ChatGPT and AI for Marketing and Websites

 

 

Ben LeDonni, CEO, Creative MMS:

We have an agenda for discussion here on ChatGPT and Artificial Intelligence (AI).

There’s been a lot of recent changes and developments in the last couple of months. I don’t know how much everybody knows about it, but we’re going to start off by talking about that. I’m going to share my screen so you all can see the agenda. But really, just to take you back in terms of what happened late last year, ChatGPT launched.

ChatGPT is an AI tool that you can prompt with a question, get a text response back, and that text response is very robust in terms of feeling like it’s written by a human. It could be research-oriented, it could be ideation-based. It has a lot of applications and uses. And if you’ve ever tested it out, it’s really robust in terms of what it says back.

So, I wanted to first talk about what it is, and the impact there, and have anybody ask any questions from what they already know or don’t know. And then we’ll go into how we’ve seen it used and how we are using it. There’s some good conversation in there, to talk about what’s on the horizon for generative AI, outside of textual prompts like video and image. And then about what some of the impact and opportunity will be to us and our clients. Then we’ll leave off with some resources that we could probably pull together into the agenda.

I’m going to share my screen now. The agenda is up, so we can talk through it, and feel free to take your own notes. Or if you have this agenda, which is linked in the Slack conversation, you can note it there.

So, ChatGPT… What is it? Does anybody not know what it is, or has not experienced it or played with it? If you want to raise your hand, feel free to ask a question. And if not, I’ll assume everybody already knows what it is and has at least seen a prompt and response, or read a freakish New York Times article.

Cool. So, no need to worry about that! So — how we’ve seen it used, how we used it. I think I’m going to pass off this one to Jim to start, because I know he’s already had some discussions — and we’ve already had some discussions — with clients about the questions that they’re asking and ways that we’ve applied it.

 

Jim Van Horn, SEM Program Manager, BNP Engage:

Sure. For the most part what we’ve been relating to clients, and how we’ve used ourselves, has been for research, for content creation and ideation, and essentially to eliminate that base research phase and not starting with a blank canvas — especially for tough niche industries. It’s a good starting point to at least get the information you need.

 

 

You can’t copy and paste it, obviously. And there’s a lot of people — the more people that use it, anyway — are finding out. They’ll all get feedback like, “I got the answer, but the answer doesn’t seem right.” That’s exactly what it’s intended to be, because it’s scraping everything it can find.

It’s not a search engine. I think that’s the biggest misconception. A lot of people think that ChatGPT is a search engine and it’s going to come in and dominate Google and Bing and everything. And that’s not it at all. ChatGPT is not a search engine. It’s actually scraping everything in a search engine.

To give you an answer, it’s more like a chatbot, where you prompt it and you get an answer, but the information is supposed to be accurate. It is accurate most of the time, but that’s not to say that any kind of answer you get — as far as any kind of content creation or ideas to start with content so you’re not starting with a blank canvas.

It’s supposed to be reviewed by a human for accuracy for tone for voice, making sure it matches how we’re speaking to — let’s say — your Target Persona. Because these are things like emotion, it’s not giving you an emotional response. It’s literally just giving you information. It’s scraping.

So, how we’ve seen it and how I’ve seen use most often is exactly for that purpose. Instead of doing a Google search and having to scroll through 18 different scrolls and go to 18 different entries to find the answer that you need, if you prompt ChatGPT with that it’s doing that leg work for you. All that research that you might spend hours doing, it’s condensing all that down so it’s at least giving you a starting point. If anything doesn’t look right, you can fix it. You can do some secondary research just to make sure.

I actually got one unique usage from one of our clients on one of our calls. They like to produce content, but the content they produce is something akin to, “How is this industry impacted year-over-year?” or “What changes happened a decade ago?”, and then they need citations and sources for that information. So, we tested it live on the call and the unique thing is that it will give that information to you if you give it the right prompt.

We prompted to give us year-over-year data including citations for the source of the information of the statistics and it did exactly that! It dropped the links right in there and the links it dropped in there were high quality — they were authoritative .gov or .org type domains and websites. It wasn’t just a blog that it picked up that somebody said “oh this is the number for you.” It actually gave us what we asked for — credible authoritative sources.

Another thing I’ve been using it for personally is getting my ideas together for an outline. For one of our Lunch & Learns (just to test it out), I entered what the Lunch & Learn was about. I asked it to create an outline for me that I could follow — because it won’t create a presentation for you it won’t do that yet as it’s still just a textual prompt and answer. So what I asked it for was just a basic outline or an index of the presentation to help guide me while I was creating it.

It really did help me gather my thoughts in order so I didn’t get sidetracked. I didn’t start to go down rabbit holes. I just hit on each subject that it pointed out in the outline. I ended up getting my presentation done a lot quicker than I would have if I would have just tried to start from scratch and research everything step by step.

