The Agent Use Cases Most Ready for Primetime

0:00Today on the AI Daily Brief, the most in-demand agent use cases right now. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes.

0:17Hello, friends. Spring Break Week continues with another interview episode, and this time I am once again joined by Nufar Gaspar to discuss the agent use cases that we're seeing come up most often. Nufar is once again a brilliant AI analyst, product and strategy leader who has consulted with some of the biggest companies in the world on AI strategy, and who works with us at Super Intelligent on our agent readiness audits and our agent marketplace. Today, we're looking at some of the agent use cases that are most in demand coming out of both our audits and our agent marketplace as a way to potentially help you understand what agent uses are actually ready for prime time and which still remain a little bit farther away.

0:55All right, Nufar, welcome back to the show. This is not exactly a part two from what we did before, but I think a lot of people will connect the dots. Before we were talking about what mistakes are common that we're commonly seeing among organizations as relates to AI broadly, but agents specifically. Today, we're talking about what agent use cases are actually ready for prime time, what agent use cases are being implemented right now and people are finding success with. And again, this comes back and harkens to sort of a theme from that previous show as well, which is, you know, how can we help organizations spend their time more effectively in this agent and AI transformation?

1:31So let's dive in. We're going to talk about some use cases specifically, obviously, but I think that where we wanted to start is talking about based on our set of experiences, which includes, you know, a huge number of conversations with different types of companies, both in the context of the super intelligent work with with these agent readiness audits, but also in terms of your independent practice and consulting. What are we seeing on average at the risk of maybe being overly reductive? Where are orgs with agents on average, at least the enfranchised organizations that we're tending to deal with?

2:02Yeah. So let's assume that there is some bias because if they're very, very mature, perhaps they will not come to us. But everything that we've seen thus far are companies that we categorize as either agent initiation or agent exploration phase, meaning that they're either just starting to contemplate agents or maybe they've started working on a handful of agents or some agents' ideas. But in general, they're very early stages. And like we talked about the seven common mistakes, in many cases, they're so not read it that they have a lot of work to do in preparation for agents even before they can introduce the first one.

2:38So they're very, very early on. We are seeing some organizations and we actually encourage them in some cases that almost have no AI adoption but are looking into agents as the way for them to bridge the gap of how much behind they are. In some cases, these are smaller organizations where it makes more sense for them to hire an agent versus hire a human employee. And in other cases, they will still have to do the groundwork of getting agent ready before they will be able to do this bridging the gap with agents. Yeah, they want to talk about also the advance. Absolutely. Yeah. So I think maybe a better way to frame this even than just where orgs are on average is sort of what's the band of organizations that we're seeing, the common band from beginner to sort of a little bit more advanced.

3:26Yeah. So from what we've seen, even the most advanced organizations, one that already have agents in production, we're only talking about, in many cases, some handful of meaningful agents in production. And I'm not talking about the personal productivity that individuals are creating for themselves. These are not, in my book, meaningful agents in production. And I'm not also going to discuss the question of whether they're custom GPT or other such assistants or agents. Let's negate them from the discussion. they do have in production in many cases off the shelf kind of agents that they built with vendors like agent force microsoft and others the other cases that we're seeing and we'll talk more about the actual use cases but of course customer support is by far the most predominant category of where agents are actually mature in production and a handful in some cases of other supporting agents in some very well-focused functions within the company so i think that that bridges us into to maybe where organizations are not.

4:27So the second thing that we wanted to explore is, again, set up to what agents are ready for prime time. There are some pretty distinct patterns in what we're not seeing. I was on a conversation with a group of chief AI officers a couple of weeks ago at this point, a number from the finance industry, some from pharmaceuticals, a pretty big range of different companies. And there was one thing that was really clear, and these are definitely more advanced, or at least these are the most advanced portions of their organizations. Maybe the organization as a whole is an advanced, but these are people who are highly engaged, highly enfranchised, really thinking about these things all the time.

