All right thank you uh thank you for uh Tren and everyone for doing the Utah Tech week this quickly became one of my favorite parts of the weeks of the whole year um who's going to like three or three or four more events this week everyone yeah me too I'm excited um well thank you everyone for coming uh we were pretty excited to see the guest list kept uh getting larger and larger so happy that we could uh Extend and have everyone everyone a part of this so I'm Brad Moss I'm the founder of enhanced
Ai and I've been in the AI space de generative AI space for about 3 years so actually got into it before the big GPT graay um and quickly saw that there was a lot of exciting stuff uh with this technology and my mind instantly went toward hey there's way more cool stuff that I can build then I'm going to have uh the City in any engineering to build um they wouldn't be Able to keep up because this technology is so uh transformative so uh we went forward and we'll show you a little bit about what
we built um it's a platform that that integrates all the AI systems into one spot makes it really easy to orchestrate and build AI agents and at this point we built um on our system been over a thousand agents probably built on our systems um and so we have uh and we didn't start off calling them agents we called them workflows but as You know any new uh space technology everything coalesces the term rag was something we were doing way before it's now a thing um we've talked about that a little bit U but uh
it's been really fun to be in this space and we're really happy to be here and talking about kind of what the future is and what we're seeing as the future of the whole technology stack um so let's uh jump in and talk about agents as a service has anyone seen this Acron before maybe when I sign up um the uh there's a lot of talk about how soft traditional software completely changing and I want to start off by going into some details uh about why that's the case so um CEO Microsoft um has recently
said Hey sassin in the past AI agents are the future and what does that even mean um but it's pretty interesting that that he's now calling uh sass think in the past um there's a a podcast with the founders and some of the managing Directors of Y combinator and all these guys about 3 months ago started saying hey agents might be the thing of the future and it could be actually 10 times the size of the traditional stat Market uh for the agents um or the market of the agents um now why is that the
case so here's a traditional Tex stack where you have this is like software right I think Point points the screen points off the screen so there's a data layer I'm just really simplifying this have data Layer with with this is the software there's the data layer there's a logic layer in between we you're actually interpreting logic and then there's a UI in ux layer at the top sometimes these are intermingled but let's just make it simple and say this is the three major layers with software and how it's built so why are agents the future
and why is this accelerating so quickly so let's talk about the logic layer for a minute traditionally when I had to Code logic I had to sit and I had to do a bunch of code I had to write a bunch of code it took me months to do to write code for logic of I have certain pieces of information want to come in write the code make sure it works perfectly blah blah and I get out but that traditionally takes months or years uh to do now with AI if you're actually passing it through
you can actually use AI as the logic ler in between so AI is doing all the logic thinking for you you Don't have to code if this then that if this is the case then go do this look up do a lookup table with whatever AI can just take the inputs and give you a really interesting Dynamic output and by the way it's almost instant results when you do it uh so this is actually AI completely transform the logic layer in software so let's go back to the other layer so so stand layer we're not
changing this um the logic layer is what now can be done almost instantly with an AI um and then the top is the UI and ux layer now ai gener ai was around for a long time but how many people heard about it before chat probably not very many right the the point is that they were actually because of the logic component they're actually able to create brand new AI or UI layers the ux layer which was a chat and they and everyone realized oh as soon as they had chat to interact with this logic
layer boom it like the Biggest explosion of any software in history right they hit over 100 million users faster than anything uh by by multiples and so this is this whole slide actually I know it's simple but if you really think about this in the whole software stack this is amazing what's happening here and new UI I actually think a lot of the future of how we interact with software and all the technology is actually in the UI and ux layer and this is very under thought And utilized right now because we're actually doing a
lot of dealing with with the logic player of AI but there's brand new interactions that we will be able to have that we've never had before hence the reason for chat or there's a lot of audio um conversations you can be having with your phone now with AI and things like that but there's there's a lot of interesting stuff that's coming around so challenges as you're thinking about AI Ag and to rethink these pieces Um so let's go back to what or what are the results from all of this so SAS now turns into as
uh software more fun to say actually software interfaces um well actually the old ones feel stale and uh and I'll show you an example in a second but how many people do traditional Google search still compared to AI search in this room that's a perfect example of it feel stale and it's old because the UI and and the layers that were currently existed are old and and Our entire technology infrastructure is right there or making this transition right now um the other piece the results of these agents guess what you had all of these there's
