In today’s episode we’re joined by Mark Good, AI expert and Founder of AI Force Training. Mark provides fascinating insights into the world of AI and explains how Salesforce professionals are currently using GPT tools and where the opportunities exist for future transformation. He talks us through some of the key points from the Salesforce AI Day, and shares the vision of his company, AI Force. We learned a lot from this conversation with Mark, including the best ways to approach interacting with Chat GPT, and we hope you get as much from the conversation as we did.
You can follow Mark on LinkedIn here:
https://www.linkedin.com/in/mark-good/
And his business AI Force Training here:
https://www.linkedin.com/company/aiforce-training/
GPT Dreamin’ is happening in August 2023 and you can find out all the details here:
Enjoy the episode!
Ben (00:00.939)
Mark, welcome to the show.
Mark Good (00:03.062)
Hey Ben, thanks for having me. Excited to be here today.
Ben (00:05.451)
Not as excited as me because I think this is the topic that people want to talk about right now. And especially with, as we’re recording last week, there was Salesforce AI Day and I’m really excited to delve into the world of AI and how that’s going to kind of impact the world of Salesforce. So, yeah, we’ll go through all of that. But if you could just kind of set the scene, tell me a bit about your early career and how you first started working in Salesforce before the world of AI?
Mark Good (00:34.294)
Yeah, back in, I guess it was like 2015, 2016, I was working for a company whose sole customer was the US Navy and I saw a great opportunity to use Salesforce inside their company. I learned about it from a really good friend and he talked about all the capabilities with me and I was like, “oh, you know, it’s interesting”. So I pitched the company I was working for on it and they were hesitant. So I did three pitches and finally got them to get on the hook. So I did a greenfield implementation of Salesforce there.
And I loved the software so I was an Admin all in. That’s sort of how I got my start in the Salesforce ecosystem to begin with. And it just so happens that around the same time I started learning a lot about artificial intelligence. Around that time you had like, DeepMind and AlphaGo, Google products, doing things like beating humans at board games. So I started researching and reading the white papers that were presented by a professor Dr. Carl Jung, and just tackling all these really nuanced, machine learning topics, neural nets and how they get trained and how they were being trained at the time for very specific things like image generation, text generation, manipulation, detection of animals, birds, things like this, fluid dynamics in photos, light dynamics and photos. And it was infinitely fascinating. So this feels like my time, right?
It wasn’t until early 2022 when I started realising “hey, I think that there’s a space for AI inside of a Salesforce world”. And so I started this AI Force but the AI that we know today that’s blowing up is generative AI. And so it was still relatively young at the time or not easily accessible to everybody. And so the opportunity to evaluate what it could do inside the ecosystem was still fairly low. But I had some ideas, right? But they were in my mind, it was years until this could even be possible to implement. So I started AI Force, and then as time went on, generative AI started getting better and better, and you had companies like OpenAI open up this playground where you could test out different things that you could do with AI, and that’s prompt engineering is a term that popped up, and then we were playing inside that.
Again, with the intense knowledge that I had in AI and neural nets, and now it was turning into generative AI in 2022, I knew and started to see the vision even more. And then all it took was ChatGPT to roll out an actual product that could be mass consumed because it was a chatbot, which is a fantastic way to interact with generative AI. And so as we’re leading into the moment that ChatGPT was released and then ChatGPT got to have way more conversations about AI than I’d ever had before. And so it’s started sparking even newer ideas. And that’s when I realised, “okay, generative AI isn’t just a chatbot tool, generative AI is going to blow up into so much more”. We had text generation, video generation, audio generation, music generation now for Google. These are sort of the front lines of AI and they’re blooming in 2023. And it’s incredible to watch.
Ben (04:03.703)
Yeah, I mean, I was saying offline that to me, it seems like it’s just all of a sudden landed. But for you, who has had an interest in this space for a while, it’s obviously been something that’s been developing, been growing and it’s been on the horizon. And here it is. But for anyone else that’s in my boat, who kind of all of a sudden, all we’re hearing about is this GPT. What does that actually mean? We had ChatGPT, then Einstein GPT, like what is GPT?
Mark Good (04:31.97)
GPT was born in, I think it was around 2016, there was a paper published called Attention is All You Need. And it was this attention mechanism that got infused into GPT, as they knew it at the time. And that attention mechanism sort of unleashed what I consider as the productivity of GPT. And so this model in 2016 sort of hit the scene, it was really quiet, it didn’t really get picked up or iterated on for a while, 2018, 2019.