5:03They're the internal champions. And it's very clear that where they want to get eventually is agents involved in core business functioning, right? If they are in insurance, they want agents to actually be making decisions beyond just providing algorithmic advice like predictive analytics have done. They want to reorganize their whole companies around agenting capacities. However, to a person, none of them really feel like agents have the complexity, sophistication, duration capability to be used for those specific purpose-built use cases that are the very core to their company. And so instead, it seems like the place that they're going is focusing on not things that are unimportant, but just other parts of the functioning of the company that are, call it lower risk, right?

5:59That still allow them to get used to integrating agents into workflows, but are on things like customer service, marketing, sales. Is that something that you're seeing as well? Yes, but I'm not sure about the observation that the technology is not ready for what they have in mind. I think in many cases, the organization is not ready, whether it's from culture, technology, skills, use cases. In some cases, they don't want to tackle the most contradictory thing that they can do to get their employees basically to create an uproar because they're seeing the future. So that might be another thing.

6:36There is also a fear to take your core business and offload it to AI because of the potential pitfalls beyond the employees. And then lastly, technology perhaps is not ready in some cases, but I'm not sure whether in all cases this is in the case that the technology is the biggest hurdle. Sure. So basically, it sounds like you're seeing the same sort of inclination towards orthogonal use cases rather than core business function use cases. You're just you're not sure that that's more technology or the organization itself or some combination thereof. Yeah. And in many cases, we're seeing that other peoples are doing what other people are doing.

7:13So there is like a momentum here of automating the support function. First, I think from a business perspective, it makes sense to dip your feet in the water where it's more safe from various perspectives and then go there. Just yesterday, we had a conversation with a company that is very bold in their agentic approach, but they're saying, let's get the efficiency first off the table and put all of these agents that can free some of the bandwidth of our employees. And then let's tackle the core business with agents, not because we think that we can, but because we want to have our employees have more bandwidth to think about those core agents before we dive head deep with those.

7:57Yep. No, it totally makes sense. So, okay, then let's move to the meat of this conversation, which is what things people are doing right now? What are we seeing most commonly in terms of the agentic use cases that are being deployed, that are ready for production, that are actually yielding results for companies? Yeah, so I can name a few and then add whatever I'm missing. But one thing that is very straightforward and easy to use is basically to use agents that others have built, whether those are agents for coding that are probably one of the most mature. I think literally every coding platform now offers an agent.

8:31Some of them are better than others. Some of them are agent-native versus others that are just introducing agents almost as an afterthought. But those are getting some good momentum and some positive feedback. and of course my personal favorite deep research agents that are doing amazing job and we're also seeing some companies basically creating their own version of deep research so that it can work internally or in their own terms and these are some very good use cases. The other probably simplest thing that people are doing is augmenting the classical like Zapier or make automations with more agentic capabilities whether it's for planning some open-ended tasks that currently agents can do and beforehand they couldn't, or augmenting them with more like NLP-based interactions of text and speech, but just adding them to the existing flows concretely or metaphorically by creating similar automations using other tools.

9:27Let's actually pause. I want to break these apart a little bit because there's a lot to dig into here. So let's talk about the augmented automation a little bit because they're, frankly, the least interesting of these things to me. They're a very obvious starting point, but this is the area where people love to debate, are these things really agents or not? What should we call automations? What should we call agents? I'm well on the record with this one that I think people should give up the ghost and agents are close enough. And actually, people's understanding of that term is directionally correct, and they should just be fine with it.

10:00But it feels like this is an area where there are certain types of tasks that are just, they're so begging to be automated more. And it's really just figuring out these sort of very slight customization improvements for existing attempts at automation that are, it feels likely that these things are going to be completely boring, rote and normalized inside out of a very short period of time. Galileo called these the digital assembly line. In KPMG's taco framework, they called these the taskers, right? Things that are very, very specific. I mean, how much are organizations getting fired up about these things versus they're already in the kind of table stakes column?

