# Models, Harnesses, and Multi-Agent Systems

- **Show:** Practical AI
- **Published:** 6 Aug 2026
- **Source:** https://share.transistor.fm/s/063cfaad
- **Transcript:** https://podcasts.opennash.com/shows/practical-ai/models-harnesses-and-multi-agent-systems-d1f535/
- **Provider:** OpenNash Podcasts (machine-generated)

## In brief

Daniel Whitenack and Chris Benson argue that AI has moved from chatbots to autonomous software agents, and that organizations should plan for fleets of agents powered by several different models. They define the key terms (model, open versus closed weights, agent, harness) and say these pieces are all just software working together. They conclude that multi-agent systems will become pervasive and that businesses should avoid depending on one vendor.

The central tension is between convenience and control. Vertically integrated stacks from large vendors are easy to adopt, but they risk lock-in, and a vendor can change terms or absorb your product. Open harnesses and swappable models cost more effort but preserve flexibility and lower costs. The hosts say several approaches will coexist. Their advice is to start small, iterate quickly, measure cost sensibly and always keep backup plans.

## Key insights

1. **A model is just a function** Daniel explains a model as software that transforms one kind of data into another, such as text into text or an image into labels. Keeping this in mind removes the mystery: agents and harnesses are also just software built around that function.
2. **Agents pursue goals; AI features answer turns** Daniel says an AI feature, such as a chat box, responds one-to-one with a user. An agent is connected to company systems such as email, calendars or NetSuite, is given an outcome, and works with some autonomy.
3. **Open models and closed models trade control for power** Running a model needs both software and the weights, the numbers that configure it. Open models like Gemma can be downloaded and self-hosted cheaply, while closed frontier models are reached through paid APIs, and Chris says the American ones still lead.
4. **Multi-agent systems will become invisible infrastructure** Chris argues that better models will not remove the need for many specialized agents, because models are still just functions that must be coordinated. He expects such architectures to be so pervasive that people stop discussing them.
5. **Vendor lock-in is the main strategic risk** Daniel contrasts vertically integrated stacks with open harnesses where the model can be swapped, and says control and agility favor the latter. Chris warns that one vendor change can crash a business built on it, so he advises starting small with backup plans.

## Questions this episode answers

### What is an AI model in simple terms?

Daniel Whitenack describes it as a function in software that takes one kind of data in and produces another out, such as text to text or image to labels. It is a transformation of data, not magic.

### What is the difference between an AI agent and an AI feature like a chatbot?

A chatbot-style feature is turn-based, with you typing and it replying. An agent is connected to systems like email or company databases and is given a goal that it pursues with some autonomy, according to Daniel.

### What is an agent harness?

Chris Benson describes it as a software system, like an operating system for AI, that accesses the models, manages the agents and orchestrates what they do.

### What is the difference between open-weight and closed AI models?

With open models, such as Gemma, you can download the weights (the numbers that configure the model) and run it yourself. Closed models stay on the company's infrastructure and are reached through a paid interface such as an API.

### How should a company begin adopting agentic AI without getting locked in?

Chris recommends starting small, experimenting and iterating quickly, and keeping one or more backup plans. Daniel adds that building agents in a vendor-agnostic way preserves control and flexibility.

