273: Where to Start with ChatGPT - with Shanif Dhanani

A lot of people are talking about ChatGPT. What is it? How can it work? How can I use it in my business?Fortunately, Shanif came on the show Shanif Dhanani. He is an expert at Chat GPT. In fact,he's a consultant he helps businesses Connect ChatGPT into their business for use it in all sorts of ways. And so he ran down a really good list of things you can do with jazz up and other AI like Google's Bard.So check this out. If you're wondering where do I start with chat GBT? This is where you start.

Welcome to Scale Your SaaS, the podcast that gives you proven techniques and formulas for boosting your revenue and achieving your dream exit brought to you by a guy who's done just that multiple times. Here's your host, Matt Wallach,

and welcome to Scale Your SaaS super excited to have you here. Thank you very much for coming. My name is Matt. Our goal today is help you grow. So we're going to talk about some amazing things and I'm really excited for Shanif Dhanani. who's with us, Shanif,how you doing,

I'm doing great,man. Thanks for Thanks for having me on the show.Absolutely. I'm

really glad to have you here. Let me make sure everybody knows who you are. So Shanif is a ChatGPT consultant,definitely something that I am in the dark with. I know everybody's talking about chat GPT. But Shanif helps businesses understand and build systems that connect their internal and proprietary systems to chat GPT,enabling internal employees,customers, and other stakeholders to reap the benefits of using large language models on their own data. Sharif has built machine learning analytics and software systems at large companies like Twitter and Booz Allen smaller startups like Aptio, where he was the founder and tap commerce, where he was one of the world's first,which was one of the world's first mobile advertising platforms. So he absolutely knows his stuff. Once again,Sinead, thanks for coming on the show. It's my

pleasure. You know, I, it's always great to chat with folks like yourself who are curious and interested about this stuff. And it's, it's just fun. So I appreciate you having me.

Absolutely. So tell me what's been going on with you lately. And what's coming up.

You know,lately, it feels like the worlds sort of change I was, you know,before the show, we were joking that six months ago, there was no such thing as chat GPT consultants, or even necessarily chat GPT. So, a lot has changed for me in the past past three,four months. You know, I've been doing a lot of work with with companies who are really interested in this technology,maybe they don't know much about it, or what it does. They just,they just know that they kind of have to use it. Because if they don't, their competitors are going to use it. And so it's been a busy past couple of months for me, but good.

I can imagine. So,how did that come to be? By the way? How would you say I'm gonna be a chat? GPT consultant?

Yeah, you know,it's funny. Um, last year, I was actually running my startup, the one that you mentioned after you and we were we were using AI but totally different way, what we were doing was helping ecommerce companies basically predict what their customers were going to buy next, or, or which ones were going to churn, and then helping them create marketing campaigns from that. And then eventually,we sort of went through that we realized, hey, look like the economy is not doing so hot,ended up shutting it down. And then I decided, look, I really want to, I really want to keep working with AI, I want to keep working with technology. So let me do some consulting. It's a way for me to get an understanding of what businesses want, and get a feel for what problems they have. And I did it for like, two, three months, and I got no business man. And so I was like struggling, I was trying to find customers. Not easy. And then eventually, I started doing a couple of projects with companies who wanted to just see what chat GPT could do. And I started doing more and more of those. And then as soon as I sort of advertised what I was doing there, I got a lot of influx. And so you could say kind of fell into it. I sort of purposefully fell into it. I knew I wanted to do something with tech, but I didn't realize it was going to be chat GPT and it was one of those things that the market just sort of drew me in.

Very cool story. I love it. Now as a chat GPD consultant, how do you approach understanding a client's needs and tailoring the solutions accordingly? I mean, right now,most clients, I would guess,have no idea what it does or how it can do. So I bet it's hard to figure out how can we approach this? How do you do that?

There's there's a lot of education. So there's sort of two types of customers,let's say there's the ones who are pretty savvy and who know a little bit or maybe even a lot about what GPT is, you know,maybe they go to the website,they use it sort of on a daily basis, maybe to do some research or something. And they're like,hey, wouldn't it be cool if we could do this with our database?Or if we could do this with our existing SAS product and maybe add a chat app on top of it. So those folks, they come to me and they're like, Hey, I know exactly what I want built? Can you do it? And then I'll talk to them. I'll say, hey, you know,you might want to think about this. And this. If all looks good, I'll start a POC with them. Then there's the other side of people who are like I have, I have no idea what I want. I know this thing is awesome. I just want to I just want to do my job better,faster, stronger, and for them,it takes, you know, a handful of meetings and sort of diving deep into what it is that they have Actually you're trying to get done what it is that's taking up a lot of their time to get a sense for what they need. And a lot of times, it's not judging beauty, a lot of times, it's something totally different.Maybe it's more traditional AI,maybe has nothing to do with tech. And so for those folks, I try to walk them through what it is that they're trying to accomplish, and help them out as much as I can to see if if we can get a project up and running for them. Or if we need to do something else.

