Headless AI, otherwise known as bots, are taking over the world. How can these headless AI help businesses get work done more effectively?
In this episode of Defrag This, our host Greg Mooney is joined with Mark Towler who is the Product Marketing Manager at Ipswitch and Greg Jankowski who is the Community Manager at Ipswitch. Greg J. discusses how he is using headless AI in the Ipswitch Community, and we also go on a tangent about how AI works and what the future holds.
Today, I'm joined with two guests in studio. They're not actually on the phone this time which is kind of nice because I usually get kind of lonely in here. Greg Jankowski and Mark Towler, how you two doing today?
Mark: Very well, Greg. Thanks for having us.
Greg Jankowski: Very good, thanks.
Greg Mooney: Greg Jankowski, he's our community manager here at Ipswitch. If you guys are asking questions about WhatsUp Gold and MOVEit, you know, on our community...I believe it's still called WUGspace? It's usually Greg who's on there talking to you, right?
Greg Jankowski: It's called the Ipswitch Community now. So now, the community does support all of our products.
Greg Mooney: And Mark Towler, he is the PMM here for WhatsUp Gold.
Mark: Actually a PMM for all Ipswitch products at this point.
Greg Mooney: Oh now, see it keeps changing over here. I can't keep track.
Mark: We're very fast moving.
Greg Mooney: Yeah, so we wanted to talk today about headless AI. And Mark, you were telling me that you were reading an article, I believe a couple of weeks ago. We tried to get the actual author in here but...
Mark: We did actually. I reached out to him, and either he was on vacation or has a lot more interesting people reaching out to him via Twitter. But what struck me was not so much this article in particular, but the synchronicity of reading this article and then another one in close succession.
There was a great article that was printed a couple of weeks ago by Jensen Harris, CEO of Textio on The Sudden Rise of Headless AI; that's the actual title. But he's talking about AI and not Siri, not the Echo, not Cortana or anything like that, but a kind of background, you know, deep business-oriented AI who doesn't care about interacting with humans but does care about mining data and looking at interactions and how it's gonna impact every single aspect of business.
If you're not using a Headless AI, i.e. one that sits in the background and doesn't communicate with humans, then you're gonna be behind. And because they're learning systems, if you catch onto this five years late and start, well everyone else's AI is gonna have five years ahead of you. So it was interesting, and it was interesting how they're talking about this, doesn't just apply to manufacturing or you know accounting or any of these other ones.
But I noticed, because I'm in marketing, that it applies to marketing as well. And I thought, "Well, that's interesting. What would we use that for?" Literally two days later, I happen to read an article on eMarketer, and here's the headline, There's a Disconnect in Connecting Marketing Tech Tools.
Our recent survey finds only 3% of marketers said that all their systems in their marketing stack are integrated, and more than a third said, "We've got some integration, but most of what we have to do to get information from system to system is manual, spreadsheets, you know, datasheets outlets, that sort of thing."
Greg Mooney: So in order for these headless AI to work, integration is needed.
Mark: Well, that's it. I'm sitting here thinking, if we've got systems that aren't talking to each other, what good is a headless AI gonna do? It's garbage in, garbage out, right?
Greg Mooney: You'd be spreading data everywhere.
Greg Mooney: So I guess I wanna relate this to, you know, IT teams.
Mark: Of course.
Greg Mooney: And you know, there's obviously a relation because IT teams are setting up usually the marketing stack.
Mark: Right, and a lot of this stuff is organic. One of the reasons here even at Ipswitch we don't have necessarily full integration is because some of it's legacies, some of it's acquired through acquisitions, some of it's pieces that, you know, just haven't needed to integrate until recently. And one of the reasons I'm so glad that Greg Jankowski is able to join us is not only is he dealing with a corporate-wide Ipswitch community that really touches all parts of our stack across not just marketing but all of our other organizations. Apparently, we are using an AI system right now already.