several DCS in the room and folks who see a lot of business cases one of the main things is hey is this a big enough market to spend the time for the money months of software development that it used to take to solve this problem guess what that's no longer the case you don't need to spend Months of software development and there's there's an opportunity in a service space that I could do something and it might just take me a week to do and there could be an opportunity in that service space that's you know
$10 million $100 million something that's actually worth the ri because the software is so cheap to build now at this point with that logic there so quick right so there's actually this is the huge piece here is that service Businesses probably see this enormous huge boom so Scott gallway uh kind of Quasi famous NYU Professor his goal is to be the most influential business influencer uh on the planet uh he's fantastic guy I I was on a stage with him years ago my former Amazon background he I love follow stuff he shows um and his
predictions he's got great methodology and one of his big method one of the big predictions for this year is that uh Service he calls it Service as a software which think about um the ultimate result here is because AI can do so much and you can build these agents with all these customized logic uh could you get to software margins on a traditional service business traditional service business has maybe 10 to 20% margins and can you get to 80% margins with a service business by using agents and software um with the way that the AI
built and so that's his Prediction for 20125 I actually think it's going to take one more year to fully get there but that's one of the predictions but this is the most interesting reason this could be 10 times the size agents could be 10 times the size of traditional software all the way from seoa um software Market about $650 billion right now the AI software markets was Tiny it's about $3 billion but the whole services in software Market is about $1 Trillion so there's this huge opportunity of all these microservices and things that would have
taken way too long to code way too long to build the interfaces for that now you can shift into what we call AI agents and using AI as the logic uh between all of these and you can build out all these systems and the opportunity is actually quite large and there it could be cut up in a small micro it could be you know services that cover a full slot use use cases There so let's go into again as the new market comes and coalesence how do we Define an AI agent I've seen this defined
very many ways by very many people um and I actually do my own little uh LinkedIn survey and I said how do you def find AG um what was interesting is that not very many people chose the selfing learning digital to Tool uh which is what I thought more people would have actually chosen but most folks considered it an automated System using AI for an autonomous decision maker that uses AI um so let's let's actually break it down I don't think there's one definition actually like most things I think it's more like there's tears of
what is an agent but let's let's say an AI agent is anything that incorporates on a most basic level an some kind of an automated system that's integrated with AI so software where youve integrated uh an AI logic layer into it that could be Considered an AI agent um Next Step Up we Next Step Up is if you actually get AI agent governance which what that means is you have a system that does something you can actually send the results to Ai and say hey what do you think and then give it parameters and then
it sends it back hey I think you should go do this instead or go down one of these three different uh branches of information so that's the next what I consider the Next level up um and then the most advanced is uh is a swarm intelligence where you have multiple of these AI agent governance dealing with multiple agents of intelligence and that's where you start getting really gnarly things and uh you probably get a lot of uh problems if you don't do this very very carefully but you might have multiple agents that are all governing
different parts like a human organization right you have you got marketing you have Sales you have HR uh there's people who are managing these different pieces imagine that on kind of an agent level when you're building these AI agent systems and by the way you can build these so quickly because of AI right and you still stick LM inside the middle so that's how I'm defining agents for now so let's do a couple of uh example use cases of what is an AI agent versus SAS traditional SAS this was the first one showing traditional Google
results uh which not as many people use anymore Google sping and I know traffic's down 15% or at least the referral traffic's down 15% uh year-over year which is the latest report uh with AI by the way my favorite goto tool that's not my own tool is perplexity um because it has a has a search Logic on top of whatever system that I want to use as the ACT bra the llm um but this is giv me more relevant results and I said hey just give me the Results that I want to WR to BU
me now Google's all about relevance but I could say even more AI hey I have this kind of health insurance uh and I'm looking for a dental place and this wouldn't do that now but that's a good agent would do that right where it knows more about you and you can actually get exact results you want right away from it so another example um traditional um customer service we've seen a lot of excitement in this space I Think it's one of the more easy loading and fruits to convert to an agent system uh with AI
having AI agents and logic there's still a lot more to do here um by the way one thing that we've I did this about a year ago so this is an example of our our tool um what happens is whenever you have an AI call come in all the prompt does is it looks at what it what you talked about before and then it looks at about what you're asking it now and that's basically a chat so when I had it the Persona I changed the personality to be Dr jeo Mr high so it would
talk really angry at me or it would talk really nice to me and if it was talking Angry it was lie to me every time it would talk to me but it was a fun thought experiment because I basically said hey if a Taffy go down this route if it's sad go down this route and it would change the logic all the time but I built this in I it was one of our experiments I did in maybe 30 Minutes maybe 15 minutes and I had this agent that was like changing his personalities right and