10:44Yeah, in most cases, I think those will be like stuff that the employees will individually create for themselves and thereby they will not move any needle. In some cases, though, you are seeing some business processes that even using this method can be automated significantly more than what they've been doing thus far. And in those cases, you might be able to see even a higher use cases that are implemented using a not very complicated technology. Interesting. So basically, there's a risk at undervaluing the simplicity just because it is simple in terms of its potential business impact. Exactly.

11:21Yeah. Let's talk about deep research and coding for a second. And because for my money, I think that these might be the two agentic augmentations, however, you know, agent categories that to me, it is very hard now to justify doing things that you used to do without them, without them today. I think that the capabilities of research tools are, it's very hard for me to see how people who, I don't do a single thing that involves any sort of research without using these tools at this point, right? And in general, my workflow often involves cross-checking two to three to four of these to see how they come up with things, right?

12:05I have three different versions of deep research running at any given time. Now, obviously, we're just scratching the surface of deep research possibilities because the versions that we're using are sort of very individually designed agents that don't have access to proprietary knowledge bases. And obviously, what enterprises are thinking about is how to plug those into other data sources. But it feels completely like not in six months, not in 12 months right now, if you are doing anything that involves sort of research or strategy and not using those tools, I tend to think you're behind. And I think that we're pretty much in a similar spot when it comes to coding.

12:41Now, coding is interesting because there is, I think, an ironic or surprising at least amount of intransigence when it comes to adoption of the sort of coding agent tools and vibe coding tools and things like that among enterprise developers. And to some extent, there are pieces of it that are understandable. right? Like the first generation of these tools that are becoming popular, right? The IDEs, the cursor and windsurf, the specific text to code tools like Bolt and Lovable, they're absolutely optimized right now for an individual sort of developer experience, as opposed to integrating and interacting with these massive legacy code bases that have thousands of people working on them.

13:22And, you know, a guy who might, you know, be typing on them, the next day, someone different is using that code. But it still feels basically criminal at this point to not be taking advantage of these sort of new efficiencies of these coding tools? I mean, we at Super have, we had to let developers go who wouldn't get with the picture, basically, to change their processes around them. You know, is there, at what point do organizations just start to mandate that these are now the way that you do things? Is, you know, if you are not agentically augmented in these categories, you're just out, you're too far behind?

13:59Yeah, you know, you're very savvy in that. But when we talk to organizations, there are so much behind you. And a lot of that comes from what you're saying, that it's very optimized for the individual user. But another part of it, which is probably the lowest hanging fruit, they just don't know the breads of possibilities that these tools can take you. Because deep research, for example, people get stuck on the deep research part and forget that it's just like a deep reasoning agent that can literally get you any complex task done much better than anything. any other AI tool. So they don't understand that.

14:31Similarly with coding, like when I'm talking to engineers of everywhere on the stack, low level, hardware, high level, in many cases, they just didn't spend enough time to understand all the use cases that they can do with them. And that's why they're not using. So I agree with you that they should, but they need to spend some more time to get comfortable with the existing tools. And then they can augment them and modify them to be like enterprise ready for a donate.

15:26That's U-S-E-P-L-U-M-B dot com forward slash NLW. Today's episode is brought to you by Vanta. Vanta is a trust management platform that helps businesses automate security and compliance, enabling them to demonstrate strong security practices and scale. In today's business landscape, businesses can't just claim security, they have to prove it. Achieving compliance with a framework like SOC 2, ISO 27001, HIPAA, GDPR, and more is how businesses can demonstrate strong security practices. And we see how much this matters every time we connect enterprises with agent services providers at Superintelligent.