How good are the responses you get from AI chat tools? Simple answer is — the answers you get are as good as your prompts. ChatGPT doesn’t do well with very broad questions. If you ask it something like, “What’s a good digital marketing strategy?”, you’re going to get back a very, very, very broad and generalized answer that may not be 100 accurate.

But if you ask it with a prompt — something like, “What’s the best digital marketing strategy during the summer season for summer dresses in the United States?” or something like that, it’s going to be able to give you a much more accurate answer and a lot more information about exactly what your question is the more precise your prompt is.

 

 

I saw a good example of this one. A developer was using ChatGPT to build a plug-in for WordPress that then didn’t exist. It was simplistic and he even said he could have built his own plug-in for a design that he wanted in a few days. But, he decided to test out ChatGPT and it took him about three times running prompts through.

He ran the first prompt through which he thought was pretty thorough with the question to get the code he needed to write the plug-in. Then, he started to build the plug-in only to realize there were some gaps, so he had to go back. If you don’t find that it’s giving you the answer that you thought you were looking for or you thought was going to be the answer, you can certainly go back and just edit the prompt. Give it more detail or more clarification and it will give you back better responses.

One of the main differences between a chatbot and a search engine is if you keep manipulating a search in search engines, you’ll still get some of the same entries that you got when you searched the first time. If they weren’t correct, you just keep on searching and searching and searching trying to figure out how to get what you need to populate. With ChatGPT, it actually gives you fresh answers every time.

It’s not going to be redundant. So the answers that you get are responses that you get. The output from these tools are only going to be as good as the prompts you put in so be as accurate and precise as possible with it. You’re going to get back much better answers that you can use. It’s going to need a lot less editing and you could just dive right into giving it more emotion, tone, or voice to actually who your audience is or who’s going to interact with it.

That leads into the “How current is it?” question. That’s the other thing. Again, ChatGPT is not a search engine. Search engines are current. For example, if you search movie times for today, you’re going to get movie times for today because these crawlers are going out multiple times every single day to gather information that businesses put online. These chatbots are not necessarily current up to today. The information is old because there’s still somebody that has to update this data and it is historical information. It’s scraping.

It’s not to the point where it’s learning on its own per se. It’s not saying “today’s the 17th of February and I need to go out there and scrape new information for today” for anybody that might be searching or prompting for answers about anything happening on today’s date.

It’s still a static tool, so it’s only as good as the programmers and the people that build it. If you’re looking for information for the day of the search, you might get more accurate information by doing the research on search engines as opposed to a prompt. Or if you’re just asking a generalized question that’s not a question about current events, you’re going to get back some good answers to ChatGPT. It just might not be the most current information.

For ChatGPT and competitors, it is branching out.

 

Jeff Aspenburg, Director of Production, BNP Engage:

I have a question about the data. Can you tell me where the data is coming from? Is it coming from the internet? Is it coming from databases somewhere else? Where’s the data coming from to generate responses?

 

Jim Van Horn:

Hypothetically everywhere. It’s accessing databases. It’s accessing anything it could scrape online. Social media, aggregation sites, listing sites. Basically, anything online is what it’s scraping. That’s why some of the information isn’t as current as if you just did a Google search because it’s scraping historical information. It’s not going out there scraping information today to use today. There’s a lag period.

 

Ben LeDonni:

You can see on the screen I just asked that question to it and what Jim brought up about the data model and where it stops in September 2021. But I think what’s fascinating to me is that just like that type of question, to do the research traditionally using search engines you’re going to get a whole bunch of results that you have to comb through and form your own kind of research decision. Whereas here you’re getting kind of a summary. Who knows how good it is or accurate it is, but you’re getting a summary that comes back to you. You could hopefully trust it is accurate.

 

 

Tyler Baber, Director of Accounts, BNP Engage:

One of the things that I find interesting about it — and Jim you kind of alluded to this — but it’s not just “I type in a question and it spits out an answer”. Finding the answer in its database. It’s creating the answer based on, I don’t know what neural network means precisely, but it’s interpreting all of the data to give you the best answer back.

So, that’s another reason to me it’s not a search engine. A search engine thinks “here’s the question you ask, here’s the best answer I see to your question.” This is looking at “here’s the 25 best answers to your question, let me write it back to you in the way you’re asking for it.”

It’s like doing a first draft of an essay or paper. You do all the research, then you figure out how to present that research in a first draft, And then maybe an editor comes and looks at it and says, “hey you could tighten this up or you’re missing this piece of information.” That’s what it feels like to me.

 

Jim Van Horn:

You’re right. That’s what a lot of people I think fell into a hole before. They took this to be gospel — I asked it a question, it gave me an answer, the answer looks right to me, so I’m going to copy and paste it exactly. And, that’s not the case. It’s just giving you all the data. You have to do the research. You still have to put your eyes on it. You still have to proofread it. You have to give it some kind of voice.