that's the that's what's exciting about where we are in this space that you can change stuff so quickly and try a lot of things and there's a lot of um I come from the entertainment and kind of video game Space so it's say boring ways of interacting with people but agents you can actually get really really engaging and uh get things more interesting I Don't think You' probably do a Dr je Mr H but uh you can do something like that with an system uh let's talk about blogs for a quick second blog Generations right
if you read this now this is and this is my perspective uh here's a Blog generation service right all aboard build your plan request your content receive and review your content leave your feedback and build your team right like that's kind of a traditional thing I built uh I'm Going to switch to this real quickly an automated blog generator agent that every single day the agent goes through and finds the latest news it knows who my target. audience is it looks at everything I've written before and then writes an article it writes it again and
writes it a third time like I would if I was drafting something I would go through multiple drafts and I have different agents thinking about it each time it rewrites it and then it actually Looks it's connected to all my SEO keywords that I want that I want to embed into my blog article and it injects those into my article and then then it generates an image for me uh for My article and then automatically uploads it onto my WordPress site every single day I have 100 40 articles and I haven't touched it in two
months now I should spend more time on it and I know and I can improve it but let me I'll show you a quick demo Edgar's going to Do more of the demos than I am but I'll show you a quick example this system um and this is our platform where um we create these AI agents and systems on the simple no code platform um and basically we have access to all the AI models again we've aggregated all these AI systems in one place to make it really easy to orchestrate them so we have text
generators we have all the visual generators generating images reading images uh audio transcriptions Uh audio voice generators some data connections for the technical folks so you connected with like vector databases and uploading things like that into those we're connected to Google Sheets and SQL databases so any way you want to De with your data understanding the web is where we're strengthening web pages we're doing Google searches various things like that um and then we have these kind of decisions you can do with some of the functionality and then Everything you build is an API or
you can call it from an email or you can call it put it in any existing system you have it's really really flexible the way we built this um Ed will do a little bit more explaining in a second but what this agent system does um basically it's doing the prepping the article information where it grabs what are people talking about today that has to do with that subject U actually in the general topic area what are they talking About today so look an all article that just came in this morning um and what you
normally search for if I was writing my own bar article and then does the first iteration of the blog article and it does a reference to all my previous articles that are down here and it looks what I have written about so it's always writing about something new and then I make it sound I don't like the way AI talks so my first agent says make this sound more like Gary D I like the way he's very direct and he's he's not too salesy he's just like straightforward um I don't say right like him I
say make it more like him and real and saly so it actually changes the the hook of the article um and then it rewrites it the first time with that information and then goes to this this last time where we're including all our SEO keywords and we're actually rewriting with some more Fess uh into the article and then it Generates the image uh right here what we do is we have ai tell us um or we give AI the article we say write an image prompt to generate an image uh for this thing um you
don't have to generate your own prompts anymore and so this is the latest image based on the style guides that given it and then we have uh a Json structure that basically it's a a structure that the website would read and then we take that and we say hey this is a similar structure rewrite this New article with the same kind of structure guidance that does it and then it pushes it up to the website um and uh again does that every single day course so that's a quick examp there um if I go SL
is it going to pick up where we were yes okay um all right so there are are agents ready for Prime Time some are let's talk about two issues that are still out there with the AI agents so one issue and don't get fooled that's why I bring this up um There's a memory problem still there's a memory problem talked about this last year there's still a memory problem with AI and it's okay uh we'll talk about um a little more detail and then agent orchestration so when you think forward about your whatever you're building
how do you orchestrate all these agents and they're interacting and working together who are on their own there's no system that I know of that's built to be able to do this orchestration AWS 3 days ago Just announced a new kind of system that's trying to help orchestration but again it's going to be on the tech layer not on like the business personer that's easy to kind of see so I talk about the memory problem real um the what I like to say with AI and why a lot of people don't get AI uh I
like to say it's the left it's the left brain of Technology meaning or sorry AI is not the left brain we've been dealing with The left brain of technology for the last 40 years we are all used to dealing with technology that is exact it's precise and know it will you put the formula gives you the output that's kind of much more like the left bra and that's what that's how people are used to dealing with technology so when you go to computer you expect to deal that way but AI is this whole new paradigm
