Ep 37: Using ChatGPT Across Different Functions of a Company

Jordan [00:00:17]:

Can you use ChatGPT across all areas of a company? That's one of the things that we're going to be talking about today on Everyday AI. This is your Daily Livestream podcast and daily newsletter. That's right, go to your everydayai.com to sign up for that newsletter, but this is yours. We talk about everything that's happening in the world of AI and we bring on experts in their field to talk about how they or them or their companies are using AI.

So before we bring on our guests, let's quickly talk about what's happening in the world of AI news. There's always a lot going on. It's hard to keep up. You could spend hours a day. Instead, just spend 15 minutes with us every morning. So, as a reminder, if you're listening live, please drop a comment, especially when I bring our guests on for the day. But let's go ahead and run down what's happening because we have some big stories that are going on in the world of AI.

EU Parliament passes first Western AI Law

 So first, and this is hot off the presses, this has only been out for, I think, a couple of minutes, but the EU Parliament has moved ahead with a monumental AI Act. So it's called the European AI Act and it is the first law in the Western world. So essentially what's happening is they are saying these foundation models, such as ChatGPT, need to include provisions to ensure that their training data does not violate copyright law. So, again, this is just kind of fresh and new and there's been a lot of kind of back and forth between OpenAI, the ChatGPT parent company and just different EU nations in terms of regulations. So this is going to be very, I'm not going to say exciting because it's depending on your thoughts on ChatGPT. But regardless, there's going to be a lot of activity on this in the coming months. And also what this means for the US. Some more on that in the newsletter.

Salesforce's new AI cloud software

Salesforce, they've really been going all in on AI lately, so they just recently announced a comprehensive suite of just AI infused software. They're calling it, I believe the software, sorry, salesforce AI cloud. And they have a starter pack. This does seem to be geared more toward enterprise companies, given its reported price tag of about $360,000 annually. So if you're a small business or an entrepreneur, probably not for you. But if you're an enterprise company that's using salesforce in every part of your business, that's something to keep an eye out on.

Meta releasing its own AI image creation model

Last but not least, the third of our news stories, and we'll have more in the newsletter. Meta is releasing its own AI image creation model, so there's a lot of popular ones already on the market from OpenAI, you have Doll-E, private company called Midjourney who I think is leading on the pack Meta has released. So this is the Facebook parent company. They've released their own version of an AI. Image creation model called I Jeppa, and it apparently is doing kind of this AI. Image creation a little different than the other models, and it's using what it says knowledge of the world to complete images rather than kind of the similar technology that the other companies use, which is looking at nearby pixels to kind of, like, autocomplete an image. So they haven't released a ton of samples. We'll go ahead and link their research paper in the newsletter that shows kind of some basic methodologies, but it should be exciting to see what comes next and how Meta can really push what's going on in that space. 

So those are the big pieces of news. Make sure to go to your everyday Ai.com, sign up for the newsletter, read it because we're going to have a lot more about these and other stories. But let's bring in our guest for today. Very excited. So we are welcome to the show. We have Erik Bowman. He is the VP of technology and security at Improving. Erik, thank you so much for joining us.

Erik [00:04:20]:

Yeah, absolutely, thanks for having me.

Jordan [00:04:22]:

We have a great conversation to get to. We're going to talk about ChatGPT, not just ChatGPT, but also how companies like Improving are using It and even talk a little bit about Microsoft Copilot. So as a reminder, we have a couple of comments for people who are tuning in. So if you do have a question for Erik, please go ahead and drop that comment. So a couple of people are already commenting and excited to hear what we're talking about. So Erik, just give everyone a quick overview kind of of Improving, what you guys do and kind of what your role is at Improving.

What Erik and Improving are doing

Erik [00:05:00]:

Yeah, absolutely. So improving is an it consulting company. We do a lot of different It consulting, software development for mid market and large clients and some small as well. We actually have offices in Canada, US and Mexico now around 1500 consultants and growing. So pretty big consulting team and a lot of different technical services. And so one of the things that is super important to us as we are both advising our clients on technology but also using technology ourselves is what AI is going to be doing for us. An internal business capability, but also as a consultant area as well. And so I do want to quickly just kind of disclaim that as we're talking today, this is not Improving internal official stance on that.

That's something that we are just like the rest of the world learning to figure out that working with our in house counsel, I'm thinking on the security side. And then we have another gentleman who's helping us from the consulting practices side. And so just kind of that we are thinking about this the same way a lot of other companies are. So as we talk today, it's more how do we approach that problem, what issues are we seeing and how are we helping our clients as well think through this? So just wanted to get that out there as well.

