Full Transcript
[Music] Welcome to the Solspace Podcast. Thanks for listening.
Hey everybody, welcome back to the Solspace Podcast. This is Mitchell Kimbrough, your host, founder of Solspace. This is a solo cast.
Why? Could I not get a guest? Would nobody come on and talk to me anymore? No. Actually, I've been getting some pretty good feedback on these solo casts, so we thought we would do one where I could talk about some work I've been doing with AI. I've been trying to demystify AI from my own point of view.
I've been trying to write some code to explore what can actually be done with AI tools on a client website so that we could roll some of those capabilities out for our clients at some point in the future. And I just want to kind of cut through the hype. I've been around Silicon Valley since the very beginning of my career, and there's a lot of hype pretty much at all times.
And I've probably never seen more hype than now with regard to AI. So I want to know what's really there and what can we really do. Alright, so we're all accustomed to using ChatGPT at this point.
I'm assuming anybody listening to this is current with it, has tried it out, had conversations. If you're like me, you're using it to help your kids with their homework, math homework. I'm also using it to help solve coding problems as I'm working on a project and maybe I get stuck with, you know, there's some sort of validation error or whatever I'm getting.
I run it through ChatGPT and get a fix. I learn, I get better, and my code gets better, but the fix I found happens pretty fast. So I think we're all doing that.
I'm a little worried about Google because I'm not going there very much anymore. I'm using ChatGPT and putting much longer prompts in to solve much more complicated problems. Anyway, you guys are all doing that, so we're seeing that too.
We're also using ChatGPT as a creative partner. So that's another use, real world, like non-hyped up use for AI as a tool as a conversation partner. It can really help in the creative process, whether you're generating marketing content or you're thinking about how to solve a coding problem or you're considering developing a marketing campaign and you need some ideation.
It's been a good companion for that and it remembers the conversations. You can go back, pull up an old conversation, and catch up where you left off. So that's been a really useful capability that I know most people I know are using.
But this company of mine is directed at and serves marketing leaders. So marketing directors, VPs of marketing, people in charge of the marketing efforts at a company, or working on small marketing teams trying to drive in leads for a manufacturing company or a law firm or maybe some sort of a healthcare type of a business. A lot of those clients are our types of clients.
So as a marketing lead, you're asking yourself, aside from how I can use AI to help me with my daily work, what could I deploy on the websites that I manage with regard to AI that would be useful to our customers? How can I use AI to do the stuff on the web better? Or is there some new things that could be done using AI tools? So what I found in some of my experimentation and exploration is that yes, there are some real world right now capabilities that you can deploy on your website, leveraging the powers of these AI tools that are available to us now. So let's imagine that we're talking about a website that's responsible for real revenue, like it's accountable for and there's a scorecard that's tracking the website's ability to generate leads or generate money for the company. The website is part of the sales process.
It's part of the sales team. It drives in leads and it's one of the top members of the sales team if you zoom out and take a look at it. Now, imagine what would happen if those types of websites could become agentic.
So this is one of the key words in the AI universe, agentic AI. So agentic AI is when you have an AI tool that has agency. It has the capability to do something more than just generate responses in a predictive manner based on a conversation it's having with you.
More than just generating code snippets or generating marketing copy or generating solutions to math problems. So agentic AI is capable of doing things and in the context that I care about on websites, it's capable of affecting not only the front-end experience that a customer or a user of a website may have, but it can also do back-end tasks and it can do those in the context of a conversation with your customers. So here are three examples that I'll give you.
These are all real. These are all things that could be deployed by you and your web developer within two weeks time, at least at a prototype beta level. And these are actually powerful building blocks and capabilities that unlock a lot of creativity for marketing teams and the web developers who support them.
So I'm excited in what can be done here. It's a different world. We should have clear expectations of what's possible, but I'll get into that in a minute.
So first off, and this is one of the things we already deployed in our own product. So we sell a plugin for a CMS called Craft and a CMS called Expression Engine. This plugin is called Freeform and it's a form builder.