## Facts

- **Daniel Whitenack** [0:00:41]: "I am CEO at Prediction Guard, and I'm joined as always by my co-host Chris Benson, who is a principal AI and autonomy research engineer."
- **Daniel Whitenack** [0:03:14]: "I give you an image and you tell me what's in the image, it transforms to objects in the image."
- **Daniel Whitenack** [0:07:26]: "You can go to Hugging Face, you can download the weights of that, and there is software that runs it"
- **Daniel Whitenack** [0:06:16]: "Nvidia and others making pleas to the community around the importance of open weight models or open source models."
- **Chris Benson** [0:09:18]: "there's a lot out of China, which is kind of the the the powerhouse planet of open models now."
- **Chris Benson** [0:09:43]: "if you're willing to do APis for closed models, uh, the the the American-based ones, uh, are still leading in that space."
- **Chris Benson** [0:11:22]: "I've heard it described as an operating system for AI, where it is it is accessing the models that you're using"
- **Daniel Whitenack** [0:13:46]: "What the software is meant to do is to accomplish a goal, and in many cases, to operate with autonomy."
- **Chris Benson** [0:23:11]: "swarm is usually a fleet, but a fleet is not necessarily a swarm."
- **Chris Benson** [0:24:43]: "And it's happening very rapidly in 2026 compared to even 2025."
- **Chris Benson** [0:25:50]: "Walmart advertises drone delivery, as does Amazon, and you know, we're seeing more of those."
- **Daniel Whitenack** [0:28:29]: "you'd look at how the pit stop evolved over time from like, whatever 20 s down to 2 s."
- **Chris Benson** [0:35:43]: "And so I'm actually actively writing a book on this topic, so, um, among other things."
- **Daniel Whitenack** [0:36:42]: "for example, LangGraph or Pydantic agents, or Hermes agent, or Open Claw, et cetera, et cetera."
- **Daniel Whitenack** [0:37:21]: "inbuilt agent features in Anthropic or OpenAI's platform, but also others, whether it's like IBM Watson agent thingy"
- **Chris Benson** [0:39:42]: "it came and it went because suddenly the the cost of that caught up to people."
- **Daniel Whitenack** [0:41:08]: "if you're primarily using Google Drive and Google Docs, then you're automatically gonna get one of these very opinionated AI integrations across those various apps."
- **Chris Benson** [0:47:12]: "all it takes is one of those vertical integrations to change something."

## Transcript

**[0:00:02] Narrator:** Welcome to the Practical AI podcast, where we break down the real world applications of artificial intelligence and how it's shaping the way we live, work, and create. Our goal is to help make AI technology practical, productive, and accessible to everyone. Whether you're a developer, business leader, or just curious about the tech behind the buzz, you're in the right place. Be sure to connect with us on LinkedIn, X, or Bluesky to stay up to date with episode drops, behind the scenes content, and AI insights.

**[0:00:32] Narrator:** You can learn more at practicalai dot fm. Now, onto the show.

**[0:00:41] Daniel Whitenack:** Welcome to another episode of the Practical AI Podcast. This is Daniel Whitenack. I am CEO at Prediction Guard, and I'm joined as always by my co-host Chris Benson, who is a principal AI and autonomy research engineer. And on these episodes where it's just the two of us, no, no guest, we like to take take a moment to really. Uh, sometimes talk about just a topic we wanna talk about and learn about, sometimes be more educational, dig into specific details.

**[0:01:13] Daniel Whitenack:** And and I think that's more of the direction we wanted to go today, right, Chris? It it is, and and actually to. We've done several M of these recently, and I think it's important to to call out. If you've been listening to some of the recent episodes, there's so much happening right now that one of the priorities for us in the show is to try to help people. You, you know, kind of really stay up with what's going on.

**[0:01:38] Daniel Whitenack:** Not just from a newsy in a practical way,

**[0:01:39] Chris Benson:** standpoint, in a practical way, so that you can actually go do this yourself in your organizations. And we kind of realized that while there are some like, we always have a a huge funnel of guests to come in, we realized we needed to kind of pause and take a few episodes and try to just get people caught up to where they need to be. Because so much is happening right now. And so that's a big part of of why we're doing this right now.

**[0:02:04] Chris Benson:** And so we hope, uh, we hope this is gonna be helpful for a lot of folks. Help

**[0:02:07] Daniel Whitenack:** yeah, it's. This is always evolving, and the way people use terms is always evolving, so it's it's useful to even sometimes redefine some of those terms. And, uh, I, I promise we won't belabor this point too long, but I do. I do think it's worth us defining a few terms here up upfront. So we're just all on the same page, we're talking about the same thing. And some of these we've defined before in other episodes, like you said, but again, uh, the semantic variance of terms changes over time, and it's, uh, it, it's worth just updating and and reminding.

**[0:02:44] Daniel Whitenack:** So the the 1st term, obviously, is, uh, an, an AI model, or a model, and very often a. And this actually hasn't changed, although maybe some of the things we refer to as models have changed. But an AI model is really if if you're more coming from, like the technical side, you might think about this like a function or a software function. Maybe if you're coming from another angle, really, it's just a transformation of data, right?