It’s not like we can click a button and say, “I want this to be in an aggressive tone,” or “I want this to be in a placated tone.” It’s still just giving you raw information that you have to mold into. “Okay, it gave me everything I need now, I just have to rearrange it and piece it together in a format that makes sense.”

 

Lisa Howard-Fusco, Digital Content Strategist, BNP Engage:

One of the things that I wanted to bring up was my experience in trying to write an outline for a client for a piece of content. What I found is that — and as Jim said — you really need to know how it works and how to prompt it.

For example, one of the things that I did was I said, “give me an outline on this topic,” and it did and it was great. And then I said, “okay, now take that outline and integrate the case study that my client has.” It gave me an outline that at first looked great, but it totally changed the whole topic. When you look at it, it was not quite right. So it took twice the time because it wasn’t shades of meaning, or where things belong.

One, you’re still going to need a human element to work things out. Two, it reminds me — and I’m gonna date myself — of the early days of computers when you had to learn basic language in order to get a computer to do what you wanted it to do. And I think ChatGPT is in its early days where you need to know how this works in order to get it to do what you want it to do.

So, I would say to a lot of content writers out there, experiment. I’m hoping that there are courses that come out that teach content people how to prompt properly, because I have a feeling there’s going to be a lot of problems in the future.

That said, Jim, I’ve heard these things are learning at warp speed though and they double their capabilities in weeks. Is that correct?

 

Jim Van Horn:

Yes, they’re learning. One thing to remember is that — kind of what you said about the prompts — a misconception is that this is, It’s called, a large language model (LLM). It’s not a natural language model. I’ve seen a couple of use cases where it seemed like a pretty easy straightforward prompt and it was confusing the chatbot because there’s slang being used, there’s some misspellings of words, the grammar wasn’t 100 point-on because of the dialect they were using and the meaning of words were different, where they’re from So, it’s not a natural language model yet.

It’s not a conversational interaction where you ask, “hey give me this information,” and it knows what you’re talking about. It’s a large language model which just means it’s collecting a lot of content, a lot of data, a lot of information. That’s where you just need to know how to arrange the prompt and exactly how to ask it to get the right kind of output because it’s not going to do it the other way. It’s not going to try and figure out what we’re saying. We need to try and figure out how for it to give us the prompt that we want.

 

Ben LeDonni:

Yeah, I saw a prediction online that there will be jobs for really good prompters. Like a really good prompter will be an actual role at a company because they’re good at it. I had thought about the same thing as when you’re standing behind somebody that’s searching in Google. And they’re doing a terrible job of trying to find the information. And they’re putting in the wrong things. And you’re like “no, no don’t search like that, do this instead.”

This is how Google works so you know what you’re going to get back. I think you’re right on that the more you know the tool the better you’ll be able to use it and there’s going to need to be translation for clients as far as like, “use it to do this, not to do that” and “here’s how to use it because here’s how it works,” and “here’s what needs to be done by human because this tool can’t do it”.

 

Jim Van Horn:

Next, I was just going to mention that ChatGPT and its competitors are beginning to become a tool for complex developer and production teams for website and application builds. Again, this isn’t to say that you can prompt it to build a complex comprehensive website from scratch and that’s exactly what it will give you. But how it’s being leveraged for the mundane, day-to-day, or one-off instances like the case before where there’s a simplistic plug-in that just wasn’t available.

I think it had to do with a randomizer, where they want the functionality on their site so it was more transparent to the user in the community that when they did these randomized drawings, they got to see it in an animation on the website and it was more accurate or as accurate at least as you’ve randomized with type websites. But there’s no plug-in that exists for it.

Instead of the developer spending a couple of days creating the plug-in himself for WordPress, he actually prompted ChatGPT to spit out the code. And after about four prompts to get the code he needed then through a couple of minutes of testing, it actually worked. He was able to create a brand new plug-in that didn’t exist the day before for something they needed right now. Nothing too complex though

Where a lot of production teams or development, even design teams, are using this to generate code to help them on their day-to-day, it’s more for the redundant tasks, the QA tasks, or simple tests that are just generating the piece of code. It’s not hard. It’s not difficult. It’s just time-consuming. Or if there’s a bug or a glitch, they’re using ChatGPT to perform QA (Quality Assurance) for that code section to figure out what the issue is, where the disconnect is, and what the fix is; instead of having to do that manual research themselves

And, there is natural integration between GitHub and ChatGPT. There’s actually hundreds of different Integrations with a bunch of different platforms because again ChatGPT is just a tool. By being a tool, it can integrate with other systems to help enhance it.