It handles form submissions, helps website owners create and manipulate the forms on their website. We built AI into our product and launched those capabilities recently. The first thing that was so obvious for us to do was to build in an AI scoring system.
So when Freeform receives a submission from a contact form or a lead gen form, the AI can be configured and activated by developers on the backend, just in the control panel of Freeform for that matter. You can activate the AI and you can tell it, you can teach it how to score the submissions that will come in so that you can then cause those submissions to be ranked and then routed to the right people on your team. So in the case of Solspace, we get three main types of submissions on our website.
One type is the type that we prefer, which is the incoming new business lead. So someone wants to engage us to do some web development work. They're going to submit our contact form that's running on Freeform.
That submission is going to come through and the sales team is going to work with those people. Sometimes we get submissions from current clients. That's pretty rare because we usually have Slack channels open and we have weekly calls with them and email conversations with them on a regular basis.
But sometimes they'll hit us up on the contact form in the case of maybe off hours emergency or whatever. And in the third category, aside from the sales and the client activity, is the product side. So Solspace, as I mentioned, sells some products, Freeform, Calendar for some CMSs.
And the product side of the business provides support for these products. Sometimes a customer of one of our products will come through the contact form on solspace.com and ask for help there. We want to route them to our support system, but sometimes they will submit there instead.
So this AI capability that we built into Freeform can score those submissions. And even if a user doesn't self-select what kind of submission they're providing or are putting into the system, the AI can come in and clean up that mess. The AI can come in and score it, look at the language that was used by the user and say, oh, you mentioned Freeform or you mentioned you had a question about some documentation.
You're clearly a product submission, like this is a product query. So we're going to route that to the right person and we're going to submit it into the support system and what have you. If it's sales, then it goes to that part of the business.
And if it's a client ops question, particularly if it's an urgent thing, we're able to flag that and notice that more promptly than the other types of submissions if we need to. Now you can imagine how this general concept can scale up and can accommodate a lot of different types of form submissions and can make these teams more efficient. The teams that handle the submissions, they can route these to the right people more quickly because the AI has stepped in and done its job.
And it's just been trained on what to do with simple prompts. So a second example, and this is more advanced work here. I mean, what I described on Freeform is pretty simple to activate.
It's just built into Freeform Pro. If you have a copy of that, get the most current version, you have access to that. Just click a toggle and configure it and you're good to go.
So the second type of agentic AI I'm going to talk with you about is a type that I've been sort of tinkering with and experimenting and exploring with. And this is the capability where you can add a chat GPT type agent to your own website. And I'm not talking about a bolt-on JavaScript widget that pops up a little chat thing.
That's been around for years and that's not as interesting as what I've been working with lately. So I built a demo that you can have a look at. It's at ai.solspace.com. You can take a look at that demo.
And when you go in there, you're going to have a conversation with an open AI type chat agent. And it's going to talk with you about what Solspace does. So I've prompted it.
I've given it a little bit of training in the form of some really clear and carefully worded prompts. And then it's going to have that material at the LLM level. And then you're going to have a conversation with it.
And you're going to be talking to that AI about what Solspace can do. And I've played with this at length. And I'm pretty happy with the result.
What I like about it is it's another way to understand what a company does. So a customer's not quite sure what problem they have or what type of a person they need to help solve the issue. They can engage with this chat agent, this tool that's been trained, and they can interact with that in a way that's sort of even more effective than all the efforts you've put into your own marketing strategy and language.
It's another way to understand what a complex business like Solspace or like a law firm or like a manufacturing company does. The reason this is an agentic demo is that as you're talking to this AI and learning about Solspace and exploring what types of projects you might want to work on with Solspace, the AI is capable of doing things to the web page. One thing that I've taught the AI to do every time it has a conversation is to ask the user, would you like to look at this web page in light mode or dark mode? And you've all seen this.
Like if you're looking at a web page at night, you can click most well-designed web pages and you can switch it to dark mode so you don't hurt your eyes late at night. Why is that interesting? Like who cares if an AI is capable of changing a web page to light mode or dark mode? I mean, if you're a creative marketing director, you see the power of this. If you can have a conversation with an AI and the AI can decide to trigger things on the front end of the web page, that opens up a lot of possibilities.