Greg Jankowski: So one of the things when we launched our new community, search was a really key component of the experience. So we've actually started using an AI tool with our search vendor, and one of the challenges you have with search...is a great example...is everyone says they want Google search. Well, you can't do that through structured data. You know, we can have a long discussion just on that topic. But the AI in our search vendor provided us with first off, easy to set up.
It saved us work which is typically if you look at AI today, it's becoming easier to use and less maintenance, but you need to have good data and you need to have typically a lot of it. So what the big advantage from us on the community side is instead of us spending a lot of time trying to tune our search engine, it's actually just looking at what people do. It's figuring out what they went to. It makes a lot of assumptions. It looks at all that data and actually tunes our search automatically for us, so we don't have to do it. It's one great example of AI, and I think like you'd mentioned Mark, you're gonna see a lot more of this moving forward.
Greg Mooney: So how do you build up the data to begin with? So are you basically collecting the data that people are typing into the search field? And it's not gonna work off the bat. It's probably gonna work eventually, is that kind of how you go about it?
Greg Jankowski: You do actually solve right out of the gate with...there needs be some amount of data, and so you have to wait a little until you have enough data for it to start working and doing what it's supposed to.
Greg Mooney: Yeah, that makes sense.
Greg Jankowski: Although having said that, I think not only in our organization but pretty much any organization, there are huge amounts of data that are just sitting there. Now, are they good? Are they accurate? Are they valid? I don't know, but I know you're in a situation...and I've seen this in other employers in other organizations where they've got a huge database of customers, and they've got a huge database of sales. And you know, they may be all in disparate places where they've got a Salesforce implementation, they've got, say, a Marketo implementation, they've got an entire manufacturing implementation, but nothing's talking to each other.
So again, from where I'm sitting, and I've got no clue exactly how to implement any of these, my question is really, is the drive not only to get the AI into the systems and get them working for the company and getting it to make you...let you make better decisions, smarter decisions, but do we not have to integrate everything first? Or is that...sounds like is that even something that could be outsourced?
Mark: Well, outsourcing depend... I think every case is probably different. It depends again, what tools you're using, what's your technology stack, if you're using AI or anything else, can it talk to, you know, multiple systems at the same time? You know, I don't think we're at the infancy of this.
But as AI in general, you know, matures further, it's becoming easier to use, less set up, and I think that crossing different types of data, you know, whether it's platforms, whether it's repositories, that's gonna be where the real power is. You know, the more it can do for people, if you really look at it, AI is meant to just let you ask questions and come out with really meaningful answers.
You know, so if you wanna understand, if you go back to the search, well, hey, where do people find...where are they having the most problems, where are they having most success, you know, how well is some content received by customers? Well, there's ways that it can infer that, so it is about letting, you know, the machinery do a lot of the hard work for you and lifting.
Greg Mooney: Yeah, and I would assume a lot of this is being done via API, micro services, stuff like that on the back end and then actually using all these APIs to actually connect all these different services. In a way, if you think about it, AI is basically...I guess it's the interconnection of all these services, grabbing all these data from different repositories as you were talking about. At least, that's how I see it. Would I be wrong in assuming that?
Greg Jankowski: I don't think so. No, I think it's a way to look at. I think a lot of the AI that, you know, you're seeing today, pretty focused Salesforce...you brought up Salesforce earlier. You know, they have a product that...it's called Line Sign, and it is their AI. They're gonna use that for search. They're gonna use that for decision making. You wanna ask a question, you know, any CRM system produces a big pile of data as a great example. And that's one that if you can mine that data effectively, you're going out, try to make a report out of a little reporting tool. You know, one way to think of that is, it's just way beyond that. It's more of a "Hey, I have question. Give me the answer."
Greg Mooney: Yeah, I mean you're gonna mine the right data too. I know there's also a big issue with data overload. You know, are you collecting the right data to have the AI get its job done correctly. How does it find and extract the right data to do what it's doing? You know, these are all obviously still big questions, and I know that MIT was doing actually something on this a few months ago. It's very hush, hush. They don't actually talk about it, but it's basically trying to figure out how to teach the AI to find the correct data so that it's not, you know, malfunctioning in a way. Basically if it collects the wrong data and does the wrong thing, then it's basically a bug in the system so...