it's actually the bright brain of Technology it's creative it's uh Procedural and it actually follows patterns that it's the top and top trillions of patterns that's why it does what it does and it actually is rally over it doesn't the only reason it knows that oh I'm sorry the only reason that it knows uh George Washington was the president of the United States it's cuz it was told that 100,000 times in the patterns that it was reading right uh when you ask it how did two po ey like it might give you Some you know
some answer but maybe it doesn't fully know right it doesn't have exact data point it's just pulls from all that's generally uh trained information information set on it right so it's two different kind of Technologies someone asked me one time and said hey can I just feed it these million rows of ordered information and get the average order value um and I just give this document of a million rows of orders and I said no Actually you can't do that AI cannot give you the average order back when you would think of Technology could right
because Normal technology you could AI can what it can or general AI cannot what it can do is it can write a SQL script that would then go hit your database and then figure out the avory values and then get back and then you then you can have it so that's it's a you have to think it's a little different technology because you can't Just expect everything from it like we kind of think some science fictions so um there's another underlying piece and we were talking about this still need to um talk about it uh
so people as as people are learning there's a concept called Vector databases out there versus traditional databases they're very different they're both storing information uh traditional database is just like a map where you can say KY uh any2 and you know the Exact spot to go to pull the information and gives it to you that's a normal database a vector database is much more like a library where you're like I'm going to the fiction section and looking at things about water and oh there might be there's M dick right or there's a couple of books
and it would return all the books uh that are in the fiction section about water it doesn't know exactly what you're looking for that's how Vector databases work it's much more Um based on semantic generally semantic search unless you change your vector datab structure but so when you're dealing with um uh the exact prec Precision out of AI you're not going to get exactly what you want that's why this memory is still a problem you don't get exactly what you want because we're still working on this piece now one last piece is that um and
Google this announced three or two weeks ago this new Transformer system and They're they're working on this it's a new Transformer architecture um the problem with AI is like my 7-year-old I tell him to clean his room there's two problems so first if I tell him to clean his room he's going to go clean his Legos and make his bed and come back and be like I'm done I'll be like you didn't clean your clothes you didn't clean up your stuffed animals uh because I just said clean your room but if I tell him exactly
what to do those five things Then he'll go do them and actually clean his room he's pretty good kid um now but if I gave him 30 things to do I said hey Clean This Clean This Clean This Clean this whatever his his braid would like explode and he and he do like three or four things then come back and be like my room's clean and it wasn't clean right CU it was too much information for him to have come in now ai also works like that that's why memory is the problem is that we
are building Essentially the Corpus colossum the thing that talks between the right and left side of the brain that's being built now in sophisticated ways of dealing with this memory problem are being work on um so when you're trying to get exact data out of AI it's not doing it because these layers in between are still being worked on the technology is not solved yet and so you're getting cool stuff out of AI it's generative but May not be exactly what you want what you have to do is you have to give it you have
to give it what it needs the right context relevance it needs to be accurate and that has to be in the information that you're feeding in to get it back into the system when you start developing now I know I'm going pretty deep but we're talking about agents here so hopefully this is helping um the last these uh is orchestration I'm not going to go very deep in here But this is a problem orchestrating all of your agents as you start proliferating these agents we're thinking you know I'm thinking 12 months ahead or 24 months
ahead when companies start pulling in agents we've had some clients coming saying well if everyone's building agents which everyone should be building agents how do you know what's built how do you track everything that's happening and there's no real systems out there that are trying to track that Right now one thing that we are doing that we have in our system is you actually can do some orchestration we're actually but we're doing more improvements to make it even easier and better understand what's triggering when is triggering what's calling what um but you do you can
have these governance agents uh where it's like okay this this agent is calling these three agents and then the results go back and then does something with it and it keeps getting Smart along the line so um we built this platform we call flux prompt and basically like I said integrates about 150 different AI mod and systems into one place to make building agents really really easy and quick and we love doing it um and uh I was showing these guys some go ones I built the other day we do it all the time do
a Friday showcase where we always talk about the fund agents that we built that we um but that's I'm now going to hand it over to Edgar uh to show a couple of examples of other agents very applicable business agents we built um that he has particularly built any of these uh that that help showcase you know uh some of the power that we're starting to see goe thanks Brad appreciate that um yeah I'm excited to to take you guys through a case study is there feedback coming okay sounds good um a case study that