It just fundamentally stands for generative pre-trained transformer. Generative is you’re generating text and pre-trained’s on your big data. And then, so you train the model. It creates these weights inside of a neural net. And then the transformer just means that it’s taking an input, the transformer does its job, and it sends out an output. Or that’s how we’ll keep it simple for this call. So GPT, Generative Pre-trained Transformer. And the interesting thing is a Generative Pre-trained Transformer can transform so many different things. So it takes the data in and it spits data out and can convert data from one form to another. And it’s just really this awesome superpower that if well-trained and if well-managed can do really amazing things.
Ben (06:12.539)
So when someone’s engaging with, let’s say, ChatGPT, or I’ve not seen Einstein GPT yet, so I don’t know exactly how that works, but if someone’s engaging with it and they’ve not kind of, if it was the first time they’ve ever seen it and they’ve not heard about it, which I don’t know where they’d have been living if they’ve not heard about chat GPT, but people would often think they’re just dealing with like a bot, right? Because it kind of has that kind of feel that, whereas a bot, just to clarify for listeners, a bot is that pre-programmed to give a certain response to a certain question.
Whereas ChatGPT is searching this big data for the answer, the right answer and not programmed, is that right?
Mark Good (06:51.222)
Kinda, yeah. Yeah, the chat bot that we typically know is the one that we get annoyed with when we’re working with customer service. It’s the pre-canned answers. This is a totally new level. This is more, and some people don’t like this analogy, but I think it’s something that we all relate to. It’s like if you go to your favorite search engine and you start typing and then it sends out, it gives you that recommendation, like based on the next word you put in, the most likely next thing that it would recommend to you. So in a very similar fashion to that, but definitely not exactly.
It’s looking at whatever your input is and it’s saying what is the most likely next output for that input that a human just dropped into this, you know, transformer and then it spits out what it thinks is the highest probable output based on the input, based on all of its training data, all of the information that was fed into it from everywhere, like billions of parameters. And so it almost has, it seemingly has almost like a super intelligence, although oftentimes very incorrect.
Ben (07:53.719)
Yeah, yeah, I guess that’s the, it’s not always right, right? That’s the thing we need to, at this stage, it’s not always right. There’s always gonna be some discrepancies with what it provides.
Mark Good (08:04.798)
Yeah, exactly. And we call them hallucinations, which some people don’t like these terms, but I still use them, I like them. Hallucinations is like, it’s gonna tell you and sometimes very confidently that something is the case when it is just not the case. And you’ll find this, if you start to get really nuanced and talk about Salesforce specific solutions and you’re asking questions, looking for specific answers, hey, where do I find this in the setup menu? It’s gonna get it wrong. Oftentimes it’s gonna get it wrong. And that’s the big challenge. Okay, so GPT is an amazing technology, an amazing tool, but the challenge is, you just show up to it and you show up to the chat bot and you’re looking for really nuanced solutions to help you do things, you’re going to get hit with incorrect information. And so there’s this moment of hesitation, like, “well, this thing isn’t really what I thought it was”, or it’s really hard to use. So I’m not going to bother with it, but I just implore you that yes, it is really great technology. And the challenge is you have to get over the hump of learning exactly how to engage with it, how to think about it, how to frame it. And, once you get over that hump, it changes your life.
It changes how you view tasks and you approach challenges and projects. It’s a game changer.
Ben (09:12.239)
Do you like, personally, do you have any data around how much time it saved you on specific tasks or anything like that over the last kind of six, 12 months?
Mark Good (09:22.634)
No, no, I don’t collect any data. Although that’s a great idea. I probably should do something like that and then post the metrics at some point. I have started to see, and we’ll get into this a little bit, I’ve started to see apps. So these are in-org apps that you could bolt on to Salesforce today that are starting to do that. So the idea is they wanna sell their products, so they want to show you how much time you can actually save doing your job in the org as an end user, as an end user using the org in this case. So that’s coming, but no, I don’t have any data. I wish I did.
Ben (09:51.951)
And why are they so different? Because I saw, like Google announced this, I can’t remember what it’s called, but it begins with a B, I think, their solution.
Mark Good (09:59.682)
Google’s Bard, yeah, that’s what they call their solution, Google Bard.