**[0:03:14] Daniel Whitenack:** Like, I give you an image and you tell me what's in the image, it transforms to objects in the image. That's an object recognition model, right? Or I give you a series of words and you tell me what word should come next. That's a large language model, it's a transformation of input to output, it's a function implemented in software, and these models generally do a a a single type of transformation. So a large language model does that transformation that I mentioned.

**[0:03:48] Daniel Whitenack:** It takes input words or text and then generates text output. Uh, language vision model changes, the input side changes, so it might take language or text plus images or just images and then produces text output. An image generation model, as the name might suggest, takes in text and goes the other way, it generates an image out out the other side of the function. Or data transformation, there's, you know, Imma, uh, basically, any combination. Just think of the model as that data transformation.

**[0:04:25] Daniel Whitenack:** And then you could have a lot of things on the input side, you could have a lot of things on the output side. And depending what type of thing, type of data is coming in on the input side, coming out on the output side, we refer to these as different types of models. So an LLM, a language vision model, an image generation model, a video generation model, uh, a forecasting model, uh, an auto regressive model, I, in that case, um, an anomaly detection model, et cetera, et cetera.

**[0:04:56] Daniel Whitenack:** There there's all sorts of types of models. S And not all these are neural network based models. A lot of the ones we typically think of today are. But a model itself could be composed of guts that take a variety of forms. And the the way that that function does the transformation, that could take a variety of forms. But at the end of the day, it's a function or a data transformation.

**[0:05:22] Chris Benson:** no, that's a great explanation. And and it really, and like, the architectures have gotten much more complicated over the years and and and able to do a lot more functionally. But at the end of the day, I think you men you mentioned neural networks and stuff is, uh, it's still neural networks, you know, kinda plus, plus, uh, uh. Architecturally, in terms of how that function is. But I I really like the notion of a function. Because, you know, you give it an input and you get an output, which in this case we call an inference.

**[0:05:51] Chris Benson:** And uh, and and you know, kinda keeping that in your mind as to what it is is pretty important, cause it's not magic. You know, we get a lot of that in the the mass media about ai, uh, and so, yeah, we're talking neural network, uh, based architectures that are able to do, uh, transformations along the way. So, um, a, and then and now we're building all sorts of stuff around it, like agents, you know, uh,

**[0:06:16] Daniel Whitenack:** And and the uh. Maybe one distinction too that's really important with these models is some are what's called open weight or or open source models, some are closed models. good point, And, uh, and so these models can take a. At the end of the day, they're all kind of doing the same, the same sort of thing. But the way that you can access them as a user might be different. And some people might have seen in the news, even just in recent days, Nvidia and others making pleas to the community around the importance of open weight models or open source models.

**[0:06:55] Daniel Whitenack:** So what this means is if you're gonna run that function, if you're gonna run that data transformation, so you're gonna take text in and generate text, or you're gonna take text in and generate an image. Some there's really two components to that. One is you have to have software that runs that model and actually executes it. The other thing is that you need what are called parameters or weights and biases. These are just a data set of numbers that configures that software function to operate in the way that it operates.

**[0:07:26] Daniel Whitenack:** And so you need both of those things, and sometimes you you as a user may have ac- direct access to both of those things for a given model. So let's take, for example, an, uh, an open model, like a, maybe a Gemma model or something like that. You can go to Hugging Face, you can download the weights of that, and there is software that runs it, the Transformers library from from Hugging Face, for example. And so you can have both of those components.

**[0:07:56] Daniel Whitenack:** You can run it, you know, on your own laptop, in your cloud environment, wherever other models are not re released, either the software component or the and or the data component of the weights are not released. And those are generally called closed models, often kind of just accessed through a managed API. they, yeah, So program programmatic interface or a chat interface or other interfaces. And those are sitting behind a a company's proprietary infrastructure. They're still running the software, they they still have the same thing that they're running on their end, um, just their version of it, and they keep that proprietary as their IP.

**[0:08:36] Daniel Whitenack:** So that's what we mean by open weight or closed.

**[0:08:39] Chris Benson:** and one more thing to throw in that because it it's relevant to the rest of the conversation is these models can be of many different sizes. Um, what we call frontier models, which are the the the largest, most complicated and produce the most sophisticated output. Uh, often require, uh, hardware that is going to be prohibitively expensive, uh, and so that is a a huge investment if you're gonna go that way, uh, on, yeah, and and and the cloud providers and these model these, you know, kinda top tier model creators, uh, have enormous data centers filled with with server racks to manage this.