I love it. It's good framework, because in what we're doing in software, which is what a lot of our audience is, it a lot of times people don't know how to solve their problem. They just don't they have a problem. Maybe sometimes they don't even know that. And so we do need to do some deep discovery to understand what really are the challenges,sometimes they come saying I need to do this. But in many cases, they don't know that they just know that there's a problem. And it's us that needs to be the expert to help them understand what's the best way and like you said, sometimes it's with Chad GPT. Sometimes it's with that some of this. And so being that expert in guiding and advising is really critical,right?

It's really important, especially now, this is such a new technology, like I've been doing machine learning for a long time. But this particular technology is really new. And so a lot of people,either they have this amazing expectations of what he can do,where maybe it falls short of reality, or maybe they just don't even know what it does.And they're like, Hey, I have this problem. I just don't know how to how to fix it. So educating people is really important. And it's something that I'm not a sales guy by trade. So what I like to do is just sort of be helpful and sort of consultative. And if I can help someone by building a project for them great. And if not, I'll try to do what I can for them. Well, the

good news is that I am a salesperson, but I took me a long time to learn, you actually sell more when you help. So you're helping you're educating, you're going to actually sell a lot. So kudos to that. What are the advantages and drawbacks of AI language models? What what what do we learn in there?

You know,there's, there's a lot of really cool thing. So let me just start with what what the heck is a language model, just if people are familiar, you guys probably have gone to the chat GPT website and asked a question and seeing it sort of type out a response to you. language models basically do one thing, they predict the next set of words that come next, you know, you give it something and they'll predict the next set of words that come next. Now, as simple as that sounds, it's really powerful. Because what you can do is when you predict the right set of words, you might look like you're thinking reasonably or rationally or you might look like you're reasoning about something and you can open up a world of possibility. So the benefits are, you can start to automate things a lot more effectively than you might have been able to in the past. So maybe whereas before you had sort of, maybe you had something that was more like robotic process, automate automation,RPA. Now, and that's like basically taking a set of rules and just coding it up. Now you can have sort of this machine start to, you know, kind of think about things for you. And then based on its output of what it concludes, you can do different things. So automation is huge. I think you've probably heard a lot of people being really productive with it. You know, they a lot of folks are saying, hey, look like I don't even Google anything anymore.I'm just typing it into chat, it gives me the answer. I'm a software guy. So I use it to code and it helps me code probably, you know, saves me a bunch of time, by giving me answers and things that I would have had to spend a lot of time looking up. So it's a way for one better automation and to just being more productive by wasting, you don't have to waste as much time now trying to find things or learn things as as much as he used to.

Very cool. And then for SAS businesses, which a lot of people listening are, how can the SAS businesses use AI to drive their growth in innovation?

There's a few things so you know, most SAS businesses, there's like, 80% of people are coming to me saying,hey, I want to be able to use chat GPT with my own data. And so, you know, what does that mean? Like maybe they have an internal knowledge base, and they want to let their internal employees search that knowledge base, you know, maybe you're a first year employee and you want to know how much PTO you get. Or maybe you're a customer success person and you want to be able to answer a customer's question on a quick chat bot. You might have a bunch of text data or knowledge, you know, and you want to be able to have chat GPT you know, ask a question to chat GBT. Now the thing with chat GPT is it wasn't trained on your data, it doesn't have access to your to your private data.Right? So it's not going to be able to answer a question. And so a lot of businesses want to be able to add to ask a question of their internal data, and then have a reasonable response, come back and be confident that that response is not made up, or what's called hallucinated. And so that's sort of where SAS companies are coming to me, this could apply to customer success.So maybe you're trying to make your customer success seemed more efficient by spending less time with each customer. Maybe you're trying to automate away your whole customer success team. I don't think that's a great idea. But it's certainly something you could try to do.Other companies are doing interesting things. Maybe they've got millions of pieces of content, like user generated content. And they want to start to categorize those content and show how categories change over time. This is really useful in the world of E commerce for people who have, you know, lots of user reviews and things. And then other companies have an existing SAS product, and they might want to just use chat GPT like may We take chat GPT and make a chat extension of their existing SAS products. So the use cases are varied. But if I had to give one use case, it's mostly hey, look like I have data, I want to be able to ask questions of it and know that the answers that are coming back are not beta.

And so is that what you do you kind of integrate it into your tech stack? Or how's that work?