shows how powerful our platform is um the company that we're reviewing today Is medical exams it we changed the title a little bit of the company um they are essentially a specialists in Social Security Disability valuations so what that means is if somebody wants some disability then they have to prove to the government that this is what what they need help with and so then they go to this company and get evaluated this company um works with doctors and nurses and once all the exams are done this company goes and basically comes through All of
the documents and then puts an evaluation together and uh some of these can be really simple like four pages right and some of them can be very complicated 20 plus pages so it can take anywhere and even the simpler ones with a template and a process that they they've cultivated over time it's still can take anywhere from 2 to 6 hours depending on on the climent um they came to us and they said look we know AI can help us with this But we we're trying all these things and they're not working and U right
they're trying gpts or they're going to different llms and they can't get figure it out um within one week of working with these uh this company we were able to create an AI agent that took their process from 2 to 4 hours down to 5 minutes um and and it's it's actually really really good like really good content um I went and did just a little experiment and took the Content that they gave me right and I put it into anthropic which is an L that we love um they are known for analysis and for
writing and it barely put together two pages 546 words and it wasn't very accurate I compared it to what we created and then I asked anthropic to kind of grade itself and they and it said oh I'm about 65% accurate and then it told me all the things I missed and as you can see there's a long list of these things so No wonder our client was getting frustrated with AI right um but our platform was able to to create a solid um AI agent that now his entire team is using every single day what
this means for them is that they actually not only did it cut down on time but now they can get more clients through the process so they'll get more money as well um so I don't want to go too much detail about how we did everything I'm going to touch on a few things here and There um but just know that it's a we've been doing this for a while now we've built over a thousand of these agents and we've learned a lot of good things some of the best practices is that if you want
something really solid you have to break it down you have to break it down into smaller parts you also have to take those smaller parts and and put some effort into it you have to think about all the different pieces that that step is going to create um you have de Feated a specific example um maybe directed to know where the information is at and where you should be able to find it um one that i' love to use is use AI to request just the verification of the facts so check itself um we're dealing
with medical documents and uh we have some clients in law so these documents have to be very precise and accurate um anyway so another thing is that one of the things that our client said was I want these Documents to be very specific format and styling we've been doing this a long time we want to make sure that it's it's exactly the way that we like them and that's one of my passions I don't think people need to change their process just to use Ai and our platform allows you to do do exactly what you
like to do the way that you like to do it um and very easily okay a couple of other things that might be of interest to you guys is Shots not in the way that some of us are thinking about shots um we're talking about like shots in AI are just the explicit instructions that you give a brad mentioned earlier SAS was built on thousands of instructions when I worked in Microsoft we had a product and it was hundreds of the pages along the and that was all the rules and and all the different things
that this product needed to do with AI you don't need that but you need some there are some things A zero shock is basically products that don't require any any explicit instruction those can work for certain things uh you've probably seen a bunch of these note aggregators or summarizers right those are great you feed it your video no instruction whatsoever and you get a summary uh and it works great for for that specific use case sometimes people don't really like that and so they they would need to add a few more shots to get it
to the right format that They want but for the cost makes sense right um but if you're talking about complex tasks you're going to need more shot you're not going to need 200 300 p Pages worth of requirement docks you're going to need a little bit and and that's but the nice thing is we want you to get those few shots going and you get your process just right after that it's just gold after that right you just feed it every single day every single hour every single minute and it just produces Exactly what you
want another thing is context window and I'll go through this one quick context window is just basically how much um how much the AI system can intake and output and it's the context window takes into consideration both of those things so as you can see there's the different systems up here have different Contex context windows so you might look at this and say oh yeah yeah yeah um you know Gemini has a million tokens of Input tokens are essentially 34s of a word think about it that way um but um that wasn't very helpful was
it anyway but you might look at Gemini and say oh yeah it's got so much context I'm just going to throw everything in there and then it's going to print out this beautiful magical thing um the problem is that there's when you feed it a ton of information then um yes it's good in some ways it can go and find connections um hey there's this one Document over here there's other document they're talking about the same thing and it kind of finds those um the problem is that sometimes it gets distracted you giving it so
much information that it might say oh I like this part over here I'm going to throw it in when it hasn't it shouldn't have done that another thing is it's very expensive and it's it's time consuming right the more you put into it and the more you expect out Is going to take a lot longer so then the other option is let's just feed it the right information so that it can get the right result that's great it works the problem is that for our client to go through all of these document and find just