**[0:09:18] Chris Benson:** Um, not all models are at that level. Uh, you can, you know, you mentioned Gemma, uh, which is, uh, some of the smaller, gradually smaller models, uh, that can be downloaded and run locally with whatever, uh, hardware you have available. Uh, if they're open, uh, well, the gemmas are open, and there's others as well, uh, both, uh, out of Europe, uh, there's a lot out of China, which is kind of the the the powerhouse planet of open models now.

**[0:09:43] Chris Benson:** Kame. Yeah, Yep, and so, uh, and that's uh. And it's interesting that we're in a moment where if you're wanting to run, uh, a fairly sophisticated model on your own, your, uh, w- you know, on your own server, you're probably most likely going to a Chinese model these days. Uh, they're the best, or or, or they're the they're best right now, or, uh, uh, if you're willing to do APis for closed models, uh, the the the American-based ones, uh, are still leading in that space.

**[0:10:10] Chris Benson:** So that's just a a general shakeout of where things are at this moment.

**[0:10:15] Daniel Whitenack:** yeah, okay, so model, open weight, closed. What about what about the agent and or the agent harness, Chris? What? What are we? What are we talking So, about there?

**[0:10:27] Chris Benson:** yeah, we're like, if you, if you, uh, you know, just we've, we've used this before. If you think about the model, that function that's trying to do something fairly sophisticated in its output, it's kind of being the the brain, uh, it, and then, uh. Historically, we've had these chatbots and there wasn't a lot. It was more of the oracle was sitting there and you gave it some input and it gave you some output. And um, we weren't, you know, it was still like, Okay, I'm having a conversation, but we're the the big change in the last year has been agentics.

**[0:10:56] Chris Benson:** Uh, and uh, uh. That's where you have one or more agents that are, uh, that is a type of software that is a task, uh, to go do something out in the world. It might be on your system, it might be in a R, a robotics, uh, it could be almost anything that you want it to be. And you can have one or more that are working together and they are managed by that agentic harness that you that you referenced, which is a software system.

**[0:11:22] Chris Benson:** Uh, I've heard it described as an operating system for AI, where it is it is accessing the models that you're using, it has it is managing the agents and orchestrating what they are doing and how they're working on that. And so it kind of gives you a body of software that you're able to use, uh, to get prod- productivity done, uh, to get productive tasks done. And so, um, if you th- you know, agents are smaller bits of that that are in the harness, uh.

**[0:11:51] Chris Benson:** But it's all just software, including the model at the end of the day, and that's important to keep. It can some of these can require sophisticated hardware, but at the end of the day, the model, the agent, the harness, all these are just, uh, s, you know, different types of software that work together,

**[0:12:07] Daniel Whitenack:** Yeah, may maybe one way to think about it is you have that one model, which is a function, uh, an input output. And the amount of code and software surrounding that to do various things can can vary, right? So initially, in the very Earl- or or in the earlier days and earlier days, I'm meaning, you know, a couple years ago or or whenever it was, um, we. We had software that was basically just a chat interface, a thin chat interface.

**[0:12:39] Daniel Whitenack:** And to your point, there wasn't a lot of business logic between that chat interface and the model. Basically, what you put into the chat interface, that was almost exactly what went into the model, and what came out of the model was almost exactly what came back to you. Now, the amount of business logic that you can apply on top of that model can vary widely, right? And, uh, the but that would be kind of like AI embedded in your software.

**[0:13:05] Daniel Whitenack:** I that's right, think the distinction of the agent, which is very interesting, is an w- What I would consider an AI application or an AI feature within your software is really something that is meant to have a one to one interaction with you as a user of that software. So a chat interface, I go in and I type something in and I expect a response back, and then I type something in. It's turn-based. Or maybe there's things like in, uh, financial software, where you go in and there's an AI model that makes a prediction for, uh, a future, you know, cash flow, uh, variance, or or something like that, right?

**[0:13:46] Daniel Whitenack:** And so those are AI features. What I think is different about the agent again, it, to your point, is not fundamentally different in terms of, uh, how you it's still software that surrounds the model, right? But it the w- What the software is meant to do is to accomplish a goal, and in many cases, to operate with autonomy. And that's where I think the agent is different. So it's tied into systems that are within your company or within your context.