Yeah, you know,a lot of what I do is sort of building the pipes between your existing tools and your existing data, and chat GPT. So there's a couple of steps that go into that. So you know, if you were trying to just go to church abt asked a question about, let's say, your Salesforce account, it doesn't have access to your Salesforce account, it doesn't know what you're trying to ask.So the first thing I'll do is I'll build the pipes to let you connect your Salesforce account to some sort of central database. And then I'll build some sort of chat interface where you can ask a question,and then I'll connect that central database with chat GPT.And, you know, I'll have some,some pipes in the middle that say, Oh, you're, you're you're talking about Salesforce. And you probably want this and so I'll make sure that chat GPT has access to your Salesforce account through my own API. So it's really building the pipes,you can think of it as like middleware that connects all of your existing data, your systems, maybe you have documents on Google Drive, takes all that and makes it available so that you can talk to your data, which at GPT, whenever you need to.

I mean, I think that's critical, because there's so many companies that have accumulated these drives of tons and tons of data and information and these backlogs of stuff that would just be mind numbing to try and search through figure out. So to be able to have an AI in there helping you out and just providing answers for you to be amazing. I bet people are loving it.

You know, I started slow, I've sort of done a handful of early projects with early customers. And they they do tend to like it a lot. You know, they they're excited about the idea of being able to get access to their own data with just a quick question. Because right now, like you said, they might have 1000s, sometimes in some cases, millions of documents and just kind of a pain to get access to them. So they're excited. I think that they're interested in the idea of improving their productivity,and hopefully making their existing employees a little bit more effective at their jobs.

Very cool. Now,Google's come out with theirs.It's hard. Yeah. And I've seen a bit about that. So how does Bard compare to chat GPT.

I'm really excited about the industry, I actually think Bard is probably going to catch up. So right now you can think of the world of large language models, kind of like how I think about the cloud. So AWS is the leader in the cloud, but you've got Microsoft, Google, you got all those guys. In the same way,chat GPT right now is sort of the leader with large language models, they're widely considered the most, the most useful. Now, that doesn't necessarily mean they're always the most accurate, although I think they are. But they have built a tool that's able to chat with you sort of in the most natural and effective way possible. Bard is is hot on their heels, they're close behind bars is something that has, it doesn't have as many guardrails, and it's sort of not as effective in certain areas,for example, maybe cogeneration,that things, but Google is going to catch up really quickly. Now outside of just Google, Bard and chat GPT, you've got probably 20other models that you could use,but I would say, you know, to answer your question, Matt chat.GPT is the leader right now. And they've got it's their market to lose. And so I'm really excited to see what happens next with them.

Do you see them continuing to innovate and make it better and better?

Oh, yeah, you know, it's one of the it's funny, because as good as they are now, there's still a lot of issues. Right now, for example,one thing I'm doing is helping a company categorize, you know,nearly 80 million pieces of documentation. And I run into rate limits on Chachi t. So I can only categorize maybe 30documents at a time. And so you know, I've had to write this write code that normally, if this wasn't the case, I would sort of write code that can take all these documents into a huge chunk, all at once. But now because I'm limited to 30documents at a time, it's slow.It's, you know, it's sort of brittle. I think they have a long ways to go for improving the rate limitations. If you're a paying customer or chat, GBT,you're still getting things like, Hey, we're overloaded.Sorry, you can't use us right now. And so there's a lot of issues that they're working on.I'm sure that they're going to solve it because they're really,really smart over there. But you know, like I said, it's so early right now, there's a little bit of a growing pains. So I think they're going to have a lot of really interesting things coming down the pipeline in the next six months.

Very cool. What would you say is the best way for someone to learn about these AI?

There's, I'm kind of a proponent of, there's nothing like trying it yourself.And the cool thing about chat GPT is you can just go to their website, free ask you to sign up for free ask a question and kind of get the response right away.It's instantaneous gratification. And I would say start playing around with just the website. And if you're like me, you're gonna have sort of this lightbulb go off over your head and you're gonna want to dive In a little bit. So once you're sort of at that stage,you're gonna do a little bit of research, maybe you're a coder,and you want to plug into their API, like I said, you're gonna start reading about their API's,and maybe building a couple of toy apps. Or maybe you're going to just start using existing tools that plug into chat GPT.And so the best thing to do is just go if you haven't used it yet, try it out. I was a skeptic myself a few months ago, but it's really cool. Try it out.And then once you want to be able to use it for yourself, you can start building your own applications, if you've got an engineering team, or using like an outside consultant like myself to connect it to your own data. So start small.