the right information and then and then you know hook it up to AI that takes probably longer than just like coming through and like annot and annotating their their results so there's pros and cons there So really the best strategy is a hybrid strategy um a place and our platform makes it so that you can take information feed it into any pro any step of the process and uh it makes it really easy for you to just find the right uh information to give and the right information to to take out now the other thing
that you have to consider is that in order to get the best quality of your outputs you can't stick with one family Of AI if you did that then you would have terrible images um using open AI um because that is not that great um but but every single one of these AI systems has their strengths right um Claude which is made by anthropic is phenomenal at at writing that's all they focus on they don't do images they don't do video they just write and they're really good but GPT has plenty of really great um
Pros gini as we talked about has large Contex window and and then we go into the images as well we got IAM who does that which does letters really well so it can create illustrations with letters fantastically well and you've got flux that does photo realistic images in order to have the best quality of your results you're going to want to use various different uh systems it's the same thing for voice and same thing for video um and others like calculations Etc our product we have over a 100 tools in our product and um dozens
of we have over 100 system AI systems in our product and dozens of other cool tools that are like plugged into CRM or Google search Etc it's so easy for you to turn and combine all of this technology and get the very best quality last thing um it's very familiar ux if you've used if you use Miro or Lucid or fig J now you can use our product it's the same concept you Have two walls on the left you drag them you connect them and voila um the other thing too is that our product makes
it so easy for you to actually customize every part of your your process so let me take you through a demo this is um let's and I'll go through this quick cuz I know we're running out of time this is uh the company um anyway but this is the product that they wanted right um at the very end what they Wanted was this document it's very very specific very detailed and as I mentioned you have when you try to do this with AI it's not going to get you really good results so what we did
is we basically made it possible so that you could go in here and take all of the documents that you have so there's the the Social Security page that you need there's a the transcript of the the conversation that they're having with the nurse the patient exam the medical Vitals and any background medical records all they do is upload those documents for each client and the very first step that it does is it comes here because they have to be accurate the first thing that we're doing here is basically double cheing every one of those
records and saying are the Le the same person and also what are some things in these documents that don't make sense and so it it outlines all of those things so now the the boss can Send this out to their team and now they can recognize and look through it and be confident that what they're doing is correct or where they need to be careful and then what we do is we take every section that that are similar sections and we create AI for it um for every single step and I'll show you just very
quickly instructions we put a lot of really great thought into the instructions what are the things that you should be looking for what are the The different uh formatting uh what are what is some of the definitions and Etc um then once we do that we'll put it all together into a document and uh now they can they can uh download it as a doc X if they want they can send it to drive they can do all sorts of things they can have they can put it into a different flow and analyze it further
if they wanted to right so um anyway that is my part of presentation I think we are a Q&A or did You want to show something else yeah so thanks Ed um so let's move on to uh we've got you should come up here stay up here Q&A uh in just a minute um we there's a couple other cool agents just uh brief examples just want to call these out um we've done uh interesting I think this is actually an interesting spot where things are moving chat Bots um why are we still filling out forms
in 2025 but a chatbot essentially could have a conversation and push all that information into a form into database system we're doing that now it's pretty cool um we have like really really long conversations with people and record like he out of the that conversation we would work with some Consulting companies that are having big long interviews with with people inside the company maybe 100 interviews they want to find out you know job performance and Whatever else you can take all that crunch it through Ai and actually spit out into like a nice uh data
set really fast um then so I actually the whole huge marke right there um the lead scoring is something we've been doing for for a company now we have about 10,000 this one company uh real estate company on of Dubai they have about 10,000 leads running through every single month and it's a lead scoring system one of the coolest things about AI is that you can now do things people are doing things they know they should have been doing but never did which is like scoring leads right A lot of times we're like just dump
into the sales team right but actually if you score these leads the their performance is now uh increasing on conversion rates and and close rates and this this system basically takes the new lead that comes in actually does a little internet research about each one of them and then Pushes it in and scores the lead according to their Dynamics deping want the lead scort uh the sales team coaching and Reporting we have a um customer who's doing actually several uh who all the sales calls that come in is connected HubSpot have another one who doing
