are AI, but   they are not, this is automation. We did build  a bot that was monitoring the mailbox and when   there is an email, taking the email, go in the  platform, open the ticket, that's not AI right   now that helped a lot of course, but something happened because you   reach a certain point, where you are at capacity and the next step is to   hire another person, you cannot just say okay  I'm gonna pay a little more and I can I can   handle no no you need an extra salary or  at least part-time right, so what did we do we had about 300 tickets a month for two full-time  employees, that's a $6 000 to $8 000 dollar   a month for those two employees, and average  about $80 000 for the year   I'm giving numbers to show you  really the impact okay what happened uh once we did  change this step of human agents, Just so you understand why we have this step,  i went a little fast on this one, we have this   step because when the ticket comes in we don't  want the engineer to work on it right away, maybe   there is not enough information, maybe it's  not a bug, the engineers are very expensive   right, we don't want them to lose time trying to  guess what the client is trying to say right, so   that's why we have this layer in the Middle  with this technical support agents. Before we used to have those technical  support, Junior, two years experience agent,   between $20 and $25 an hour, two full-time employees  so that's about 80 000 a year for those two people.

To take care of those tickets, there  can be a delay about 24 hours, based   on the SLA, based on the contract, but when the company   is closed, nobody is going to answer,  you have to wait tomorrow morning those two people were at capacity, and if we  want to handle more, we have to hire more people   now after we did Implement AI in this department this is what happened, instead  two people we still kept one part-time   and I will explain soon why this person  is still paid between $20 and $25 an hour   that's about $19 000 a year, now with this, there is a near real time answer to the tickets   unlimited capacity, there is no limit anymore, it's  not going to cost more, and to show you, you see on   the right the implementation cost of the AI uh for  for this department was $20 000   for the license, there is other fees, the  processing fee because each time the AI   is thinking it's costing money  right, of 50 Cent per question   so the total cost for the first year  was $40 800 if we compare this to the $80 000 when we  had two people, this is already 50%   saving on the expense, and I'm not talking about  the user satisfaction, to augment the capacity   people are very happy, and they're engaged, and if  we want to take on more clients we'll   not be worrying about, holala, we have to find someone  and hire people, no that's covered now also you see this, I'm saying here  the cost for the first year because the initial $20 000   implementation this is only the first year, so  the the next year   we're going to make even more savings, so it's very interesting. It's a simple example,  but I think it's very effective  to make you understand   how it can be done, now regarding the technology,  when I say the right way what does it mean   in this scenario the AI is going to  communicate with clients, the ticket   comes in, what does the AI do, the AI will read the  ticket, understand what the client says or want   it's going to go check who's the client, what  project the client did with us, the history of all   the tickets that were opened by this client, who  is taking care of this project, and based on this   reply to the client, we need more information, is  it the same as this this and that, or thank you for   raising this ticket, we'll take care of  it as soon as possible, and the AI will assign the   ticket automatically to the right person in the  company, so you all this job usually require the   human intellectual job, you see it's not automation,  it's really thinking, seeing, understanding and   then taking action, that's what's done by the AI, in here Now, let's move on to the next example this company Dayoris Doors, so those are  based locally, in North Miami, they   offer Modern Luxury interior doors  and they manufacture those doors   now when they get an order, this is the  process they have today. I'm telling you   I will not be able to share financial  information, I'm not allowed to do it   they did not accept, but I can give some details  on what we did, and what was the gain out of it.  