Ben (10:04.163)
So I played around with that and I wasn’t that impressed. I felt like actually ChatGPT was a lot better but then ChatGPT is only up until the data that they’ve got in their lake is only up to 2021, right? Whereas Google saying it’s live now. Like the example I gave to Google Bard was to give me a list of companies in Australia, FinTech companies with under a certain amount of headcount that was similar to company X.
And they were giving me like global companies that had huge head count. And then when I put that into ChatGPT, I was getting much better answers, even though the data is not so, so relevant or so up to date.
Mark Good (10:41.214)
Yeah. Isn’t that interesting? And so the secret sauce is how you are, what data you have that you’re using to train the model. Uh, you’ve got ChatGPT that was trained by OpenAI and Bard, which is trained by Google and how you train, you know, the data you use to train the model and then the method that you use to train the model and then the sort of the human, the feedback that you perform after you get the model trained all feeds into what I consider the sort of the secret sauce of whoever owns the language model. And Google has, they’re just not hitting them relative to OpenAI right now. And so ChatGPT is the preferred tool thatI talk about all the time. If, and, but if you are doing more searches, so like if you’re not, if you’re not, let’s see, I don’t know how to say it, but if you’re thinking about searching for things like searching the internet for things, go use Bard, why not? But if you’re looking for trying to use GPT as a tool, the solution for things, I would say go to ChatGPT just in general, it’s a much better product overall, it’s much more user friendly. It’s much more accurate than Bard. And especially in the Salesforce ecosystem, ChatGPT is preferred over Bard right now, at least.
Ben (11:47.887)
So your two passions have just, they’ve come together and everyone’s talking about in our world, but right now as of today, how have you seen Salesforce professionals making use of these tools?
Mark Good (12:03.946)
Yeah, that’s a good question.
This is interesting because we’re all looking at Einstein GPT and sort of like awaiting for the features to arrive. And that’s supposed to be sort of the promised land where Salesforce professionals will go to utilize GPT inside Salesforce. I see GPT as just life-changing in general. So I’ve seen professionals use it for, um, preparing for interviews. So using GPT and we can get into details on that if you want on how to prompt it to get it into a mode where you’re actually, it’s actually operating as an interviewer for a specific position that you might want working for a certain company. And these professionals are talking with GPT, sometimes even using voice to text so that you can get that vocal as well. And practice interviewing with this, you know, 24/7 fake assistant. They’re doing things like updating LinkedIn with, you know, information and keywords that are geared towards positions they’re looking for. They’re doing things like updating very quickly, updating resumes and cover letters inside of Salesforce, they’re solutioning with it. So they’re brainstorming solutions. They’re creating training documents for solutions that they already have. They are solutioning for really cool things like we train on this kind of stuff, like validation rules, formulas, which can kind of get complex at times. Writing Apex, you’re using Apex, writing Apex, writing test classes, writing triggers. I see posts that are popping up often now where people who don’t have as much experience are writing Lightning Web Components. And the list just kind of goes on and on and on for in-org use. Or it’s technically that’s not in-org use. That’s out of org use, but for solutions for Salesforce. And then there are people that use it to help them schedule out priorities. Like they’ll copy and paste sort of a list, maybe their Excel list, their sort of upcoming tasks and whatnot, maybe with potential due dates. And they’ll drop it in there and ask for sort of a list of sub tasks that go with it that they can tackle and dates to go with that. And it spits out really great information. Just the list goes on and on. Actually, we’re recording a series called GPT Trailblazers, which is just collecting these stories because I think that most people still don’t have a concept that all these different use cases are available out of the GPT tool that we have today. I think a lot of Salesforce professionals are still waiting for Einstein GPT. And I just advocate, don’t wait, you know, get into what we have today. Einstein GPT will come and I’ll tell you more about some of the other tools, but you’re missing out if you’re not taking advantage of it today.
Ben (14:43.767)
Yeah, that makes a lot of sense. Now, just on your first point around interviews, that’s an interesting obviously being a recruiter. That’s an area that I’m interested in as well. Like I think a lot of people might go on to ChatGPT and say like, “ask me questions for a Salesforce Admin interview”, right, but I’m going to have a crack here at saying this is what I would prompt if and you can tell me if it’s kind of the right kind of thing. So my understanding of it is you need to
the input you give is going to really directly impact the output you get, right? So would you go as far as like researching who the person that’s interviewing you is and saying, “and play the role of a Salesforce hiring manager that comes from an IT operations background, who is looking for a Salesforce Administrator with four years of experience and with experience with Flows, you know, day to day Salesforce Admin tasks and comfortable engaging with senior stakeholders”, something like that. And then that’s obviously going to give a lot more detailed questioning than someone that just says, “ask me Salesforce Admin questions”.