with with uh sales of course um any call that comes in it goes and it reads interprets that call and then scores it according to how the sales manager wants to score it's like he just Changes dynamically like how how he wants his team to be doing things there's no preset software he just goes in and just types it right in the system uh and then he gets a a weekly report about everybody so we take all their calls about every single agent all El they did and then CR that an AI and gives a
weekly report for the sales manager on that system um and then we have a cool well just another uh interesting one I think is more creative Uh is this storybook app that we've partnered with with this group do a pretty cool story book app you put in something uh simple and you basically create a whole story book in a matter of minutes with images and everything printed off personal stories family stories whatever you want to do with um it's pretty pretty cool system so uh oh important questions last thing uh anyone feel free to uh
Reach Out out to us at a point uh team in enhanced AI we're also Launching I asked back in November uh who wants to know who wants learning about Ai and AI agents teaching them and I got um a lot of really positive responses from it so we're actually launching a course that we're going to do uh limited seats but we have several people signed up already um we're going to be doing cohorts and uh first cohort uh we're doing a course was teaching people like the Deep uh pieces of building AI agents particularly Building
them on our and inide of our platform and that's what what this is all about um but we love to help train and corporate training on on doing some of this too we feel like we just need to tell people how to do this stuff so um all right let's let's take a pause there and go to any questions in the room about anything we talked about or even anything we haven't hearded about do you have modules that you can rent out or give by subscription yeah uhhuh yeah we We've actually built exle that if
you don't want to do the whole platform you could actually just do one or two of these modules the other thing is that we've had some clients who are building their own modules and then selling them and we're we love it we think it's great so you can build a module build some really simple software on top of it and put it inide of system in fact we have one that's um uh some of these guys in this room uh have built a pretty cool uh Marketing agent um agency kind of module that's pretty awesome
that they using with the platform so right what other questions how sophisticated does the uh uh front end integration now look you know when you take what you're building in the back end which is very technical very PR now I want to build an app or I want to build a platform on top of this like how do you think about that and is that part of your process or is Generally where the issu is going from the back end to front end you know yeah so and that's why I like that slide the very
beginning of there's the data layer the logic layer in the UI layer we've been primarily all in the logic layer over the last couple of years we've actually been doing stuff in the UI layer now also and you're this is one of them so we built something we call a form Builder like a better term I'm sure come we come up a better name for it but Basically you can input it's a form that you could after you build a flow you can then um create a form that someone just spills out and those are
inputs inside the flow that can trigger anywhere inside the flow so this is how we're solving it the all of these flows that we build this is foundational everything you build is an API so you can call it from an API you can call from a web hook you can just say you know send one shot to it and and you can respond back to Web hook respond back to an email there's an API system inside of it so you can call apis during the workflow so you can send data or get data um you
can go and very the flow um but then there's also uh this kind of frontend user interface here here we also have a chat system that you can build your own customized chat agents inside of our platform and that's one of the these two are like the more prent we're seeing that's why we're spending some Deb time On building chat agent uh that's really flexible and we're still building some pieces to make that even more powerful and then kind of Reform agent that you can use but if you wanted to you can just put it
in we have plenty people who are just putting in the software and just using our platform as a quick logic layer to the existing sta that they have with their AI or whatever set their set stock when creating that agent flow using so Many different models and made so many different steps how do you keep track of things like loading and costs for the AP see what's yeah so yeah he's asking how do we keep track of rate limits and how do we keep track of all the calls so we actually built there's a big
process we did last year uh we built a a currency conversion for all the all the models and all the systems so we actually have a currency verion layer across every single model that's in our Platform there's a cost there's an input token cost output token cost sometimes it changes if it's audio and all that so we built a whole conversion layer so everything converts into a common what we call flux token inside the platform and uh and so you know token usage we actually every time you run something there's a dollar amount that will
show exactly how much that cost to run that uh execution of the flow make it really simple to like optimize your your Workflows and your agents and uh we also um you don't need to have an account with everything with all these models and systems we have a kind of a master account now you could you know uh work through putting your own but really it's a master account that's it makes it much simpler so you don't have to have an account with you know whatever 150 different systems that we have set up in here
you're like well let me try Gemini let me try you know all these other ones To make it really simple to use yeah imagine excuse me the graph that you got at some point is just some sort of like a Json representation right have you Tred to take your particular products the jonhs hydrate one of those grabs and have LM right your for you that's right so um that is the division and we're working on it now um actually one of our guys in here somewhere uh is actually working on that project right now so