So usually the client requests an estimate, when  they request this estimate, they send   a plan, or they give the requirements,  what they want, I want twenty doors, ten   doors, usually you don't order one door for  your house, you change a set right   and the agent, the person on the phone, will  take the information and prepare the estimate   they send the estimate, the client approved the  estimate and when this happened, they generate   what they call a master sheet, a master sheet  is a document where there is all the information   detailed information, the size  the way the door should open   Etc... the material that should be used, there  is a lot of information in this document   now they generate these documents and they have  one person going to the client, on the field, going   over the master sheet with the client and based  on what the client says, adjust maybe this door   supposed not to open like this, but like that  this material he wants a metal instead wood   we don't know, there is always change orders  now they take all this information   in the in the master sheet, and then go back to  the office when they get to the office they   do what they call a Reconciliation, whatever  information that has been added or modified   in the master sheet need to be reflected on the  estimate, now that's what I call "the pain point"   the pain point meaning, this job was heavy, risky,  and for this company was a real problem   because there was about 5% to 10% issues with those reconciliation, so once this reconciliation is done, then there  is a change order that is sending out, update   the estimate, client sign and to be honest, when you  see it like this, it seems pretty solid, bulletproof   almost, and uh and and supposed to be fine they  used to have four people to do this reconciliation   and per project, it was taking up to three hours,  so just to make you realize, it's a lot of work   for a lot of people, it's not a simple task, now  after we implement the AI, this is what happened   before, they had the four people full time  they were at capacity, not everyone could   do the job, and you cannot use someone that  knows the job, that you hire and   they can work right away, you have to teach them,  make them understand why and how it works, and   all, this so it was difficult to increase the  capacity. 3 hours per project, there were weeks behind   on a reconciliation, that's a problem because  reconciliation is not done, the actual final   estimate is not signed, a final estimate is not  signed you cannot start the project, so it's    not the minor problems, it's  a huge problem after we implement the AI   there was one person full-time, we kept one  person, the capacity now is unlimited because   it's AI right, the time for this person went from  3 hours / 4 people to 20 minutes 1 person   and for the the reconciliation,  at least the AI part is near instant   and if they want to do more, they don't have any  problem anymore, so just to give you an idea   this is in term of time and resources  97 gain that's huge   they are very happy, and now we are working on other departments  but uh I I can tell you the the when AI comes in   it's not only a 10% gain that you will have, it's  way much more, remember, the AI is working   way much more than the human being, if trained  properly, can do a better job so it's a big gain   the last time we saw such a big uh benefit and  advancement was the last Industrial Revolution   in the late 1800s beginning in 1900s, when human  labor was replaced by machines in the factories   Did people suffer from it? yes, okay   Did we benefit from it? civilization I mean..   yes, Was it avoidable ? no, and what  is sure is that the one that did jump   in the train, those are the ones that  stayed after that, the other one disappeared so I think it's pretty clear that to have an  AI today is not an option, you will have to   take it, don't wait too long because your  competitor will not wait for it, and just   be very careful on on how you do it, always  ask yourself what's the liability, what will happen   if it goes wrong, I'm not talking about just it  doesn't work while it can cost a lot of money   so you see that's one risk there is no  real liability behind it unless it kills someone   right, but when there is an AI doing  something for someone you must be careful Now if we want to talk about the future of AI The future of AI, I think that's my opinion is  the convergence between machines and AI   the body with the brain right, the body with  the brain, actually the future is already here   and it's moving very fast, there is a company,  and I did find this example very interesting   who did a point as its CEO an AI, and this company  it's a big company, large company, they in the   Stock Exchange and they have thousands  of employees, is now managed by an AI   between the time they did appoint the new CEO,  and about six months later, the stock went up by   10 percent, profit skyrocketed, without them  having to sell much more or firing anyone   so you see that that's a pretty impressive, just  by operating better without causing any damage to   anyone, if I can say you can actually turn things  much better, there is some more examples about that   but I did find this very interesting, I'm pretty  sure soon, we will see much more of those,   helping us in our personal lives  and also in the business. For now, replace the   employee completely, no, I don't think so assist the  employees, yes. Thank you very much [Applause]  