GitHub is one of the biggest ones I know of at least right now that a developer would use. But it integrates with a lot of mail systems — everything from Gmail on Salesforce. So it’ll be interesting as this kind of matures and grows out and these Integrations. Get a little bit easier to make for the connections, how some of these processes through multiple Industries can be streamlined.

Some of these day-to-day tasks can be taken off the hands of a human and let them do the more complex things that are needed that a chatbot can’t do. And a lot of automation could run through these chatbot responses. They use a lot in real estate. They use it a lot in local government like HR Municipal resources and places like that.

 

Ben LeDonni:

Jim, I want to open it up to the Dev and design team. Based on what Jim just said from a code and an initial standpoint, any thoughts or anything you’re already seeing or have seen?

 

Jeff Aspenburg:

Definitely. I’ve seen some extensions for visual code. Visual Studio —that’s what the developers use to write code in-house — where you can just ask a simple question like how to make an array in JavaScript, show me an example, and it will write the code out for you. Or you can even ask a JavaScript question like, “how do I do a click on hover?” or a toggle switch on JavaScript. And you can type out the question and it will write out the code for you now on how to do it.

Normally, we’re searching Google and finding other forums of people showing us how to do that code or different examples. And you’re spending that time researching. Now, I can just type how I want to do it into, what I would call, a ‘code internet’. So I’m trying to figure out how to do something and I type in the question and it shows me how to do it. Where normally we are searching forever on how to do that.

 

Jim Van Horn:

Just a quick side note. When they built ChatGPT, they actually built this from the start to understand most programming languages. So, if you need to be more specific than that — say, for instance, you need to be in Python — if you prompt it for Python, it knows what you’re talking about.

 

 

Jeff Aspenburg:

And, I’m seeing a lot of coders using it to code things and I’m thinking “Wow, this is great!” But, it goes back to what Lisa said about the prompts and definitely looking over your code or your content or what it’s spitting out. Is it accurate? That’s the part that scares me too.

If I just copy this and put this in as a junior developer and run this production, most likely it’s going to break. We need to be checking over anything that we’re getting out of this system, and verifying that it’s correct. We, as the experts, are still going to be needed for a very long time because there’s so many flaws with technology. We still need to be looking over this and or just using it just for startups, in my opinion.

 

Jim Van Horn:

From a marketing perspective, I’ve already asked it to write a schema I wasn’t familiar with. I’ve asked it to write event triggers that marketers would use. It wouldn’t go on every website, it won’t be something that a production team would just instinctively do on a given page. So, I prompted it to write some new schema that was available and I asked it to write different event triggers that I could ship over to Leandra to add to like a button so I could track analytics.

I knew what to ask you though because I know exactly what I needed. Which goes back to what we all keep on saying — the output you get it’s only as good as the input that you give it. I know exactly what I need, so I just typed in. However, I needed to type in to get that information and it was really accurate.

As soon as I gave it to Leandra to test it out, I didn’t have to reset the search. I’d have to double-check because I’m not in code every day. I know how a line of code looks, like an event trigger, but I can’t write it over and over again just for memory because I don’t do it enough.

Having it, seeing it, and knowing that it gave me that in a couple of seconds, within 10 minutes we were able to start tracking. It helps the marketing team with some of these one-off code updates or tracking updates that might be needed on websites.

 

Tyler Baber:

I wanted to add on to Jeff’s point. When you’re saying people see us as the experts, one of the things that has given me comfort with ChatGPT AI is that it feels very similar to disruption we’ve had with Google. We’ve had this level of disruption before. What it means to be an expert is still true for us.

To take Lisa’s example, we are not expected to be subject matter experts on everything. Our clients know we are expected to have the critical thinking skills to be able to know how to get that information. So, in that way, ChatGPT can be a tool for that the same way if you stop and ask for directions. You ask someone who’s local. You don’t need them to tell you the name of every street. You need them to tell you how to get there with the information you already have.

It used to be an expert on Python or not Python. To be an expert on basic, you’d have to have coded thousands and thousands and thousands of lines of basic and that’s what made you an expert developer. That’s not what makes somebody an expert developer anymore. Now, you have to know how to do it right and, if you can do it right, you can do it for almost any language.

Same with if you’re writing a blog post or if you’re doing content strategy. If you know how to do it effectively, you can adapt to the tools and the information you need. That’s what makes you an expert. It’s not that you know the most about that company’s business over anyone else in the world. It’s that you know how to communicate that effectively and there is a degree of people who have been calling themselves experts for years. That’s something ChatGPT can just flat-out replace.

They weren’t actually experts — they were just good at summarizing other people’s research. We’re talking about being able to do an effective prompt. That’s a skill that you probably can put a dollar amount on, but if you can write requirements, if you can write an outline for someone else to write to, if you can read someone’s work and say what’s wrong, you’ve got the base skills to be able to do an effective prompt. And if you can’t do those things, then you’ll have to learn how because that’s the skill set that’s always been most valuable in what we’re doing.