Mark Good (15:43.798)
Yep, absolutely. Exactly. You hit the nail on the head and, and this is a great point. I’ll drive it home here is that what you just came up with is based off of experience that you currently have with ChatGPT and you’re spot on, the more and more experience you get trying to tackle a task like that, the more and more you realize how nuanced it is how much you can actually improve the output for what you’re looking for based on your amount, your personal amount of experience using the tool. And so this is why it’s so important to start like yesterday. And if you haven’t started yesterday, start today.
And even if you start today, you know, try to use it as much as possible. So going back to your prompt, this is a great tip for anybody who’s watching here, listening. If you want to get great interviews, you can do a prompt. The first thing you want to do is, you want to set the tone for GPT. You’re going to, you’re going to tell it who it is. So you are a, and then insert, you know, whatever it is, Salesforce hiring manager, professional with whatever title and the more information you give it, the better during when we’re using it for these creative type things. Give it its tone. Tell it who it is and then create some lines and then give it context. So now it knows who it is in the prompt. You can give it and feed it information you know about the person who’s going to interview you right. You tell them maybe even give them like oh this person likes to play golf or whatever the thing is and it actually it matters and then next thing you do is create context so tell it who you are even paste a copy of your resume potentially watch out because it’s a you know that information could end up in the model in the future so just do your research on that first before you dive in. And you give it context, as much context as possible, who you are, what your skills are, the type of job you’re looking for, what your strengths are, your weaknesses are, and then, and then, so you gear it up. You give it its voice, you give it context, and then you can say something like this at the very end. The sort of these instructions to GPT. These are like the guide rails that you put it into or you give it. And it’s like, “okay, now ask me, you know, interview questions, act as if we’ve been interviewing for five minutes already”, and even you can throw in another thing, which is “after each interview question, offer me a quick tip on how to improve my answer”, because as you answer, it’s going to use that answer for the next prompt and then it’s going to evaluate that as an expert. If you tell if it’s an expert interviewer, then it will give you an expert response for how well you did on responding to that. So now you’re learning as you’re going through the interview process and not just practicing an interview. So yeah, spot on. And there’s some there and that’s a really sort of, in my opinion, general way to use it. There’s so many more nuanced ways that you’ll find out that you can use it, things that you can use it for as you’re going through and want it for.
Ben (18:39.047)
Yeah, and that’s why training is so important, right? Because I think a lot of people aren’t using it to its full potential, because they are using it like a search engine and just putting in one line, and then seeing what comes back. So you’ve told us how people are using it now, where do you see it going within the Salesforce world? Like where, what’s the, I’m not gonna say end case, because who knows, right? The world is evolving so quickly, but the world of technology. But like in the next kind of six, 12 months, where do you see AI, GPT, and tools being used for good within the Salesforce ecosystem?
Mark Good (19:13.066)
I think the next six months to 12 months are going to be riddled with Salesforce delivering on various GPT. So Sales GPT, Marketing GPT, Einstein GPT, whatever that might be. They’re rolling out features. They put out a roadmap on AI day and that roadmap was full of different features that they’re going to have in the org. And so I think if you’re a Salesforce professional looking to sort of be on the what to do with all of this. So these in-org GPT tools are coming. They’re coming from two different sides. One is from the Salesforce side where they’re going to deliver on their promises on these random, these various GPT clouding tools. AI Cloud, I think was the name that they’re going to bundle it all under. And then you’ve got the sort of the Salesforce partner side where you’ve got these apps that currently exist and there are vendors that I’ve been working with that are already
So Salesforce, these ISVs with their apps, these are the things that are going to start making their way into the org. And that’s a big deal. And I think many people don’t get this yet. I think, oh, and we have GPT Dreamin’, which I hope you attend coming up in early August. That’s going to sort of like, unravel all of this. So this is sort of a sneak preview for anyone who’s joining us here. So the next six months are going to be riddled with all of these GPT infused tools hitting Salesforce in the org. So our job as Salesforce professionals is to know the tools, know their capability, know what their risks are from a security standpoint, going through the trailhead that exists from Salesforce right now on AI, learning about the AI, learning about the neural nets, getting all as get as much information as you can. You want to be on the bleeding edge of this. Trust me. Get trained. It’s worth the money to get trained today. Do not wait. Go ahead and get trained. Get all the free training you can. If there’s any paid training, take a look at it, evaluate it, see if it’s worth it to you. These GPT tools are going to change the game. It changes the way that we, it changes the way that we administer Salesforce, and it’s going to change the way that end users use Salesforce. And every company is going to want it, in some way or another. There are so many use cases for automating things things up inside of every company’s orgs using GPT tools and if you are a professional who’s got a GPT and we call them out of AI Force training GPT specialists if you have a GPT specialist title or some similar background or capability companies are gonna compensate you for that and you’re going to want to be on the front end of the ability to implement that. There’s a lot of security concerns, but it’s solvable, and we already have a path for that.