Jordan [00:06:09]:

Yeah, and that's a great point. So let's even talk about this because I'm sure you have people internally who are looking at you and you're obviously working with other decision makers as well. So whether you want to talk about it through the lens of improving and what you guys are doing, or just businesses of your size in general. But how do you think companies should be handling the use of AI tools, specifically ChatGPT? Because there's great power but also great responsibility. So what's your take on just how you can moderate it? I guess respectfully use it in an office setting.

Using ChatGPT and CoPilot in a business

Erik [00:06:48]:

Yeah, no, exactly. And I do think that's something that there's so many aspects to it, right? Because it is a super powerful and capable product today. And of course, some of the things that we're talking about too, like Copilot is coming soon and so not using it really is not an option. Right. You are definitely going to cripple yourself competitively if you decide we're just not going to use it. And so, as I think about it, I'm a technologist, but I also have that legal and compliance background as well. So I have almost competing minds as I think about things like this.

You talked about the copy of your law, right, coming from the EU. That's also something that already existed as a concern. We think about recent cases where software developers put code, proprietary copyrighted code in ChatGPT, didn't realize the extent would train on that and then it would later spit that out and now they violated US copyright law, probably we would assume, but also enable ChatGPT to do that. And so as a consulting company, or as just any company that has your own intellectual property yeah. Are you exposing it? Are you losing it because you are using a product like ChatGPT? But also, how do you use it and get to where you need to be competitively? Because again, like I said, not using it, it's not an option.

Jordan [00:08:09]:

Yeah. And I'd say when companies are hiring improving, whether it's in It consulting or software development, I'm sure on the client side there's maybe different expectations or maybe sometimes unspoken expectations in terms of your use of ChatGPT or like we talked about Microsoft Copilot or other AI tools. How do you handle that kind of balancing act between, like, yes, even if a client asks for something, how do you handle that? Because it is kind of a gray area, at least here in the US. For now.

Erik [00:08:47]:

Yeah, well, first and foremost, we'd all refer to our clients own decisions on that. So if they, for example, have said, hey, you know what, we are using Copilot, we are using ChatGPT, we would say that's great, give us the contours of how you want us to use it. Do you already have policies? Do you already have accounts and services, all those sort of things? Do you have your own controls in place with them to protect your intellectual property? If you do, awesome. Problem solved. Right? Because they have made those business decisions, they've put the relationships in place, they put the safeguards in place, and we kind of know the contours of that. That's the minority.

Jordan [00:09:24]:

Yeah, that's a good point. Explain a little bit. And this is something I'm very curious about, Erik. At a larger company such as Improving, it's always evolving, right? Like even what ChatGPT means or can do four months ago is much different than now if you have access to the Internet and the ability to bring in third party plugins. But what are some of the more, I guess, meaningful or impactful ways that you guys have used it so far? At least internally? Because I know with clients it's obviously on a case by case basis, but what are some ways that you've used it internally that you think kind of the everyday person can learn from that?

ChatGPT - a helpful teammate for input

Erik [00:10:11]:

Yeah, I think what's amazing about it is that it can really reach into any vertical of your company. Right. So we talked about software development. That's kind of one of its oldest use cases. But even as we think about compliance and policies, for example, it's a great tool to get you started. And I guess that's the other thing. My other caveat is I know that it makes mistakes and so as I'm using it, I'm always going to double check and make sure that what it's giving me back is valid. Think about a policy. Think about we'll just use privacy policy, for example. It needs certain things to be in it to be effective.

And so as a professional, I could do all the research, I could write that out and I could then produce a perfectly good privacy policy. Or I could also say, hey, ChatGPT, what are the good elements of a good privacy policy? And make sure that if it says point number seven is XYZ, oh, yeah, that's also something I should be including. Right, so it's almost like that teammate that's giving me input into the topic, I'm not relying on it, I'm not having it write the policy. I'd rather do that myself still, but it can absolutely give me that input. Likewise, if we're in marketing, what are the good elements of a marketing campaign? It's going to spit out something and there's a decent chance it'll remember something that you've forgotten. So thinking about it as that friend that sits next to you that has some good ideas is always an easy use case, no matter what department you're in.

If it compliance, marketing, salespeople who are working on a proposal and they're like, could I say this more effectively here's all these words that I've written about what we're going to do, paste that in again, assuming that you're not violating any sort of intellectual property and say, can you summarize us? Right? Can you give me the three bullets that summarize these 500 words so that I can be a more effective communicator?