 

Lisa Howard-Fusco:

I almost see it as running as an editor. You say, “Okay, ChatGPT, you’re my writer. Go ahead. Write it,” but you have to bring it back and check. Did they do it right? Is this working? You need to look at it that way, instead of the end-all-be-all, bow down to what it says every time.

 

Ben LeDonni:

One of the examples that I heard given that I really loved was Disney when technology came out to replace artists that were drawing everything out and they could generate a background and they could generate things and use them a lot quicker. And if you think of that Innovation, there are probably a lot of junior people that got replaced by this technology. But, in a sense, Disney grew from that and became much bigger, leveraging the technology. I think that’s an inflection point where we might be at as well.

Lisa you brought up a question around in the chat which is a deep question but based on what you heard in the podcast people will be able to create their own websites and website designers will be obsolete. I think we’d all want to say we completely disagree with that, but it’s good food for thought and conversation.

Before I dive into that, I want to just show one thing that’s related which is not just about text. AI engines are also about how their engines can be used to create things, like images. So, OpenAI also has this DALL-E tool which you can use to generate an image for me or a logo or something and it comes back with something generated from a system.

 

 

But all that to say, the interesting thing is that it’s not just text. The AI can be used to create other things, and those things will stack on top of each other to eventually be able to build a whole website. I have a prediction on that one last thought and then I’ll open it up.

We will probably be able in the future to train models to replicate the digital style guide for a brand, the editorial style guide for a brand, the voice and tone, the way it gets spoken about, and that model becomes our model. Then it can write chat responses to us, create things, and whatever, but I personally don’t believe that we’ll ever fully replace humans. This is because of the things that Lisa and everybody has already brought up.

 

Jeff Aspenburg:

Where does that copyright come into that? Images are one thing because I’ve seen that there’s been some copyright issues with the images. What about the content that ChatGPT spits out? What about copyright infringements? How do you know if you’re not stealing someone else’s stuff?

 

Jim Van Horn:

There are programs that detect AI plagiarism but for the most part, that’s why we keep on trying to iterate that this is a start. It’s a foundation. It’s not meant to be a ‘copy and paste’ mode. Even much of the information that we may generate using ChatGPT, we’re revising it a lot — enough for it not to trigger anything related to plagiarism online. It’s not giving because it’s not coming from one direct source.

So, unless you’re asking it to cite sources for you — which it can do when you’re asking it for statistics — if it’s pulling from 8,000 different microblogs, websites, aggregate sites, listing sites, social media accounts. If it’s pulling one more from 8,000 different places, there’s nothing to cite because there’s not enough to plagiarize. One word out of one social media post is not plagiarism.

 

Tyler Baber:

Everything that I’ve seen on this — especially on the graphic side — is that, by the letter of the law, there’s not a lot of intellectual property or risk of using these tools. But, there are a lot of ethical questions. These are the same kinds of questions frankly we should be asking ourselves when we are being influenced or copying a line of code from GitHub or whatever — the things we do every day. Where’s that line between borrowing being influenced by and copying is a whole other hour or more of talk.

But the intellectual content, like the intellectual property side of things, everything that I have seen, these companies are fairly protected because they’re databases that they’re modeled from, and databases that have been made available to them. Now, people might have contributed to those things like Wikipedia or online art sites without realizing what the terms of use were when they contributed.

One of the other pieces here is to read the terms of use before you load anything online because you could see your words rewritten somewhere else later. And you shouldn’t expect to get paid for it.

 

Megan Manning, Senior Design Lead, BNP Engage:

I’m old enough at this point where I feel like I’ve been through a few waves of things that were supposed to end web designers’ careers. And I’m not saying it never will happen because I will be the first one to admit I’m one of the most paranoid people about stuff like this. I’ve watched way too many post-apocalyptic movies.

But I feel like where it is right now, to me it sounds like exactly like when I was in college and Google Translate came out and everybody was like, “everybody’s gonna use this for all their language classes and nobody’s gonna know,” and it’s like, you tried it and you gotta see it. I’m not saying I didn’t do it because I had other things going on. I didn’t write that paper in Spanish.

But I think it– Tyler just hit on something as well. It’s like when we go out and look for inspiration – whether it’s through visual or music or whatever – you can take the super cynical approach that there is nothing original anymore. So, influence, right? So, if I’m looking on deviantART, that person put that up there without any expectations to be paid. Or, you know, another company puts their logo out there and you see it every day. You can’t help that there’s an amount of osmosis.

Much like if you’re writing text, I feel like if you want it as a designer to potentially use something like this where you said, “draw me a logo” for the first time. You can use it as a starting point. You might take that file just like you would from any free Vector site sometimes and go, “this has got the 10 lines that I think I need to start.” And then take it from there.

