Enterprising Insights: – Market Insight Report Spotlights, Episode 31

Hello everyone. I'm Keith Kirkpatrick, research director with the Future Arm Group, and I'd like to welcome you to Enterprising Insights. It's our weekly podcast that explores the latest developments in the enterprise software market and the technologies that underpin these platforms, applications, and tools.

com, focusing on text to image generation in the enterprise and on customer data platforms. I'm gonna go through reports, pull a few key highlights out that I think are newsworthy and, and noteworthy, and then of course, conclude my show with the rant or rage segment where I pick one item in the market and I either champion it or criticize it. So without further ado, let's get right into it.

So, as I mentioned, I actually wrote a couple of different reports that just were published recently. These are what we call market insight reports. And really the idea here is, is to give folks an example or, or give folks some insights into a particular product type or technology in the market.

Uh, why don't we start out here with the report that I wrote on Text to image generation in the enterprise. So what are we talking about here? Well, I think everyone is pretty much familiar with this, uh, new capability that, that Adobe and, and Canva and many other companies Microsoft have come out with, which is allowing people to generate images simply by typing in a text prompt.

Basically, it's, you know, you would say, uh, draw me or create an image of, uh, of meadow setting with cows and a barn and make it look like it's from the 1850s. And the idea is that that prompt is then going to go out to the large language model that has been trained on various images, uh, and those descriptions of those images, and then is going to pull together, uh, what it believes is your intent in terms of creating an image. And of course, the quality or accuracy of that image is wholly dependent on how descriptive you are with that prompt.

Now, of course, the other issue that's at work here is, you know, how well was the underlying model trained on the data, how well those descriptions, uh, match up with what is actually in the image. Uh, so obviously it's, I'm sure most people played around with the various technology. Uh, as I mentioned, Adobe has a project called Adobe Firefly where you basically can go and try this out for free.

You don't need a, uh, right now it's, it's free for sort of a basic usage. But really the thing that's really interesting about this is it has so many applications beyond just sort of the g whiz factor. Uh, when you think about, uh, text to image generation, it really does enhance creativity and productivity.

Uh, let's take a, a great example, which is of course, uh, looking at marketing use cases. In the past, if you wanted to create a very, very personalized campaign, you had to have a designer create a new image for every particular scenario. So, for example, if I'm marketing a shirt and I want to place it in, uh, a number of different scenarios, you know, at the office, uh, at a party on the beach, in a nightclub, well, you would really need to, it used to have to actually have, uh, a designer create different images for each of those scenarios.

Uh, with generative ai, it's so fast and easy to create a number of different variations on that same image, using the same structure, using the same image type, and then just changing things like the background that really make it easy to create these very, very personalized campaigns. But at scale, and when I talk about at scale, I'm not talking about 10 images or 20 images, I'm talking about means of images that may incorporate a number of different types of variables there, uh, different cultural, uh, aspects to the image. Uh, for example, if you think about, uh, how, uh, you know, what we think here in the United States as a relatively benign image of a cow, if you were to go into other parts of the world, there is a whole different, uh, level of reverence toward that animal than there is here, let's say in the United States.

And being able to create campaigns, taking all of that into account is really powerful. So a little bit about the report itself. com and download it.

You just need to put in your basic details. It does not cost anything, uh, and it's available for you to download. Um, one thing I really wanted to pull out, uh, about this report is I do go into detail into how this technology actually works and what that, and, and the reason all of that is important is because we've heard a lot, and I've certainly written a lot about some of the challenges in terms of getting these text to image, uh, you know, uh, applications to work consistently and to actually reflect what the, uh, requester's, uh, desire is within the prompt.

And if, and I kind of go through exactly what's at work here in terms of how these models are trained, you know, what goes into it. And then of course, I talk a little bit about why it's so very difficult to get sort of that consistency when you try to generate an image, uh, every time you put in the same, even the same exact prompt. Uh, so I think that's important to really understand how it works and why it is, you know, why essentially we are not there yet in terms of being able to, you know, use this in the same way that we use other technology, uh, because it's just, it still needs refinement.

Uh, it still needs guardrails, obviously it still needs, uh, you know, people who are essentially designers or creative types, uh, to take that output and then fine tune it so it can be used, uh, in a commercial campaign. And that means, you know, making sure that, uh, obviously, you know, uh, as an example, I believe, uh, Adobe and, uh, Shutterstock, they both say that all of their models are only trained on content that is licensed or is in the public domain. But even so, it's still incumbent on companies that use this technology to really make sure that any output doesn't infringe upon any kind of intellectual property or any type of, uh, you know, copyright because that could have massive, uh, a massive impact on them if they were to get sued by using something that looks very similar to another, um, you know, a copyrighted image.

And it's found that, well, that was actually went into it as, as, as sort of the, uh, the fuel. So, uh, again, I think the important thing to remember here is that this is still new technology, very, very powerful, but it certainly needs oversight. Uh, one of the other things that I bring up in this report is the risks to the enterprise when it comes to things like bias and misinformation.