**[0:14:17] Daniel Whitenack:** So if it's a personal agent, maybe it's tied into your email, it's tied into your WhatsApp, it's tied into your Google Drive, it's tied into your messaging interfaces, it's tied into your calendar, right? If it's a company agent, maybe it's tied into your net suite, and it's tied into your workday, and it's tied into your transactional databases, et cetera. And you express an outcome to that agent, and you or or en- encode a goal with that agent, and it operates at least with some level of autonomy without a back and forth with a human to accomplish that outcome, or or maintain that goal, or accomplish that goal.

**[0:14:58] Daniel Whitenack:** I I think that's that's really a a key differentiator when it comes to an AI application or an AI feature versus an agent. It's more important than ever for businesses to move super fast with updates to their website, to their landing pages, to their SEO, to their AEO content. That is surfacing things in answer engines, and it's just not sustainable when a new landing page or website update turns into a pile of tickets and handoffs. Our our partner framer actually makes that process move so much faster.

**[0:15:39] Daniel Whitenack:** Teams can collaborate in real time, iterate on the same pages and publish instantly. Thousands of businesses already use Framer as a pro website builder, and I'd really encourage you to check them out. You can learn more and get more out of your site from a Framer specialist or get started building free today at framer dot com. Slash practical AI for thirty off a Framer Pro annual plan. That's framer dot com slash practical AI for thirty off framer dot com slash practical AI.

**[0:16:16] Daniel Whitenack:** Rules and restrictions may apply. Well, Chris, uh, the there, that was, I think, a great, a a great set of hopefully, hopefully. We didn't muddy the water with any I don't think with any of so. our definitions of of things, but I think that was that was some good context setting when it comes to some of the terms that we're about to to use. Did did we miss anything? And what I, what is your thought? W- What, what ki- kinds of questions have been on your mind recently in terms of clarification?

**[0:16:46] Chris Benson:** yeah, I think you set the stage really well. And I think, um, I, I. In terms of like that connection that the agent has to, kind of real world things that y- that a person's company, or it could be in their house. It doesn't have to be a business. There's all sorts of different contexts here. But but you know, the you're, you're giving those agents access to things that matter is the bottom line, and you're giving them tasks to do on how to do that.

**[0:17:11] Chris Benson:** And you were kind of addressing single agent, uh, moment. But the the next level above that is to start having multiple agents, um, that each have purposes, you know, to a task level, and then they are starting to interact. And when you get to that level, um, you really start. Uh, that's where the kind of the the autonomy of what agents are, or the potential of what agents can do, uh, is really realized. Because instead of you as a human kind of being the recipient of each thing there, and you know that, um, that the agent is doing, and it's really kind of almost a an assistant to you.

**[0:17:48] Chris Benson:** You can actually assign agents to go out and work together to accomplish multiple tasks in a complex environment, and where different agents have different, uh, purposes and expertise. Um, and so they are gonna be able to, uh, to kind of focus with that, and then, uh, after that, they're going to be able to, uh, interact with each other and get this work done.

**[0:18:13] Daniel Whitenack:** If you're listening to the Practical AI podcast, I'm guessing that you value practicality, not just the hype around AI, which is why I think you should check out the Midwest AI Summit. This is an amazing event. I'm gonna be there this year. It's happening October 15th in Indianapolis. This is an event like no other I've been to. There's actually an AI engineering lounge where, for free, you can go up and get expert advice from practitioners and get feedback on your architecture, your design, your agentic harness.

**[0:18:47] Daniel Whitenack:** Whatever you're looking at, you can get feedback on in real time, in between amazing speakers that are on the main stage. So don't miss this event again October 15th in Indianapolis. And you can use the code Practical AI 20 for 20% off. So go to Midwest AI Summit dot com and grab your ticket today. Use code Practical AI 20 for twenty off Midwest AI Summit dot com.

**[0:19:15] Chris Benson:** So that was a great explanation of getting us kind of, you know, from from basic models into kind of the world of human model of chatbot and into the world of agents. And what's different there if you introduce the agent into the process? Um, after that, you can start multiplying out, you know, these resources, so you by by virtue of having that agent, you have the harness that the agent is working in. It might be fairly trivial if you're just barely getting into it, but you can actually start expanding those and looking at what different types of more complicated tasks you wanna address.