Greg Jankowski: Or you don't require which a couple of search companies do that today where they have a big room of data scientists. They look at the data and try to figure out, you know, how can it do this? How can we understand that? Most companies, almost all companies, don't have a big room of data scientists.
Mark: Right, and it's an interesting question because maybe the first job for the AI isn't necessarily... And maybe this is where we're going, because let's just say, people are implementing and it's being made available as part of platforms like Salesforce, but maybe the first really successful headless AI is gonna be the one that just dumps into your entire network and goes, "All right, what's not talking to each other and how do I set it up so that they are? And how do I determine what data is good and what data is bad?"
And coming back to Salesforce, there's been a couple of companies I've worked at where there's just no trust of the Salesforce database because it hasn't really been well taken care of, and people have dumped stuff in in different procedures, in different policies. And you know, you can talk to someone like, "Yeah, we've got 60,000 customers, we think, but probably a third of that data is old or outdated or bad or wrong." And no one's gonna go through 60,000 records manually. And on the other hand, if you had an AI that would do that and...
Greg Mooney: Or an intern.
Mark: Well, yeah. No, you know what? Even the interns won't. I think these days...
Greg Mooney: That's really mundane.
Mark: You'd have real trouble getting one, but even so, I mean I gotta say, I can look at this and think from, you know, a business perspective, yeah, I'd like to have an AI that would let me do things like the whole Target customer, Target knew she was pregnant before she did. You know that story? Should I explain? No one's familiar with it?
Greg Mooney: I didn't catch that one.
Mark: And that was the thing that's interesting, it's kind of pre-AI and that this woman started getting coupons for diapers and baby formula, things like that, and she's like, "Why am I getting this?" And her buying pattern through Target because it's tracked by the loyalty card, was such that they're back end systems. And this wasn't AI. This was a few years ago, so I think this is just a certain amount of automation. It was like, "Oh, if you bought these five things, within the next four months, you'll go on to buy baby stuff," and so...
Greg Mooney: No, I do remember this now. This was basically a way that [Target] predicting that you were pregnant before you even knew it, which is really creepy.
Mark: Yeah, exactly. And the interesting thing is with AI, I think the difference is it would know that it's predicting rather than just going you know, "Here's a series of if-then's." It's gonna be able to say, "Okay, it looks like these costumers are pregnant. Therefore," and it would probably also have some subroutine to go, maybe don't send them the "Congratulations! You're pregnant. Here's a coupon," when they might not even know until they've, you know, made some sort of...
Greg Mooney: It's like when I was shopping for engagement rings for my fiancé, and I would go onto my computer... Thank god she couldn't get on my computer because the minute you open up my Google, what do you see? You see ads for rings from like Barmakian and Zales. All right, I mean she's smart enough. She works in tech too that she would probably catch on pretty quickly that I was shopping around for a ring. So like things like that, the stuff can backfire too.
Mark: Oh, yeah. And at the moment, and again, I'm not 100% sure of this but a lot of that are comparatively dumb systems that are simply adding a lot of factors together as opposed to a true AI that is sifting through multiple different locations and basing, you know, these decisions and these advertisements and all of this other activity on a more coherent and conscious, if I can say that, understanding of who and what you are. And when I think of the fact that, you know, like everyone else, I've got a fricking microphone in my home called Echo that listens to everything I say, we know for a fact that most of the apps on our phones listen to whatever we say and are always...
Greg Mooney: Well, I think we talked about this in episode 3, I believe it was.
Mark: Yeah, we talked about this last time. Right, so how long before the AIs start integrating all of that information, and you're not just getting, you know, "Oh, you seem to be shopping for wedding rings. Here's a whole bunch of options," but you know, really super, super detailed kind of thing.
Greg Jankowski: Okay, when it says, "Hello, Dave," when I wake up in the morning, then I got a problem.
Greg Mooney: So back to the community, how do you picture this, you know, AI actually helping you in the community? Is it really just understanding the problems that people are asking without...