And what I mean by that is, if you think about some of these, um, some of the companies that actually offer these solutions, maybe training their models with images they find wherever, you know, out on the web. And the challenge with that, of course, is, or using that in a commercial setting, is that you could start to wind up getting images that might incorporate sort of historical biases. Uh, an obvious example is if you were to type in some of these, uh, show me a picture of a nurse, well, you might get a very stereotypical, outdated image that reflects what perhaps society would see a nurse looking like.

Not from today in 2024, but perhaps from several decades ago, you know, with a, you know, a white uniform and hat, uh, it might be of a certain ethnicity, all of that kind of stuff. So, you know, really the, the important thing is to make sure that whatever model you use, you really make sure that you know the prompts that you put in. They, they will provide an output, make sure that, of course, you know, that it matches what you're trying to do there.

And that there isn't sort of that incorporated bias or, or inaccuracy. I mean, in inaccuracy is another big one that a lot of these companies got, you know, hit with, where someone would put in a prompt and, you know, with a historical figure and that historical figure would not be represented accurately. And again, that's partially due to sort of the tuning and and training in that model to hopefully eliminate historical biases, but perhaps they went a little too far the other way.

So, again, the, the main takeaway here is making sure that, but any type of use of this in a commercial setting that you're looking for all of those things. Uh, let's see here. Um, the reports also include a small roundup of, of vendors that are representative of active players in the market.

It's a good starting point if your organization is looking to evaluate different applications or different, uh, packages there. I would say that the, the most important thing, of course, many of the more powerful things are not free in terms of the actual compute cost it costs in order to generate images from tax over time. We're gonna start to see some shifts in pricing.

I think some of the main vendors out there already have sort of a, uh, consumption model based on certain tiers. So you get a certain number of generative credits. Once you go above that, you wind up paying more.

So if you're, you know, generating new images every five, every five seconds, you know, putting in new prompts, that cost will add up. And I expect that ultimately we're gonna see some new pricing models to make sure that the providers of this software are able to recoup their costs and eventually turn a profit. Uh, so I think, um, you know, it's worth, you know, please, you know, go to the website, take a look at the report, download it.

Uh, it's certainly good food for thought, you know, if you're a buyer out there in terms of at least kind of get, you know, digging into sort of the basics of this technology and how it can be beneficial, uh, to you and your business. So with that, I'd like to switch gears to talk about another report that also came out, and this report's called Leveraging CDPs to support access to Data, AI and Automation. This is about CDPs, which is an acronym for a customer data platform.

And this is really, we talk about CDPs, we're talking about software that's really designed to unify an organization's data into a single centralized location. And why, why do we even need to do that? Well, if you think about organizations, a lot of them have, you know, important customer information sitting in their cm, CRM, maybe they have some information sitting in another database somewhere else, maybe they want to incorporate data from another third party if it's all spread out all over the place, that's a real problem because ultimately, if you can't get all of that information and derive a single source of truth, you know, when was the last time a customer interacted?

Well, if your CRM says one date and your point of sale system says another, that's going to be a problem. So what a CDP is designed to do is unify all of that information. And this is not a new sort of, uh, offering that they've been out for for many years.

But I think the real catalyst for more interest in this has been, of course, ai, because AI is value really can be, you know, can really amplify everything when it's applied across entire processes, entire workflows, and across an entire customer journey. So essentially, if you think about how AI is being used, you know, it's being used for personalization, sure does. That's not right.

It's only really powerful when it can identify me as an individual and knowing exactly how I've interacted with that company, you know, throughout my entire life or, you know, with that company through across every touch point. The only way you're able to really make sense of that is to make sure that there is a source of truth that captures all the data about me in one single location, and it's accessible. And that way when you're actually using ai, whether it's a, uh, self surface chat bot or, or if you're using an AI predictive engine to figure out a next best offer to me, it has to be in one place.

And that's what a CDP does. So, um, the report kind of delves into some of the, sort of the basics there, looking at CDP software, pricing models, how CDPs function within the technology stack. I talked a little bit about some of the benefits of using A CDP, and then of course, uh, I get into some of the, uh, trends.

And of course one of the big ones is of course, ai. Uh, the other one is that increasingly organizations are looking to fear out a way to leverage information that may not be their own information. It could be information.

I'll give you an example. If you think of customers and purchasing behavior, there is me who purchases who I have my own sort of customer journey information, but perhaps an organization wants to be able to understand what other people who are like me, you know, might be interested in purchasing and applying it to me. In terms of presenting me an offer.

I'm thinking about external data sources, syndicated data sources that take, uh, you know, whether it's survey data or, uh, a purchase data across a big aggregate base of customers that say that people like me about my age, uh, like to purchase, you know, these items with these particular attributes, bringing in that third party external data and harmonizing it with my own journey information and provide even more personalization opportunities. And the only way you can really do that efficiently is having a central repository to pull that in. Some of them offer sort of zero copy, uh, abilities to actually take that data in a federated way, not copy it into that record, but still utilize it.