**[0:19:54] Chris Benson:** Where you have, uh, not just one agent, but you might have multiple agents that are each addressing different kind of specialties, if you will, um, with differen- And potentially each of those agents has access to different resources, um, within maybe a larger enterprise, where not one agent can go everywhere and do everything, but they have to start interacting to get job done. And you mentioned prior to the break that, you know this, that as you get agents, you start moving into some level of autonomy, um, and permissions in terms of what they can do.

**[0:20:28] Chris Benson:** And so, uh, as you get multiple agents, uh, and you start deciding what agent roles are going to be and how those agents should interact and what they have access to. Um, then you can really start getting a lot of productive work done, uh, through your your group of agents at this point, um, that is, that is largely autonomous, uh. And you can give the system instructions that have specific outcomes that you're looking for, and and you can either you can, you know, encode those into that.

**[0:21:01] Chris Benson:** And they're gonna go and work within the parameters, you know, as we've learned with some of the security breaches lately, within the parameters that you set, assuming that you set them effectively enough at this point. Or you have agents doing that for you. Then they're gonna go do a certain number of tasks that are that you've identified as productive, with specific outcomes tied to those. Um, and you can get lots of of, you know, really productive, uh, work done this way.

**[0:21:29] Chris Benson:** And this is kind of where things are right now. Um, the one other thing I'll throw in, uh, before throwing it back over to you, Daniel, is that, um, along with multiple agents, people are starting to recognize that different types of models can contribute. You can have larger models for some things, smaller models for others, some are more specialized, some are more general. And depending on what you need to get done, you can start having an architecture with multiple models and many agents, uh, doing this.

**[0:21:57] Chris Benson:** Uh, what, what's your? what's your experience so far in this realm?

**[0:22:00] Daniel Whitenack:** I, I. Well, I, I might have a follow up question for you, cause I know you obviously spend a lot of time thinking about autonomy. And even, you know, fleets or or swarms of of agents, or, um, a autonomous entities. But I I, I'm wondering. People may be very familiar with the uh, you know, some listeners may have used AI systems a lot. Where the the it was that turn-based thing, like, Oh, I'm gonna, I'm gonna.

**[0:22:31] Daniel Whitenack:** Always have my chatgpt tab up or my Claude tab, or, you know, Claude code. And there's a back and forth going on with with those things. What c- could you do? You have any examples in your mind of things that might be impossible in that mode for, like, a single agent, maybe interacting with a single human? But that would be possible for a fleet or a swarm of agents that's operating, you know, more autonomously in the background.

**[0:23:04] Daniel Whitenack:** Y- Do you have any examples that come to mind, or or things that you can think of that that people could grasp onto con- concretely?

**[0:23:11] Chris Benson:** Yeah, I'll try to do that. Uh, a little tricky, cause of the the space that I work in. Um, but I'm gonna try to genericize this a little bit. The uh, the you mentioned the notion of fleets and swarms of of agents, and um, and those are actually two different things. Uh, swarm is usually a fleet, but a fleet is not necessarily a swarm. Um, and I don't wanna get us off into an autonomy conversation about y- there's what that is.

**[0:23:33] Daniel Whitenack:** a changelog episode about a, a, a about this, so maybe we'll link that in the show notes.

**[0:23:40] Chris Benson:** uh, uh, we could do that. So, um, so when you have a lot of agents and you're and you're, uh, you are re. You are giving them each certain capabilities and access to certain things. Then you can start thinking about, um, situations that you might have been afraid to use, um, uh, systems in before. In places where you definitely want a person controlling everything because of the time sensitivity, um, or or care that they have to, that they have to bring to bear.

**[0:24:09] Chris Benson:** And so, um, we talked about cybersecurity in last week's episode, um, and and that's a good place where, um, if you're under an attack and your organization is needing to defend, uh, it might very well be. There are many agents, as as we we saw in the in the last episode, uh, that were involved, but you might. Your defense against that as well might be many agents that are operating, uh, either as a as a bunch of kind of individuals with some crossover, or a swarm where there's a lot of crossover communication and task sharing between them.