Jordan [00:12:05]:

Yeah, I think you bring up a good point because for smaller businesses as an example, had a great entrepreneur on the show yesterday. And entrepreneurs are really, I think, using ChatGPT more for content creation or just AI. Programs in general for content creation, where it sounds like maybe a better use case for a company of your size is more of ideation research. Brainstorming, some of those things. So what I guess at least if we're talking just ChatGPT, what other barriers or kind of boxes to check still have to happen in your mind to be able to use it more at scale because you kind of talked about ChatGPT making things up? Hallucinations that's something that we're actually going to talk about later in the week on this show. But what steps do you think have to be taken on the ChatGPT side or features until you can use it much more frequently across all areas of kind of a more enterprise sized company?

Accuracy and trust in AI language tools

Erik [00:13:11]:

Yeah, I think it's going to have to be twofold. One, obviously the safeguards around my information so that I feel more comfortable sharing because it's better the more I share with it. Right? Everybody knows that if I don't tell it what I need it to produce, it won't produce what I want. And so I do have to share a decent amount of information with it. And I know OpenAI has the checkbox that says we won't use your data anymore, but also that has other features that turns up other features that are useful. So I don't think they struck that balance yet. But then the other side of that is how can we measure its accuracy with some confidence, right?

So if it's reporting back to me in a subject matter that I know, then I can be like, I think you're making stuff up. If I've asked it to talk about something that I don't really have that subject matter expertise in, it can make stuff up. And I'll be like, that makes sense to me. And so how do we put that layer of self evaluation almost of can it detect its own hallucinations I'm not even sure, but give you a confidence of like, hey, I'm highly confident that I'm not making stuff up or you know what, I probably made a bunch of stuff up. You should go fact check what I just bout.

But I use the example here of foreign language translation, right? So we're international canada, US, Mexico, we have three languages within our company. I received an email from one of our Spanish coworkers. I can read enough Spanish that I know what he's talking about, but I can't write Spanish in a way that wouldn't sound like Google Translate just for an experiment. Did it in Google Translate and then also did it in ChatGPT and had it output what I would have wanted to say in Spanish and I read it, I'm like, I don't think that's what a native speaker would have said. I think it's grammatically correct, just like Google Translate tends to be grammatically correct. The ChatGPT answer, I think was right, but it wasn't actually how someone would talk. And so if I just said, hey, translate my entire website to Spanish, I believe I would be creating a very embarrassing piece of marketing for Spanish speakers. Right. And so how do you take that expertise? Again, it's an area that I have enough knowledge that I can believe it and not enough knowledge that I can validate it. That's I think maybe the question that I want to know how to trust it. Right, sure.

Jordan [00:15:36]:

You bring up some fascinating points. Even asking ChatGPT, how confident are you in this response? Actually, a pretty frequent listener of the show, I'll give her credit. Mabritt says that she asked ChatGPT, give me a confidence score on this. Is it 0%, is it 100%? And she says she's getting varying degrees. So that's an interesting take to ask ChatGPT how confident it is. Daniel, thank you for the comment saying exactly, kind of echoing what Erik is saying, that you need to be checking the answer regardless of where you get it from, whether it's stack overflow or ChatGPT. That's a great point, Erik.

Let's quickly talk though Erik about Microsoft Copilot. So for the everyday person, if you're not super familiar, microsoft Copilot is going to be a it's getting slowly rolled out, so it is kind of live for a select few customers. But if you think of what ChatGPT is capable of and you have to log into a browser, think of that throughout your entire operating system experience on Windows. So regardless of what Windows program you're using or on the actual operating system, you have that AI capability there. Erik, what does this mean for improving, specifically when that is rolling out and also what does it mean for other companies and just the future of work?

Copilot: produces documentation & incident reports automatically

Erik [00:17:03]:

Yeah, I think this is going to be super exciting when it starts to produce the promised outcomes. As with all marketing, it sounds amazing, it's going to be wonderful. Reality will take some time. But consulting. For those of you who haven't worked with consultants before, not only do we produce the work you ask us for, but one of the things that we also have to produce is a decent amount of documentation around what we're doing. Status reports, summaries, kind of all the sort of things that show that we are doing what you asked us to do and in a way that you can report to your leadership. Copilot is one of those products that promises to help us do that.