That's great advice, maybe I'll go play around with it. I have heard a lot of people in the content side on businesses like marketing, they like India, now I have seen that Google and other search engines can identify AI written content. But some of the best content people said don't let it write your stuff, let it give you an outline, so that you can write it and and you put some human element into it. Have you heard that the same thing

I have, it's it's sort of a double edged sword or a double edged coin here, because on the one hand,you have this capability of churning out large amounts of content right away. On the other hand, one, you know, the content is not going to be one frequent criticism I've heard is the current content isn't as human as somebody who just wrote it,like as a human who wrote it.Now what that means, you can argue, hey, look, maybe it's got a little bit less character a little bit less humor, it's,it's fine. It's good for producing fact based content,it's actually pretty good for producing blogs and things too.But you know, a lot of people say a lacks a little bit of something. And so what I've seen work really well is using chat JpT is something to maybe you produce an outline. And then maybe you have it produced the highest level topics for every paragraph or every document. And then you can use it to fill in the content, but you probably want to create your own content,and then use it to refine it, or use it to either use it to guide you with an outline or use it to refine it, I wouldn't have at least not right now there is a risk and having it produce everything for you. Because that, like I said earlier,there's no guarantee what it produces is 100%, factually correct. And two, I think that people are going to get better and better at identifying AI generated content and sort of glossing over it. There's something to be said about the human element. And so it's one of those things where you want to incorporate it into your workflow, rather than just have it totally take over.

Yeah, I think I think that's great advice. I've heard the same. So what do you think are some of the common mistakes software leaders are making when using AI with their business?

It's funny, you know, I used to be, I used to guide people on sort of machine learning projects. And I still do. And one of the things that they just they they need to start doing is just collecting their data. As simple as it sounds, a lot of people don't even sort of store their data.Or maybe they have user interactions or chat messages that they're not saving. So the first thing I always recommend people to do is look, data is cheap, you can store it, and you don't have to pay more than a couple of dollars, even with large amounts of data set. So just start storing it. And then I think a lot of people run into the, into the roadblock of trying to force using AI on something that doesn't really need AI. They're trying to like fit a they're trying to find a solution. Or, you know, they're,they know that AI can do powerful things. But they might not have a problem that's required by AI to solve. And so the second thing I would do is whenever you're thinking of taking on a new project, start with the problems you're having,or start with where you think you can get better, and then think about the solutions for that problem. And if AI happens to be one of those solutions,then you can experiment with it.But if if it's not, you know,you're the risk of being rude the risk of doing an AI project is you're spending all this time all this money, paying all these engineers to do something that never comes to production. I think I read a stat somewhere that, you know, a huge amount of AI projects just never see the light of day and never go into production. And so there's a risk that you're going to waste your time and money on that.

So it's almost like people are trying to invent solutions for something that's not even a big problem.

I think that either they don't know what their problems are, like we were talking about earlier, they don't know what their problems are, or they sort of have a problem. That's not an AI problem. And they're trying to apply AI to it. And when they do that they're sort of not setting themselves up for success. When it comes to building a project.People are smart, you know,they're everybody these days is,is really busy. They're really overwhelmed. I'm sure everyone's looking for automation solutions. Sometimes AI will work really well and but not always. And so, you know, just make sure that what you're doing can be solved by AI before you start a new project, sort of what I'm thinking.

That's great advice. As we wrap up, if we were to kind of give a piece to somebody who's listening, they can take it away. What advice would you have for an early stage software leader who's not using AI now but wants to look at getting AI into their company?

Yeah, as as the flip side of what I just said, I think that there is a lot of opportunity for people to use AI AI and automation is sort of the easiest way to get started, you know whether you're, if you're doing something tedious, and you find yourself doing it every day. And it's not mission critical, especially if you're a small startup and you're resource constrained, if you're doing something that doesn't have a lot of value, outsource it, and you can outsource it to a human, or you can outsource it to an AI. And the more you start to automate away these things,and the more you start to systematize your processes, the better off you're going to be.So as a person who's done a couple of startups in the past my my opinions would be well,one you know, know your market and know your customers before you dive all in, but to once you're all in, start to automate processes and things that aren't crucial for your business's success. And then if you do think that you can automate things that are crucial for your business's success, just plan it really well and use someone who's an expert in machine learning or or software engineering at the very least,to help you think about how you go forward with your processes.

I love it. Super cool advice. I think we'll look at that ourselves internally.But Jeff, this has been great.Thank you very much for coming on the show. How can our audience learn about you and what you're doing?

Man I appreciate you having me so you know I've got a website you can reach us,reach me at lo que Civ ello see you si ve.com My name is Sherif Sharif Lok, use of.com. Feel free to shoot me an email if if there's anything I can help with. Love to be resource if again.

Okay, cool. Well put that into the show notes. If you're listening, click that.You'll see all that right there.But Sinead, this has been great.Thanks so much for coming on the show. Hey, Matt, thanks

so much for having me. I had a great time to chat.

Likewise, I learned a lot myself. And I hope everybody else out there. I hope you guys did too. Thanks for listening. And thanks for watching. Make sure you're subscribed to the show. You don't want to miss any of the cool stuff we have coming up for you. So hit that subscribe button. You'll get all the goodies. It's been great having you Thank you. Take care. We'll see you next time.

Thanks for listening to Scale Your SaaS for more help on finding great leads and closing more deals go to Mattwolach.com