Ben (22:15.355)
So who, and I’m sure there are going to be people that are scared by this because you know, like you said, at the moment, it’s not always going to give you the right answer, but at some things are always improving, you know, the data lake is going to be more up to date at some point with ChatGPT, you know, like whatever Salesforce have planned from their solution, like is going to be able to solve problems in or like you mentioned. So should certain skill sets like are we seeing the end of the admin role as an example? A Developer is going to be less required because Admins can write code because they can put the requirement into ChatGPT and just copy and paste the code that it writes. What’s the look ahead for these skill sets and more, I guess?
Mark Good (23:05.194)
Yeah, it’s really hard to say. It’s really, really hard to say. I think in the short term, no. In the short term, you still continue down your typical training path that you go through to get certifications, and GPT, you’ll see it as a bolt-on, sort of. It’s an extra. You learn that, and that’s an extra, and it just sort of accentuates all the knowledge that you currently have.
That’s the short term. That’s how we’re guiding everybody from a professional standpoint. We even had a picture out one time on LinkedIn and it was like 3X your cert value. So if you have, and this is because GPT does give you wrong answers. So you still have to have some base knowledge in a core area in order to evaluate the answers and either know if you can use that code or you can’t use that code or how long it’s gonna take you to iterate over it to make it usable code. But even in saying that, there’s definitely something happening. There’s, I feel a shift occurring and I think a lot of people aren’t going to feel this for a while, but if you watch this back later, you’ll know a shift is occurring where professionals are going to start to see and start to question the actual value of a cert. There’s a, there’s a specific levelling of the field that’s going on because of what you mentioned, because if I’m not a Developer, but now I can all of a sudden decode code and faster than any developer could have decoded code before using a GPT tool. I think that those implications are gonna resonate and there’s going to be this massive ripple throughout the market and it’s going to flip the structure of how we do what we do from a learning standpoint for Salesforce and administering Salesforce.
And then on top of that, it’s like, and I don’t think Salesforce is ready for this yet, but the GPT infusion inside of Salesforce is going to change what they’re going to have to teach from like a Trailhead standpoint, a search standpoint, and as end users get comfortable with GPT and we get familiar with the safety and comfortable with the safety of it’s hard to say. It really changes the game across the board. And I know that sounds crazy and it’s probably going to be a lot of comments about how “like that’s a little bit, that’s a little bit much Mark”, but honestly, just get ready. It’s coming. Things are going to get really weird, especially later this year, early next year.
Ben (25:21.487)
But you’re obviously here to tell me that there’ll always be a need for Salesforce recruiters, right?
Mark Good (25:26.61)
Yes, yes, yes. Yes, every industry is going to be impacted by GPT. So if nothing else, as a Salesforce recruiter, you will adopt the technology, you will learn how to automate your own processes with it, and you will just become that much more productive as a recruiter in general with GPT.
Ben (25:44.475)
Absolutely, absolutely. And by the time this is listened to by our audience, a lot of people will have seen the post you put out just at this weekend, the time of recording, where you actually have passed a Salesforce certification, not the actual cert, but one of the practice ones on one of the practice websites with the use of ChatGPT. And I think the past, like you got 80 odd percent. So that does go to show there were 20% of questions that they got wrong, right? But that in itself is, you mentioned potentially people are going to question the value of certifications if people can go on to ChatGPT and answer these questions without already having the cert themselves. Like, you know, that it’s going to be an interesting few months as things like that kind of unravel.