So I think a great example I saw this is a slightly different tangent, but let's say you have an incident, you have a production outage, and from 07:00 P.m. To midnight, the team is just shooting emails back and forth. They're chatting in teams, all these things are happening and it's midnight, the problem solved, everybody just wants to go to bed. However, there's an executive summer who needs that kind of After Action Report. Well, Copilot promises to say, hey, you know what? I can read all the emails that were related, I can read all the teams messages that related, and I can produce that After Action Report and I can even do it in PowerPoint for you with graphs and pretty things. You can go to bed and have this executive report automatically sent. You can really just say, hey Copilot, create an After Action Report for this incident with these people and send it to Jordan and I'm going to bed again. I don't believe that's going to happen this year. That it's. That amazing.

But that kind of thing is going to change it because you think about all the documents that you're processing as a human, you think about all the emails that you're reading, all the meetings you're attending, all these a ton of information just flows past us and we just sort of pick and choose what matters to us. If we have that copilot sitting next to us and it's paying attention to all that, it can help us catch the things that we missed and create those summaries or even and I like the idea that if I have an Excel document full of data. I can say, hey, analyze this data, summarize it in a Word document and a PowerPoint and send it to my CEO. Right? That's a thing they're promising. I think that again, five years, ten years maybe that'll do exactly what we want. But just the fact that we can see it now is what I think is amazing.

AI improving productivity and output in companies

Jordan [00:19:23]:

Or at least to get a start, right? Because someone's ultimately going to be responsible to be turning those reports in. And instead of starting from scratch, maybe you have a document that's maybe 50% complete already, which is a game changer, I think just right there. So as a reminder, we are kind of getting toward the end of our conversation, but if you do have any last comments for Erik, please drop it in the chat. I will ask one Erik as someone, and we talked about using ChatGPT throughout just all different parts of the organization.

So combining that with Microsoft Copilot and kind of what you just previewed there, how do you see AI in general impacting. Just I won't say jobs because that's a whole nother discussion. But how do you see it impacting kind of productivity and output for larger companies? Do you think the expectation will be, all right now that we have all these tools that make us much more efficient, we should be doing 30, 40, 50% more business? Are those expectations going to come? Or do you think it'll be maybe just companies aren't going to be hiring new positions because of that increased output? I know it's a big that requires a lot of deep and thoughtful conversation, but I'd love to get your quick take on it.

Erik [00:20:53]:

Sure, yeah, definitely a quick take. I do think that there's a certain segment of jobs that are doing what I just described today that should be figuring out how they can leverage these products to continue to do their job. Right. Because otherwise the products will probably do their job for them. But the truth is most companies don't have all of those employees today anyway. I don't have somebody who can read all my emails and summarize them for an accident. I don't have that person. And so I think it's more to your point of improved productivity.

I become more effective doing what I do because those products exist and are able to take the information that I'm producing and distill it down in a way that then becomes more useful. In our case, we're not losing people because of it. We're actually getting more productivity out of it and in a way that doesn't impact jobs. It's not that, hey, Erik is now 30% more productive. We don't need a second Erik. They probably didn't need a second Erik anyway.

Jordan [00:21:55]:


Erik [00:21:56]:

But now I can actually do my job better. And so it is an overall net benefit to both me as an employee and improving as an employer as opposed to a risk. But no question that there are job segments that are threatened today by this because if that's all your job is, is to produce these after action reports, if you're at a company that's big enough to have those that I could see being an impact. But I think most companies, again, don't actually have that staff today. And so this is more enabling than threat, more opportunity than threat.

Jordan [00:22:30]:

Yeah. Obviously these things require a lot of conversation, but I really appreciate your quick take on that because I think that my takeaway from hearing you say that is if you're in a position where you can be using AI, you should be right because it is the way of the future. And even what you talked about, Erik, when companies like Microsoft are infusing this type of technology into their core product offering, that's a signal of what's to come.

Erik [00:23:04]:

Right? Absolutely. Yeah. It is going to happen whether we want it to or not, barring some crazy legislation or something. That's not the topic of today's show. Yeah, this is coming.

Jordan [00:23:16]:

Yeah, we'll see what's happening. Unfolding in the EU Ever makes its way to the US. That's another conversation for another day. But Erik, thank you so much for joining us on the Everyday AI show.

Erik [00:23:29]:

It was absolutely glad to be here.

Jordan [00:23:31]:

Thank you. All right. And just as a quick reminder, if you haven't already, please go to your everydayai.com sign up for our daily newsletter. Also, if you're not listening or if you are listening to this live, we come every day at 07:30, a.m. central standard time, but it's also put out on a podcast, so go to Spotify, Apple Podcast. Wherever you get your podcast, make sure to subscribe and follow us there. So we hope to see you back tomorrow and every day on Everyday AI. Thank you.

Gain Extra Insights With Our Newsletter

Sign up for our newsletter to get more in-depth content on AI