But where we are right now in terms of AI is – even with accessibility – there’s still a point where it reaches it’s got to be a human thing. I just hope it stays that way or that I’m not here when that overhaul happens.

 

Lisa Howard-Fusco:

One of the things if you saw there was an article that someone had asked ChatGPT to write a song in the form of Nick Cave, and gave it to Nick Cave. And Nick Cave is like, “it’s only a copy of a copy.” And so I’m wondering if that’s all AI will ever be able to do: a copy of a copy. It’s aggregating everything. It always has been.

 

 

And also, to the point of not having critical thinking, if it isn’t doing critical thinking, I know a lot of writer friends that are freaking out and going, “I’m going to be obsolete. No one’s going to call me as a freelancer.” The reality is they might not if you don’t know how to use ChatGPT.

But if you know how to use ChatGPT and I think about it sometimes as, “oh, good, ChatGPT can do the boring stuff that’s really long and tedious, and then I can get to putting it together in my creative way – in my critical thinking way – and offer that.” And I think overall that’s what they’ve been trying to tell us. It’s not AI and that’s it. It’s human + AI. But like you said, let’s hope we keep that value because I don’t think that’s ever gonna go away.

 

Ben LeDonni:

Yeah, Lisa, I love that. I mean, that’s kind of what prompted this conversation is that it’s always innovate and die. Digital marketing is always moving at light speed and changing and evolving. So, that segues into a question that’s on the agenda for us to discuss. Where is there opportunity for us to leverage it where we should be leveraging it?

Not to do it just to do it. But already for the things that we currently do. Jim told me we leveraged it to come up with the first ideas for frequently asked questions for our client, and then that wasn’t final by any means. It needed to be revised and refined.

But I keep thinking of the bookends where there’s strategy and planning and thinking that the human needs to do. That’s the prompts and things that need to go into the system, and then it could do a lot of things in between. But then, it needs somebody on the other end to be able to actually refine it, make sure it works, and speaks the language the way we want and is not broken or anything.

But I kind of want to open it up. We all know what our company does. Where are there use cases for this technology right away where we can either create efficiency or differentiation?

 

Tyler Baber:

I’ve got a suggestion. Everyone who gets a task that they wish they could delegate to someone else, that’s the kind of stuff that we should be seeing. It’s like AI is an intern that can work way better than any other intern ever could, right? It’s the smartest junior-level team member any of us have ever worked with. So, one of the things that we could immediately be using it for is any task, anything that comes across that’s like, “I wish I could delegate this to someone else. I wish someone else could do it and I could review what they did.”

Whatever that is – and maybe it’s not ChatGPT – but whether it’s like designing logos, anything like that, there’s something there that we could be using tools like this. And to that point, some of these tools are free, readily available, already being used.

It’s the same as developers using a plugin rather than rebuilding code. If we think about it that way and if everyone individually starts to think about like, “oh, this task. I’m doing it but I wish I could just review what someone else did.” That’s an opportunity to see if there’s a tool that could help automate this and do it faster.

 

Lauren Devens, Digital Marketing & Social Media Coordinator, BNP Engage:

This reminds me of one video that Ben sent me for a use case on social media. You can ask ChatGPT to generate a bunch of interesting quotes about whatever your industry or service is. You can say, “create five quotes about the importance of digital marketing for B2B brands,” and then it’ll come up with those five quotes, you can copy and paste those into an Excel spreadsheet, upload them to Canva, and then create graphics based on those quotes that you received. And then you have five readily available quote graphics that you can share across your social media.

So, that’s a really easy way to just trim down the time that you would spend designing individual graphics for your social media. Even aside from graphics, you can use it to generate hashtag ideas, to generate entire posts. You can ask it to be fun and engaging, or you can ask it to be a little bit more professional-sounding depending on who your audience is.

You can also use it for email marketing if you want to use it to generate subject lines to make sure that people are engaged and can open up your email and see everything that you have to offer. Of course, you have to start out with an interesting subject line. So, I know that there are paid subject line generators out there, but this one is really great because it’s free as of now.

There’s a ChatGPT Plus coming out, but with the free version, you can just ask it to make some email subject lines and choose from there. There’s a whole bunch of use cases, but I particularly liked that one about being able to create those quick graphics for social media.

 

 

Jeff Aspenburg:

And I believe that ChatGPT is integrated in Canva too. So, if you’re having problems accessing, I thought I saw this hack. If you’re having problems accessing ChatGPT, go to Canva, and then you can log in through Canva. That has the integration that will allow you to get in right away.

 

Lauren Devens:

Awesome. That’s really helpful.