Mark Good (26:30.43)
Yeah, and please don’t get me wrong. And I hope I don’t get fired from the community for this. Please go get your certs. The value of the cert is still there. It has not gone away. It will not go away for a long time. But just know that there is a shakeup that’s happening with GPT being available to everybody and that you wanna be using GPT. You want to be someone who is using it. So if you’re, say you’re a Developer, you’re not using GPT and you don’t for two straight years from starting today, you will have Admins pass you, I mean Admins today will pass you even if they aren’t certified a Developer, yeah, we put out that post it was for the first time ever and we recorded the whole thing so there’s nobody can question it. It’s all online it’s all available on our AI Force YouTube channel or a Force Training YouTube channel so you can check it out, but a GPT tool passed the Admin practice exam and then within an hour of posting that someone on LinkedIn got inspired and went to ChatGPT and GPT4, they had the API, so they were able to do this where it wasn’t readily available for everybody else. And they went in and they didn’t record it, so we’ll have to test it later to see, but they passed the PD1 cert with GPT4. And so, you know, so imagine we go months, you know, GPT shows up to the scene for everybody. Then we go six months and, you know, one day I post that GPT passes the Admin exam. And then within an hour, someone posts the PD1 cert.
I mean, you have a Salesforce trained professional tool 24/7 available. I mean, think of the implications of that.
Ben (28:12.1)
So we just saw Salesforce AI Day, which I’m sure you’d have been tuned into and one of the first online. What are the key takeaways or what were the key messages that for anyone that didn’t attend may have missed?
Mark Good (28:29.334)
Yeah, there were a couple things that really stuck out from AI Day. One is Marc Benioff is on board. I mean, he gets it. There’s a lot of language. I did a reaction video. Some, if I haven’t posted it by the time this goes out, I will, but he was using a lot of language that showed me that he gets it. He absolutely understands why this is such a big deal. And so I assume that he’s driving the team to do amazing things behind the scenes that we’ll find out about later on.
There were a lot of great speakers that day and they just really laid it out. I thought it was going to be GPT Dreamin’ that really delivered that message, but Salesforce on AI Day did it. And GPT Dreamin’ will just double down and just expand even farther on that. What they said was two things, two major things that stuck out three. Okay. They announced a couple more GPT tools for the clouds that they were allowed. Okay. We knew that was coming. They talked about this trust layer, which was they know all the clients are going to be concerned about trust, any personally identifiable information or intellectual property from the company that you send to the tool is going to go through this language model. And the question is, where does that data go? Where is it stored? What about the prompts? Where’s all that data? And can we trust that it’s still secure if it’s left Salesforce in some way? So Salesforce is rolling out the trust layer. So we need to become intimately familiar with that trust layer, how it works, so we can bring back to our own organisations and convince them that this is going to be secure enough for us to implement and start to bring them some real robust value out of GPT. So that trust layer is going to be key. Secondly, is they announced AI, or they talked a lot about Trailblazers. So they get it that they’re going to have to train up a whole fleet of people, of professionals who know what AI is, how to use it, how to deploy it responsibly. So they launched a Trailhead and a badge for that. So if you haven’t checked that out, go find it. It’s going to be really important. Of course AI Force Training’s got continued learning for that, so check that out. Go get trained. Okay, so that was their message, go get trained. That’s what I read loud and clear. Trust layer, go get trained. And then they talked about GPT automation. So I, and they use some terms that I had come up with in a small team of mine have come up with on our own. So I know that they’re on the right track for understanding how GPT is going to fundamentally automate. And I use the term autonomous bots. I don’t know if anybody else is comfortable using that yet, but there will be autonomous bots inside the org. We, as professionals will be managing these bots and these bots will do things for us as a human typically would have in the past. So that’s my sort of take in the future. Salesforce will come out with that eventually, but for now they’re using the language that leads to that step in the future for automation. So that’s a big deal. AI Day also, what else did they talk about? It was all in org talk, which is really good. Oh, one of the big things they’re allowing you to, they’re setting up the system, they’re setting up the org behind the scenes to allow you to use any language model you want. So you talked about Bard earlier, right? There is OpenAI’s language model. You’ve got Google’s language model. You’ve got Meta’s language model, llama, which has been touted by OpenAI as one of the best language available. And so as GPT specialists, we’re going to have to look out and we’re going to have to know what the language models are. We have to know what the best ones are for specific use cases. And then Salesforce is allowing us to utilise whichever language model gives us the best result for what we’re looking for. So imagine you’ve got prompt engineering, you need to learn prompt engineering. First of all, we can teach you that, but learn prompt engineering. And then beyond prompt engineering, you have to know what the language models are and which prompts need to go with a Salesforce scientist managing these bots inside your org, functioning the business. Imagine this. So people are going to think I’m crazy. Seriously. Check back a year to two years from now. This is not sci-fi. This is reality.