16:01Many of these compliance frameworks are simply not negotiable for enterprises. The problem is that navigating security and compliance is time-consuming and complicated. It can take months of work and use up valuable time and resources. Vanta makes it easy and faster by automating compliance across 35-plus frameworks. It gets you audit-ready in weeks instead of months and saves you up to 85 % of associated costs. In fact, a recent IDC white paper found that Vanta customers achieve$535 ,000 per year in benefits, and the platform pays for itself in just three months. The proof is in the numbers. More than 10 ,000 global companies trust Vanta, including Atlassian, Quora, and more.

16:37For a limited time, listeners get$1 ,000 off at vanta.com slash nlw. That's v-a-n-t-a dot com slash nlw for$1 ,000 off. Hey listeners, want to supercharge your business with AI? In our fast-paced world, having a solid AI plan can make all the difference. Enabling organizations to create new value, grow, and stay ahead of the competition is what it's all about. KPMG is here to help you create an AI strategy that really works. Don't wait, now's the time to get ahead. Check out real stories from KPMG of how AI is driving success with its clients at kpmg.us.ai. Again, that's www.kpmg.us.ai. Now, back to the show.

17:44embedded in these other tools as opposed to a standalone thing that was introduced as a breakout kind of standalone thing, it's perhaps being underappreciated in terms of just how differentiated it is, right? We at Superintelligent have been among the companies, and I've talked to lots of people who have had this experience, who spent a bunch of time trying to build our own kind of system to wire a bunch of things together, only to just see if we could run it through deep research and have it produce way better results. And we need some like, you know, 90 letter German word for just do it with deep research instead of trying to build it yourself.

18:19But I think that there is still, we're just scratching the surface of those use cases. And it's going to take some amount of time of diffusion of people sharing their specific uses of deep research for it to fully embrace. You know, on the coding side, I think that you're right. I think that it's going to change rapidly. I think that, you know, what I'm seeing is you're starting to see more discussion even within the enterprise around what things these tools can be used for right now, right? So you don't want to use it as your sort of primary coding environment, but you should deploy these things for refactoring or whatever it is, right?

18:51That you're starting to be more discreet about it. You're also seeing just an absolute flood of companies race in to try to fill the current gaps in capability and new challenges that these tools are arising from, right? So you're seeing on the consumer side, you're seeing companies that are coming in to try to make it easier to go from, okay, I've got this code base that I don't understand. How do I actually make it live on the internet and do things? There are companies that are coming in and doing that. On the enterprise side, you're absolutely seeing companies that are trying to come in and start to maximize for those enterprise use cases, even though they're more complex.

19:28So I think that that's gonna change pretty quickly. Anything else on those before we move to sort of like the big 800 pound gorilla in terms of capabilities or things that people are doing now with customer support? The only thing that we need for Super is to have the deep research on API. So if someone in the decision-making process can create a very good API-able deep research, the better. Yeah. Dear OpenAI, I know some of you guys are listening. Please let me know when the API is coming. All right. So let's talk about customer employee support. I think probably the area that is most discussed when it comes to agents.

20:05Yeah, so we talked about it before that customer support is probably the most mature agent EQs case out there, but there is an abundance of flavors for customer support, right? We're talking from all the way from a simple, very like a FAQ kind of an agent, all the way to the very impressive, completely autonomous end-to-end customer support agents. We're also talking about other flavors of that that can be agents that are helping to upsell or cross-sell your product because they identify opportunities. So we're starting to see these implementations. We're also seeing similar notions in internal employee support, whether it's IT support, HR support, legal support, payroll, basically everything that requires someone to answer questions in various capacities.

20:53These are perfect agents. And you can even kind of extend those to other types of support in the broad sense of the term, whether it's to help with employee learning and development or onboard new employees. In many cases, these are the most primetime ready agents that we have out there. And lastly, and you will probably claim that that's a category on its own, is everything related to outbound communication using various voice agents to create more sales or maybe reach out to candidates that we want to hire and so on. Yeah, I do think that those are, I think there's a couple different distinct categories there.