Mark: No, I think actually the more meaningful, you know, application of that is...so let's say I use WhatsUp Gold. I'm asking a question about x, y or z, is not just a simple search button; what other things can it tell me that would, you know, surface up the right thing for me to find. We have all kinds of information, and if you search just right, no matter how much time you spend on the search engines, you know, you'll sometimes get good answers, and you know there's content that you'd like to find. It really is about information, so when you're in there and you're trying to get help, it's making that more intelligent and a broader, richer experience.
Greg Mooney: So like the suggestions that would pop up.
Mark: Yeah, you know it's a really simple way of looking at it, but if the suggestions that usually pop up and usually aren't very good, if they really were contextual, really had a better understanding of what you were, who you are, what, you know, those sort of things, it can give you better answers.
Greg Mooney: This is gonna be huge for IT because I know as somebody who has worked in IT, as you said, like I will search for something on Google and I can't find an answer anywhere. It's like, I go to the forum. It's if it's on a forum, you'd think Google would be able to find it for me, but then sometimes you don't even know the name of what you're looking for, for instance. So it's like finding all those little pieces of snippets that you're typing in and having this actual program figure out what you're actually trying to think of, you know?
Mark: And this is where it's fascinating, to see how well Google has done some of this already. You can go to Google and type in, "What's that movie with the guy who climbs the mountain and falls in love?" and it will go, "Oh, you're looking for the English man who climbed up a hill or went down a mountain," or whatever. And it's amazing how many movies and things...
Greg Mooney: The movies, it works well with movies. It doesn't work that well with tech things though.
Mark: And that's the question because maybe there's too much information out there, or you're using the wrong terms, or it doesn't quite get it. And I see this...
Greg Mooney: Then there's misinformation too. It picks up a lot of the BS out there that isn't helpful, and so how do you... That is bad data. I go back to bad data, like that's where it backfires.
Mark: And this is where I think this is another area where AIs, I think, are gonna start proving their worth, and we're getting a bit beyond the business issues, but Facebook is out there desperately trying to find a way to counter the fake news because the one issue we've had with the explosion of megaphones where everyone can say anything to anyone all the time on the internet is there's almost no fact checking and there's no way to actually label things "Fact" or "Crap."
And Google is putting an initiative where they're trying to do that or to get it trusted. And I know Facebook is having problems because they're losing subscribers left, right and center. Everyone's lying and why doesn't Facebook do something about this? Well, they're trying, and maybe this is an AI issue because they've had humans trying to do this and trying to screen it, and it does...
Greg Jankowski: That's never scalable, that's...
Mark: No, exactly. It doesn't scale, and this is interesting too. It's got a big human cost. They've got an entire team that does nothing but look for illegal and obscene stuff that does not meet Facebook's terms of service, and these guys, you know, are quitting after six months because if your entire job nine to five is doing nothing but look for, you know say, child porn, that's not something you can do without a lot of counseling.
Greg Mooney: I believe they had an issue with the live streaming on Facebook with like those killings and stuff like that.
Mark: Yeah, exactly. So again, think about the advantages if you could turn that over to AI.
Greg Mooney: They took it down in five minutes by the way which is pretty quick.
Mark: Pretty impressive.
Greg Mooney: Unless that was misinformation, they just told me that to make me think that they took it down in five minutes. Now, that's the question.
Mark: If only there was some AI that would tell you whether or not that's accurate or not.
Greg Mooney: All right, well so, I think that's gonna do it for today's episode. Thank you, Greg, for coming in, Greg Jankowski, the community manager here at Ipswitch, and Mark Towler who is the PMM for both of our products, MOVEit and WhatsUp Gold. So thank you so much for coming in.
Mark: It's our pleasure, Greg. Thanks.
Greg Jankowski: Thanks.
Greg Mooney: Thanks, and until next time. I'm your host, Greg Mooney, and remember you can follow us @ipswitch and @defrag_this on Twitter, and make sure to read the blog at blog.ipswitch.com. Until next time, stay safe out there.