 

Tyler Baber:

I mean that’s– it’s an open source database. I think you can download your own instance of ChatGPT and not have to use the public one but I’m sure that’s–

 

Jim Van Horn:

Yeah, it is open source.

 

Ben LeDonni:

Wow. Anything else in terms of application? And I guess I’d lump in, too, are there any tools that we should be using? I mean, I’ve been hearing in the agency world about Jasper a lot. If you’re looking at my screen, you can see there’s templates that it gives you for common content to start with like email subject lines like Lauren was saying, product descriptions, SEO blog posts, stuff like that. And relative to how much time it probably saves, it’s nominal on cost.

 

Lisa Howard-Fusco:

I’m wondering if the image prompts are going to create pressure. I was reading HubSpot and I can share it in the link. It was how Nike and Tiffany– the collaboration was overshadowed by AI. There was this whole we’re anticipating this wonderful collaboration and everybody was on AI going, “what’s it going to look like? Oh, let’s see, it could look like this.” And it came out and everybody’s like, “the AI was better.” I’m wondering if there’s going to be pressure to use AI to see what could be possible and then build upon that.

 

Jim Van Horn:

Yeah, Shutterstock integrates AI now. AI image generation. Now, for Shutterstock, you don’t necessarily have to choose anything in their library. You can generate through AI.

 

Lisa Howard-Fusco:

Lisa, that’s such an interesting idea, too, in terms of even just ad creatives or marketing taglines or whatever. You can spin up so many more ideas that can be filtered down much more quickly. Not to say that you have to release that to a client right away, but internally we could be using it to come up with ideas that we then say, “which of these five do you like?” or something like that. So, that’s cool. The Tiffany thing was crazy. That happened really quick, too. Twitter just blew up and hated what came out and generated a lot of better images.

 

Tyler Baber:

So, in terms of how we fit there and who gets replaced, I think probably the Fiverr-level designer or copywriter or who’s a freelancer. We’ve hired blog services to write blog posts and those blog posts are written by people and they were garbage. So, those probably do get replaced and then the companies that charge hundreds of thousands of dollars for their great brains, and then everyone who isn’t their great brains looks at it and says, “Okay, I guess all logos look like Google and Slack logos.” Those costing the same amount probably isn’t the same.

I think the type of agency we are where we’re not looking to be a full-service vendor because vendor work feels like something that gets replaced quickly here. But we’re trying to be a partner and we’re trying to help people work effectively and efficiently. Something like, hey, we can give you lots of ideas.

We can help streamline those ideas that might have been 40 hours of a marketer’s time coming up with 30,000 log lines, and we don’t need to spend that time. We are in a good position I think to work quickly and effectively to take advantage of the fact that a lot of things that used to be either seen as cheap that will now be seen as free or seen as unattainably expensive will now be seen as, “I have a hundred thousand dollar marketing budget. I feel like I should be able to get more for it.”

We are the type of agency that says, “yes, you should be able to get more for it. Let us help you,” as opposed to the larger agencies are like, “well, we’ve always done it this way and we’re not going to change.”

 

Jim Van Horn:

I think it’ll help smaller agencies leverage– because before all this happened, we couldn’t really play on the scale to be a content marketing agency. Now, we can. While we may not have a hundred copywriters sitting in the room next to us like some huge agencies around the globe, agencies our size could now almost spit out at that same volume if we just have, say, five good-quality copywriters to review all the content ideation and research generation done through chatbots.

 

Lisa Howard-Fusco:

Yeah, and I’m looking at what Jennifer popped into the messages. She’s saying content calendars, messaging pillars, elevator pitches. And I know Lauren and I were talking about that the other day. The staring at the blank screen that is sometimes sitting there going, “how do I start?”

That takes time and that’s the worst part of coming up with content is staring at that blank screen. It’s scary and it’s a little time-consuming because you’re like, “am I gonna get an idea?” And then you’re like, “yes, I did!” This can help feed ideas which is really– I wouldn’t discount that. I think that’s really valuable.

 

Tyler Baber:

Just one last point. One of our goals as a company is to talk about how we can do scalable services. In my mind, the speed that things happen and the speed especially, clients who aren’t going to understand… We have clients right now who think building a website, designing a website, is you click a button and it happens, right? They already think that. So, it’s hard to get them to see the value of our time sometimes. “I’m paying you 15 hours a month. Shouldn’t I be getting more for that?”

If we’re able to crack the nut on here’s the value you’re getting from us as experts beyond the time that we spend on it, then something like, “I’m starting my Harvest timer and staring at a blank screen for 15 minutes. I’m still working,” becomes a lot less important. It’s something that we don’t need to worry about as much if we can help communicate the value. This isn’t about the hours it took us to do. It’s about the value you got back from what was produced.

 

Lisa Howard-Fusco:

Tyler, do you think that’s going to affect agencies and how they bill?