21:28I think that the part of the reason, though, that you might want to connect them is that all of these have a common thread of talking in air quotes to a person or finding out some information about them or from them and then integrating that with some preexisting set of information. Right. And this is just all the versions of that. AI is really good at right now, right? And so you're seeing this Cambrian explosion of basically every type of that interaction that AI can do. So let's talk about voice agents for a minute. I think that part of why voice agents are such a hot category is that this is a capability that is really useful right now.

22:17I mean, this is something that we observe. Part of where our voice agent interviewing came from was observing that in other areas, voice agents actually were doing a pretty good job, right? So I was looking over at the hiring space where companies were already deploying voice agents to do initial screening interviews and things like that. And they were working pretty well. The voice capabilities are good. Advanced voice mode had come out. So latency was better. All of these things had kind of come online as capabilities. and we had the thought, well, maybe you could basically turn that into a consultant whose job is just to sort of ask the right questions and grab a bunch of information.

22:55And it turns out that a hundred other startups or a thousand other startups are basically going through that same process of thinking through every other version of asking people questions, right? So you have voice agents for market research, you have voice agents for, you name it, right now they're coming out. And so I think that it's hard to call, voice agents aren't so much a category as much as sort of a common underlying technology capability that begets lots of categories. But I think that as companies are thinking about where they could be getting value from agents right now, it is not unreasonable to ask, what are current functions that involve us talking to people, literally talking to people?

23:35And would any of them be well-suited for one of the copious number of voice agents that are out there or rolling our own version of that? And there is so much value if you combine the voice agent with a deep research or deep analysis agents, then you get even a full blown consultant that is autonomous, basically. I mean, we call the sort of underlying technology behind, or at least I call it, I don't know if you'd call it this, but I call the underlying that we use for the agent readiness audit an agent consultant engine because that's what it feels like, right? Its job is to ask the right question, obviously with us helping kind of give it some initial ideas about what the right questions are, and then come up with and do analysis on the basis of some particular goals and particular knowledge, which starts to look very proximate to what consultants do.

24:25I think it's worth mentioning briefly, too, before we sort of broaden out again, the sales agent use case. This is one to me that feels very much, again, like I don't know that I've ever run across or at least in the last six months, run across any sort of sales organization that couldn't take advantage of the sales agent, the sales type, SDR type agents that are available right now. Now, that's not to say that they are. You can just grab one off the shelf and it's instantly good to go. There is more work than I think people might imagine or might want when it comes to getting their SDR agents up and running.

25:04However, sales is an area where there is no risk, I don't believe whatsoever, of there's always more leads. You always want more potentials. You always, you know, if a sales agent, like a human sales agent or sales representative had access to an agent that could get them 10 ,000 times the number of leads, they would be nothing but thrilled because ultimately more, more, more is the goal. And so I think that one, from an internal change management perspective, sales is a really good area where it seems highly unlikely to me that we're going to see big cuts in the sales organization because of agents.

25:42We're going to be straight, not in efficiency AI, but opportunity AI, where it's just how much more can we do? How much faster can we grow? How much bigger can we get? And I think that that's going to be a very useful bridge as employees try to figure out what management's goals are as it relates to agents. But two, they're also kind of an area where we're starting to get a little bit of a preview of the future. Lindy's Swarms came out recently. And Swarms are basically agents that beget other agents, at least in Lindy's case, where they start to do parallel processing. So instead of it being an agent that's doing data enrichment around a particular lead at a time, it's an agent that's fragmented itself into 100 different agents that's doing data enrichment across a whole set of lookalikes all at the same time.

26:28And it's all just efficiency. It's how much more can it get done in a given period of time? And again, organizations are just starting to put these systems online. I think that there's still a lot of work, a lot of customization that's necessary. But I do think that to the extent that, again, companies are looking for a place where they can dive in right now, get their hands dirty, and probably get some pretty clear ROI from agents, sales and SDR type agents are a pretty good place to look. Yeah, I agree. And also probably out like personalized outreach marketing, same methodology. Yep. Yep. I think it's sort of the same bucket.