This demo that you're going to look at is also capable of triggering actions on the back end. So what I've taught it to do is to take in information from a prospective client, ask the right questions, and direct that client to contact us. Would you like to get in touch with somebody at Solspace to talk about your project? Yeah, sure.
Why not? Okay, well, can you give me your name? So the AI is going to ask you for your contact information, it's going to gather it, and then it will submit the contact form behind the scenes on your behalf. Why does that matter? Like who cares? Well, I think that's pretty powerful. If you can get an AI to be agentic on your website and you can get it to do things on the front end to alter the user experience on the front end web page, and you can get it to do things on the back end, submit forms, look up data in the database, check inventories, find out lead times on manufacturing, this is powerful.
This is a really useful tool. The great thing about it is you don't have to build everything at once. You can kind of tiptoe into some of these capabilities.
You can introduce a few things as you go. The third possible thing that you could do with agentic AI like this, building on some of these concepts from this second example I told you about just now, the third thing you can do, and I don't have a demo of this yet, but you could use an agentic AI tool to serve as a product configurator or, I mean, imagine a law firm, even a law firm could do this. You know, two or 300 attorney law firm could have an AI tool that's capable of talking to a new client and it's capable of assembling on the fly the right legal team to solve a certain type of a problem.
Now, your first question is privacy. So you're going to have these clients or you're going to have these manufacturing customers put in pretty sensitive information in order to get to a solution. Well, that's something that we have to deal with at the security and infosec level, but people are already submitting a lot of information into these forms and they're doing so securely.
Just because it's an AI tool doesn't mean it's insecure. So there's some issues to get into there, but you could build a configurator that is agentic and it's capable of doing a lookup on the back end. Let's imagine the law firm example, that it's capable of doing a lookup on behalf of the client who's having a conversation with it on the front end webpage, go into the database of the attorneys and their specializations and the states in which they have taken and completed the bar or examples of awards they've received or whatever the criteria is that a client wants to evaluate a set of attorneys for a legal problem they want to solve.
And that agentic AI could configure that solution set, put it in front of the client on the webpage, and then the AI can ask, would you like to contact this firm about this set of attorneys? Very similar idea with manufacturing. Would you like to have me assemble for you this robotic arm that you need to replace the one that's in your factory in Detroit that's on the assembly floor? Would you like me to check to see if this is in stock? Would you like me to check to see if this is a quick ship item? Shall I check on pricing for you? All these things are available in databases and by API calls to other systems. An agentic AI tool, custom built for a specific client website, would be capable of serving that set of customers and their set of needs.
So some of this stuff sounds like it would be really expensive to build, but the surprising thing is there's a lot of capabilities that are available right now. You don't have to be a giant enterprise company with in-house computer science PhDs to do this. The reason there's so much buzz about AI is tools like this are pretty easy to deploy now.
Some of the more advanced versions, you'd really want to run that through some solid QA. For the reason that these AI tools, one of the things you have to change about your expectations is that they are what they call non-determinative. So if I write a line of code or several lines of code, like I've been doing for 25, 30 years, I know exactly what it's going to do under what circumstances.
With AI, I'm not exactly sure when it's going to ask the customer when they want to change to light mode or dark mode. I don't know when that's going to happen exactly. I don't know when it's going to submit the form exactly.
It's non-determinative. So that makes things interesting, but that's all part of the new flexibility and capability that AI tools like this bring to bear. So those are three solid examples of agentic AI.
The first one is something we built into our Freeform plugin for craft. The second is available on the demo that you can see at ai.solspace.com. And the third is something that's really just a few weeks away in coding time. I just need a client who wants to do it.
At any rate, reach out to me at Solspace, Mitchell at solspace.com. Get in touch. We can talk about this some more. If you have a project that you want to explore, we can do a proof of concept and see what's possible.
Again, you would be surprised at how approachable and inexpensive some of this stuff is. All right. Thanks for listening, everybody.
Hope you have a great day and talk to you next time.
[Music] You've been listening to the Solspace Podcast.