Ben (32:58.223)
So what are these bots gonna be doing then? What do you think a bot would be capable of within an org?
Mark Good (33:04.714)
Yeah, I don’t know yet. We’re doing our own testing. So we’re not at the bot level yet, but we’re at that sort of that prompt engineering, which model does it go to? How do we make sure it’s safe? We’re fleshing that out in the training. The next step beyond that is the bots. I can say that my first initial concepts are, if you’re looking at a business’s process and you think about as a human when I show up, okay, say you get a case in from a customer. Let’s say it’s an email to case. So you get an email from a customer and the email comes in and it gets created as a case and you know, you assign a couple of different things, maybe you have an automation that does some things and you sort of queue it up for. You sort of queue it up for a human, right?
So you queue up this case for a human to come in and what does the human do? They read through it, they think about the customer, they research the customer a little bit, they research the problem, they maybe look at a knowledge article, maybe recommend the knowledge article in a response, this sort of activity. That activity is prime use case for automation by a GPT bot. So GPT reads text, GPT can do research, GPT currently in its current state today. I’m not talking about the future, this is today. GPT can reference knowledge articles that are stored in a specific location, read those knowledge articles, read the customer’s problem, and then create in a very human like way the response to that customer’s problem and then send it as a very as a response to the customer that actually taps into the customer’s personal information so meaning like okay maybe we have some information on this customer on like certain topics whatever I don’t know how to explain all this but in the best way, but the response cares the most about that customer and really connects with the customer in a way that the customer is not typically connected with not used to being connected with before. Okay that’s a lot. That’s a great use case for one of these bots so imagine a web to case comes in and then within seconds the customer has a response or maybe you auto time it for like five minutes later because you don’t want it to look like a bot. And the response goes out in five minutes and you solve the customer’s problem. So where a customer would think typically maybe, maybe like 24 to 48 hours, get a response for a problem. It’s almost instant or it can be instant.
Ben (35:35.991)
Yeah, that’s interesting you talk about like, like kind of empathy, right? Because being a recruiter, we’ve always been told that at some point, bots, robots are going to take over our job. And the response to that has always been well, a bot can’t show empathy, right? Like you’re dealing with people’s careers, like, you know, there’s lots of nuances and emotions that go through that job search experience that can’t be managed by bots. So yeah, who knows, right?
Mark Good (36:00.598)
Well, like I can tell you there was a study done and it was related to doctors and bedside manner. And GPT was used and the patients didn’t know or weren’t didn’t know if the response was from a doctor or from an AI, and I’ll just cut to the chase. The result is that the customers based on the prompts that were used for that AI at the time preferred the bedside manner and the professionalism of the bot over the doctor, the actual doctor.
Ben (36:32.635)
Scary.
Mark Good (36:33.874)
So think about that now as a customer, like that’s perfectly transferable, right? But again, it goes back to the professional as a Salesforce professional, because no one else is going to know how to do this. The GPT specialists are going to have to go in and create those prompts and know how to withdraw or draw out that hospitality out of GPT.
Ben (36:38.838)
Yeah, 100%.
So you’ve started a business, obviously, AI Force. I’ve been seeing a lot about it. I know people are getting certified by you, which is awesome to see. But how big a deal was that for you personally to set up your own business and go into the world of running a business? And then also, what’s the vision? What are you looking to achieve?
Mark Good (37:30.73)
Yeah, yeah, it was a big deal. Uh, it was a really big deal. It’s really challenging and it is today too. And I believe it will be for the next year to two years to three years, maybe even depending on how AI shakes out. Uh, because it is new, there’s no textbook written on how to do this thing. It’s almost like content that I published today could be outdated a month from now. And so it’s been really challenging from a business perspective to be in the training space on AI, cause it’s, you know, it’s one thing to train but the real question is how do you apply that to your job today? And how you apply it is changing very rapidly. So it’s very hard to be in a business with such a rapid changing environment. We’re tackling it the best we can. I think we’re crushing it. I think every certified GPT specialist that comes out of the program is stoked. We’re really making a massive impact and these are the professionals of the future. So I’m really happy about where we’re going to be in the next six months to a year. What’s the vision? Yeah, our vision is really just to empower Salesforce professionals from the ground up with whatever the best possible GPT tools are available from Salesforce or from third party vendors or things that we can come up with on our own at AI Force. And we’re empowering as many Salesforce professionals as we can with those tools. And just trying to inspire anyone or everyone we can to get in the game on GPT.