That's all. There's a lot of power there in really creating personalized marketing sales engagement campaigns with all of this extra data. And why is this all important?

'cause in the end, we're seeing the phase out of third party tracking cookies, which is, if you think about the internet where you used to go to a website and they would track you where you went, I went from Dick's Sporting Goods and then I went to, uh, you know, another store, and then I went to a music store, and then I went to a clothing store. Well, they all, it's almost like bread crumbs, and they'd be able to track you and ascertain certain behavioral aspects. Well, because that's seen as being very, very invasive.

It, they're being phased out. So ultimately, marketers are needing to rely on first party data, which is data that is actually generated by that customer on their own properties and their own behavior. So all of that information has become very important.

And the way to manage, it's to be using a, a CDP to make sure that there is essentially a single source of truth that the, that a marketer or salesperson or a support organization can use to then engage with that customer, making sure that, that all of those attributes are, you know, linked to a single identity. So, uh, again, um, one of the other things about the report that I think is really important to take a look at are some of the features, really essential features of CDPs. I go through those and of course, uh, perhaps the most important one is, you know, how well, if you think about a CDP vendor, how quickly can they implement?

If you think about time to value, that is a, a critical, a critical component when it comes to any type of software purchase. And you think about A CDP where you're trying to incorporate different data sources. If it's not done fairly quickly, you could actually wind up, you know, being bypassed in terms of an innovation cycle, in terms of features, in terms of ai, all of that kind of stuff.

So that's another thing that we get into within the report. And as always, I have sort of a list of representative vendors, vendors that are representative of the market. So again, it's a great starting point for customers or, or folks who are looking, uh, potentially to, uh, to find a suitable CDP for them and their organization.

So again, this report is available on futureum group do com. Uh, so I encourage you to go up to that website and download it. Again, it's free.

And as always, if anyone has any feedback, please do feel free to reach out to me directly. Would love to hear from you both on the vendor side as well as any buyers out there. Alright, with that, I'm going to move to the final segment of the show.

And of course that is the rent or rave segment. And this is of course, the time when I talk about something that I'm either really excited about or I'm really disappointed, and I'm gonna give a, uh, you know, after we're into the second half of deer, I'm gonna start it off with a rant. And this one actually doesn't even come from me.

This comes from, uh, a friend of mine who, uh, was traveling recently and I believe she had to go to Las Vegas and went to a hotel with her and her family, and it was for, uh, I believe her daughter's volleyball tournament. So they were staying at this large property, everyone knows it. Um, it is on the strip.

And she had asked the person at the front desk if she could get a late checkout, and the person said, no, we don't do that. She said, well, you know, I am a platinum member of your rewards program and I'd like to speak with your manager. And the manager came down and said, hi, I hear you would like an, uh, an extended early checkout.

And she said, yes, I would. I'm a company of, you know, platinum member. And he said, absolutely, I I'd be happy to do that for you.

It's an extra $170. And of course, my friend thought that was ridiculous because that is one of the things that you get as a member of this rewards program is a, uh, you know, late checkout. That's one of the courtesies and benefits of being a member.

And of course they went back and forth and the manager decided that, no, not gonna happen. It will be extra. And I called to everyone's attention, not to, you know, hammer any one company up, that's why I'm not naming names.

You could probably figure 'em out, okay. But it goes to the issue that I've been talking about for quite some time, which is these organizations have all of the data, they have policy data, which is, hey, if you're a rewards member, you get this benefit. They also have the data saying that, okay, obviously if this person is rewards member, they're here a lot.

They probably will continue to be, you know, patronizing my business a lot. Why that manager decided to throw up another barrier was it's beyond me. And it gets the issue, which is you can have all the data in the world, you can have all the best systems, but ultimately it comes down to, in terms of providing great customer experience, training and aligning everybody's goals the same.

The manager, maybe they, they were new, maybe they weren't, but they need to be explained that, you know, the extra x amount of dollars that you think you're generating, you know, for you does not do anything in terms of long-term customer value, loyalty, uh, satisfaction, all of that kind of stuff. And I think that's where a lot of organizations really fall down. I, you know, when I talk about a lot of these, uh, examples of bad customer experience, most of them are not from small companies that don't have, uh, updated systems.

Most of them tend to be larger ones, and they ultimately usually wind up being some sort of either disconnect between data and people or people who just don't want to implement the, the good corporate policies or best practice corporate policies or whatever. And that's what really causes a downfall in that, in, in their customer experience. Which, you know, when you, when you think about it, you know, people tell their friends, that's why I know about this story I'm telling all of you.

And it's really, uh, ridiculous when you consider the, you know, how easy it is to create a great experience as opposed to creating a negative one. So that's my rant for the week and uh, certainly if anyone's interested in hearing more details, feel free to reach out. But that's all the time I have today.

So I want to thank everyone for joining me here on Enterprising Insights. I'll be back again next week with another episode focused on all the happenings within the enterprise application market. So be sure to subscribe, rate and review this podcast on your preferred platform, and we'll see you next time.