[Rod] Can you answer some questions ? sure if any ones have questions yes [Person in the public] How long from like when you started  working with the door company until you train the ai ? and then how do you monitor it once  you put it in place to make sure   very good question okay so for the  company to prepare the data set   because that's really they took more time  than than us actually uh to put in place so   they took about amounts to get the documents ready  because uh we are dealing with documents we have a   software that is specialized in what we call hyper  automation, so intelligent document processing   we can just once those documents are ready give  them to the AI, tell the AI what to verify on   those documents, you have the estimate you have  the master sheet, this should match with this   this should match with that once this is done the  training goes very fast the training takes about   an hour or two after that you need to use the AI, so  that's the time that took about maybe three   to six months, and uh within that time, but  don't get me wrong when you put it in there   it's deployed, it's going to be maybe at 50  capacity, I mean capacity of doing the job   this is why we keep a human being in the loop, the  human being will see what the job the AI is doing   correct if something is not okay, the same way  you will do with a new employee, you have a new   employee you go over, you say okay here you did  a mistake, here it's okay, and after six months   you tell him, okay now leave me alone, do your job I  have no time, you know, so it's about the same thing   yes [Person in the public] What's the risk when  we use article intelligence in yes political campaign ? The main risk   I can think of is hallucination,  so hallucination the the technology   the AI technology and again in  political, it could be used for prediction or prescriptive, tell us what should be done, or where we should go, or there is different ways to use it, but I   believe the main way today people will use  it is more to generate content, for example   generate content, make sure to use a technology  that will be evidence-based, meaning the generative   technology should be used to understand the  request, to generate an answer but not as the   base of the knowledge, because it has so much  in the brain, if you ask it about something   and it doesn't have the information, because it  has so much, it's going to make up information   and that's when problem happen, we have  another technology called GVK not GPT   this technology will take a set of  information that we will provide   get the question, generate an answer based on this  knowledge that we did provide, and give the proof   the source of the information along the  answer, that's the right way to use it   so yes the the I believe that would be  the the highest, the highest risk, yes [Person in the public]..   kind of slim down our overhead and I'm  extremely fascinated with AI and the future if you just have a couple questions that I want  you to uh just to you know give me a quick answer   to or some information, um I've I've been told the  tractors have told me AI is too limited right now   AI does not give access to current information  to limit dangers, AI databases are one to two   years outdated they're pre-loaded with certain  information certain access to this information   and for me I just made it to you on what AI is  learning and also that AI has no access to Legal   databases, so if you ask AI about a particular  case it won't know so I was curious about that   [Johnny] sure sure that's, that's a very good point if  if you don't have the knowledge of the different   ways technologies are working the the AI is not  one one type, there is different types of AI and   different ways to process the information, what you  just said is very true, for example chat GPT again,   they are not gonna like me a lot but  when you go in there and, I like this one, I go in there and I ask it not so long ago, who  won the the World Cup soccer and it was   still saying it's France, so for me that's  okay [laughter] , but you can imagine if you ask it about   something else that you need in your business  and current information, that's not going to work   this technology I was just mentioning the  GVK is working a different way you take   the current information you provide it  as a document PDF, the AI will ingest it   in a vector format, so I'm not going to get too  much into the technical, but basically what it will   do is when you ask it something it will understand  go in the base of knowledge that you did provide   look for the information, we are doing right now an  experiment with, and that will tell you   it's not so far from the example that  you have, but we do an experiment now with the   city of Miami website, where you know when you  do construction, it can be very complicated very   quickly, not just for us people trying to change  something in our house, you know, but even them, when   you call them sometimes you know, they need time to  find the answer, and that can be a problem, so we   went on the website did an experiment, took some  of the information, just by printing them as PDF   we did put them in the folder, and we used  the generative AI we tried with ChatGPT, Bard   some others that could be installed locally  in your company if you have privacy issues   and uh we plug those technologies with the  GVK technology now what happened, when you ask   the question, the the GPT technology will be used to understand what you say   look for the information get back the information  generate the answer and give you the link to the   the actual file where it did find your information  for you so if you use it that way you will not   have problems because if it's not in the base  that you provide it cannot make up an answer [APPLAUSE]  thank you thank you so much

2023-09-19 11:13

Show Video

Other news