And then we’re also trying to forge as many partnerships we can with sort of the best producers of GPT tools. I think AI Force and AI Force Training are going to be seen as sort of the premium and preferred and most knowledgeable and the biggest army of Salesforce professionals who are sort of on the front lines of understanding and deploying GPT inside of the Salesforce org. And I think Salesforce set the scene with AI Day and said like, that is going to be the focus. That is absolutely the focus of Salesforce. Marc Benioff even said that this generative AI is the thing that unlocks the next 10 years, next decade of revenue increase for Salesforce. And so to me, that just means money’s going to be poured into research and development. We should be ready for the tools. And I’m trying to help close that gap and get us there.
Ben (39:52.283)
So if anyone’s listening to this and now they’ve heard you for the first time and they now have this itch to scratch where they want to find out more, obviously they’re now going to follow you and follow you on LinkedIn because I know that you post some really great content there, but what other resources or even people are out there that are spreading some good messaging and some useful and insightful content around this world that we’re now in?
Mark Good (40:18.038)
Yeah, I’ll say, yeah, first and foremost, follow me, I’ll take you to the promised land. That’s my goal is just to take you there. And so I’ll post whatever I need to, to empower everyone and just get everyone engaged and just knowledgeable about AI inside of Salesforce. There’s, if you really want to dive deep, there’s a gentleman named Andrej Karpathy. He actually works at OpenAI. He worked at Tesla prior to OpenAI. He is responsible for training the neural nets
vehicles autonomously around today. He started OpenAI, he’s back at OpenAI now working on the language model at OpenAI. So he’s got some really nuanced and really important information out there on YouTube. So look him up, Andrej Karpathy. Again, this is if you want to go to my level. So I’m hoping to inspire here. Riley Goodside, he is like a prompt engineering expert over at Anthropic working on Claude. So prompt engineering is going to be key. So I’m learning as much with experts on the topic and you will want to learn prompt engineering as well. Rachel Woods is an ex-Meta she worked on the LLamMa language model with Meta, formerly Facebook I guess. So Meta, yeah so Rachel Woods publishes some really great content you can learn about all kinds of ancillary things there’s ChatGPT but then there’s all kinds of things going on with different products and it does a great job of promoting those items and showing them to the world. And then, yeah, of course, from a Salesforce career perspective, follow me. I’m trying to gather as much as I can from all of these really intelligent people, reading white papers and boiling it all down to like just actionable information for us inside Salesforce, because that’s where we live. And that’s where we want to put our time and get the value for ourselves and for our clients and customers.
Ben (42:15.867)
Amazing, well I’m going to go away and try and become a prompt engineer myself now just in case the recruitment role does die a death. But look, thank you so much, it’s been really awesome, I’ve learnt a lot and yeah, I’ve got lots of questions to go away and explore now as well which I really think our audience will as well. So yeah, like you said, make sure people are following you, LinkedIn is the best place to reach out as well if people have got any questions.
Mark Good (42:38.938)
Oh, for sure. Please reach out. Like I’m not inundated yet with too many messages, so I can still reply to just about everybody that hits my inbox. So feel free to message me. Also, we’ve got a free intro to GPT from AI Force Training at www.aiforcetraining.com. You can also reach me there as well. And we have some really cool stuff being rolled out in the next couple of weeks, so you’ll want to be paying attention to see what we have.
Ben (43:02.591)
Excellent, we’re looking forward to seeing what’s coming through and yeah, really exciting time.
Mark Good (43:07.094)
Yeah, agreed. All right. Thanks for having me. Yep.
Ben (43:09.083)
Thank you very much.
Make sure you’re following Mark on LinkedIn and feel free to reach out to him with any questions regarding the topics covered in the podcast episode. You can check out AI Force training here:
You can visit our Salesforce jobs page for up to date opportunities. If you’d like to become involved in Talent Hub TV or the Talent Hub Talk podcast as a guest, we’d love to hear from you.