the Idea would be really cool where you come in now we're going to have a a smart B where it will help you build exactly what you need right so it's over there on the right side you're telling it what you want to do it will understand what's on your diagram or um on your canvas and then it will generate noes it will generate connections and it will also edit things so that's uh that's a really cool exciting thing be the case the back end what are You using open source sdks or are you guys
writing IND call for each mod you have so we for our own system we have all our own libraries uh that's running and tracking lot these things we have plenty of Open Source pieces that are put in software like almost every software company does right but for our own pieces we're tracking and having some of our own uh you know what are all the definitions to create that AI agent right because that agent Would need to know definition of terms and a bunch of other information that we work through but the vision is basically come
in here tell them what you want we working prototype uh you tell it what you want it builds the whole thing out for you and you're like cool that's a way of getting started right that's even we have 120 we didn't mention 120 templates in the platform now but our thoughts like wait that's not I mean that's old right templates are old like AI the identic stuff is the future instead of a template to tell it what you want it'll come up with three different versions of it then you can go from there so that's
definitely yeah Vis how do you think about The Human Side specifically like overseeing these agents right get more trying to operationalize this business making sure stay positive or just kind of from that orchestration view agents AR cently they have rights They don't have responsibility so at some point a human has to have enough insight to say I'm sure we compliance B whatever that responsible for so I'm just curious about that our action you guys as this system you knowers more and more action what systems You're Building or how you already have buil yeah that's a
really that's a really good question Um yeah um well right now one of the things that we built already in place is that we give a lot of data to the evence um it it makes it really easy for them to to access any agent uh understands how much they cost to run understands like what they actually do they can turn off systems so if you want somebody using CH GPT you can turn it off so that nobody has access to it so we're giving it a ton of capability there that's just the beginning of
that then we've been Talking about um adding a dashboard where you can easily access a lot of these different agents uh much quicker a lot faster so you can monitor them a lot better does that make sense yeah so and then the dtail on that the um right now in the system when there's a lot of human things that you can do so we have an email something that you can kick off an email whenever something triggers right so I actually do get an email with my blogs C every morning but I can have An
email where it it sends the information to me and then and you can have any outputs you want in these notes I could put 50 of them so you could I imagine you know how you orchestrate the r we're like hey I just want to see the output every time it spits it out and then once it's done okay now move on to the next step the next process so you can be doing triggering these to send to you proactively so you're tracking a lot of stuff the stuff that happens but I Think one thing
that does get under uh um appreciated is that we have this whole system for uh orchestration and management of all the llm models of token usage like you can turn it all on and off and throttle it that's that's the benefit of having a platform like this that our customers have have loved by having consistent right there yeah so if I can thr this on this was not planted but we we're Advanced AI users we love it and I just got to say like It's incredible like it is the limit of your imagination and how
skilled you are on a technical base like Ping some technical people ask for their help but like the the creativity is like limit your imagination limit of what you can do building po your envir for yeah yeah so those pieces our platform primarily acts as a pass through um and We don't store the information like if you're connected to the database you pull it pass it through and then it erases uh like instantly so we have several medical clients that are doing that there's more work to be to be done I think in certain cases
uh to make sure um every you know pass all typ of stuff but yeah we we've been aware of especially as we've been dealing with plenty of H medical clients yeah and the nice thing about our system too is that You can use whichever Al you feel comfortable with if if you want to use open source I if you want to use um you know CH or whatever um it's just whatever you feel comfortable using it was the same question I just that's a huge like black box for me you know it's with any company
you're going to have a data set it's are you willing to share that with it pass through Cho the right one that's what that's you can do open Source not for particular L want to be using um just curious when you pulled up like when you asked your model and it said that it got 99% accuracy what metric was using to get that accuracy um I'm going to be completely honest uh it was our client like he he and I reviewed many of them and he got to the point where he's like well this is
really good got to the point where it was even better than some of the manual stuff that they were working on and so that's What we felt you know um honestly felt like that's the the level that we got to um now if we wanted to go through and actually analyze that completely doable that's 100% doable right and when can go in um one thing that we could do is take that output and run it through a couple of other AIS and just say okay I need you to fact check all of these things which
we could absolutely do all right I think we're out of time but we'll we stick around for a little Bit happy to answer any of the questions but thank you very much