 

Tyler Baber:

I think time and materials is going to continue to be something that gets pushed and pushed and pushed, but that’s always been the case.

 

Ben LeDonni:

To Tyler’s point, there’s opportunity to innovate right away on some of our current clients where that’s the case. Where they see us as delivering value and it’s a question of either how much more value can we provide inside those same dollars or what else can we be doing with the technology to increase the value that we provide for even increased dollars.

Knowing how to use the tool when a team doesn’t have a copywriter, let’s say, and to Jim’s point, didn’t ever see us as a content strategy or content marketing agency, and now all of a sudden we’re leveraging into that is super helpful. And I think, Tyler, you’re right on in terms of the tasks.

Jennifer just went through that list of ideation for content strategy, content creation draft form, social media posting, calendaring, and getting it out there and even any of the dev stuff, which is not super clear to me as far as how that could be leveraged. But I would just challenge you all to free to try and use it.

Even if it speeds up how quickly you get things done and you’re able to say, “hey I used this to do it,” great. If you need tools for it, raise your hand and say so because I’m always looking to innovate and show people what we can do.

 

Brent Cannon, Digital Account Manager, BNP Engage:

Question from my side. If we were doing a content planning session, are you able to look up content ideas around X, and then it’s giving us some ideas around that, potentially? So that we could use that as a starting point for sessions like that?

 

Jim Van Horn:

Yeah, you could. You just want to be careful with how you, like I said at the beginning, you don’t want to be general and just say, “give me content ideas around milk.” Like if it’s for Straus Family Creamery, you don’t want that. You’d want to prompt it with, “give me 15 popular, most searchable topics related to dairy farms. Incorporate sustainable or sustainable manufacturing operations.” That’s the kind of prompt you want to give it that will give you more output that Straus could use.

 

Brent Cannon:

I’m just thinking for our sessions, that’d be cool to just try and search and use this and come with some ideas that help us to get the creative juices flowing. But the intern did some of the work back there.

 

Jim Van Horn:

Yeah, absolutely. You just got to be careful with how you prompt it. You want to make sure that you’re incorporating what’s the ask. What do you want out of it? Content creation ideas, but then you also want to incorporate– you don’t just want to give it that blank of a canvas to say, “give me content.”

You’ve gotta narrow it down a little bit whether it be by industry, by target persona, by target audience. Some kind of limiting factor, so it doesn’t just give you a whole bunch of generalized content creation ideas. If you did that now, you’d just probably get content creation ideas around Valentine’s Day, President’s Day, different holidays, different national days in February.

 

Brent Cannon:

Look, here’s the content calendar, right? I love this.

 

Jennifer Greenjack, Director of Marketing, BNP Engage

I think one thing that it can’t take away from us, though, is the strategy that we put behind our work. I mean, yes, it can maybe help us to find an audience or a persona, but it can’t build out a journey of where people go step by step. So, there’s definitely more in terms of what we have to offer when it comes to the thinking and the strategy and the planning behind something versus just the execution of what it actually is.

 

Lisa Howard-Fusco:

Like the whole Nick Cave thing. It’s only going to recycle everything that it can glean. There’s no innovation. It’s up to people to innovate.

 

Jim Van Horn:

I don’t know if we ever want this thing to start learning or thinking ahead of itself. No.

 

Brent Cannon:

What’s that Will Smith movie?

 

Jim Van Horn:

I, Robot.

 

Tyler Baber:

Terminator! Is this a Skynet?

 

Jim Van Horn:

Yeah, again, that’s when you get into the realm of this could really go south quickly if it starts thinking for itself and asking questions for itself.

 

Lisa Howard-Fusco:

Well, you saw that conversation the other day in New York Times? I posted it.

 

Jim Van Horn:

No, somebody gave it– when this was first released, ChatGPT got asked if it was God and I forget what its answer was, but it was terrifying.

 

Lisa Howard-Fusco:

Yeah, and the constant, “do you like me? Do you trust me? Do you like me? Do you trust me?”

 

Brent Cannon:

It’s like M3GAN, that doll!

 

Tyler Baber:

Am I not supposed to be thinking that constantly? Am I an AI?

 

Jim Van Horn:

I think I’m a robot.

 

Ben LeDonni:

So, we have one minute left. Thank you all for joining. I appreciate geeking out and innovating here. A couple key takeaways. First of all, any notes that anybody has that you want to put in that agenda, I’ll share the agenda to put things in there. The two key takeaways that I have is us all knowing the tools that apply to our job will elevate us. That’s a really good takeaway just to start studying what’s already out there for you and your role.

And then getting good at prompting came up a lot as far as what is a good prompt and how do we give good input. So, we could do some more research on that, but just starting the conversation here. Let’s keep it going on Slack and I appreciate you all joining and geeking out.