**[0:24:43] Chris Benson:** Um, because your humans that, uh, that suddenly your systems are under attack. Maybe it starts with a denial of service or something that you notice, but then you notice that there are breaches in different systems. And it's happening very rapidly in 2026 compared to even 2025. And so at this point, um, it's very difficult for the humans to keep up with that, as we talked about last week. And so you can have agents assigned to operate your cybersecurity defenses in different aspects of your organization, and each one to be responsible for different parts of that.

**[0:25:17] Chris Benson:** So that as breaches occur, you have different agents responsible for that. So that is, that is kind of a broad use case, uh, for the cause. Everyone needs some level of cyber in today's world, and if you're not addressing cyber, you're gonna feel pain at some point. So that's that's something that goes across industries. Uh, another quick example, uh, without diving too far into it, is that, um, though most people are not used to pervasive robotics everywhere now, uh, at least at least in most Western countries, they're not.

**[0:25:50] Chris Benson:** Um, that's we're right on the cusp of that. And as, and and as production, uh, capabilities are getting less and less expensive to do it, and the and the capabilities of what different types of robots and drones can do, we're seeing more of that. And I mean, case in point, you know, Walmart advertises drone delivery, as does Amazon, and you know, we're seeing more of those. You'll start to see robots popping up around. I know you have already done that, uh, near your house.

**[0:26:14] Chris Benson:** I remember talking about that, and this will happen more and more. And agents, uh, in robotics are incredibly useful because there's a lot of things that have to happen. Both from a functional standpoint, from, uh, well, what I would call a mission standpoint, but you could call it an objective standpoint, and from a safety standpoint. In terms of not accidentally doing harm out there in the world, which is easy because you have a physical device that's doing something and interacting with real world things, including people.

**[0:26:41] Chris Benson:** Um, and so, um, there's a whole set of considerations. And having agents that are designed to handle those tasks and having the resources on board. Uh, so that they don't lose contact with it doesn't. They don't have to worry about losing contact with a cloud provider. Something is is is becoming increasingly important, and we'll see more and more of that. So these are all, you know, one is kind of out in the world with us, and one is is cyber.

**[0:27:04] Chris Benson:** But there's a lot of places that I'll bet people can find in their own industry, where where the necessities may be real time or near real time, or just such complexity that having lots of agents interoperating together is gonna become necessary.

**[0:27:19] Daniel Whitenack:** Yeah, and I think you can start to think through some of those, uh, th- those flows in your own business as well. Like, if you think in manufacturing, for example, you may get notifications from some of your suppliers that then create a trigger for a first agent to actually process that notification and realize, Oh, we've got a vendor problem here. And then maybe there's another agent that does research for, you know, vendors to replace. Your raw materials that are now at risk within your supply chain, and that's automatically triggered and just happens.

**[0:27:59] Daniel Whitenack:** And then that generates something that is then output into, uh, orders or or draft information that goes into NetSuite and interacts with that system. And all of these things could be happening in the background. One of the things that I'm experimenting with Chris, and I forget if I our our listeners will will forgive me if I'm repeating a s an example that I already used. But I've been experimenting with this, uh, with this example in workshops and discussions that I've been having.

**[0:28:29] Daniel Whitenack:** Um, over the last, over the last few years, I've got into to motor racing quite a bit, not myself, uh, doing any of it, but just as a fan. And I remember always, throughout my career, everyone would use the example of the F1 pit stop as in kind of like leadership discussions and team discussions about how a team should work together. Well, right? And and often how that would go is like you'd look at how the pit stop evolved over time from like, whatever 20 s down to 2 s.

**[0:29:01] Daniel Whitenack:** And then you realize, well, the the way that they're able to do that is they literally have a single person do do one a very specific thing, task and nothing else. Like, move the tire from here to here, and that's all. That's all they do, and everybody coordinates well with one another. And it used to be in those leadership or team building discussions, they would talk about, you know, how the team could work together well when it was people.

**[0:29:27] Daniel Whitenack:** Now, I think what's interesting is every individual human working in a business like me as a CEO, or you as a research engineer, or whoever that is. We're actually elevated from doing a a single task really well in our position to we're actually elevated more to the team principal or strategist level of the F1 team. And now each person can actually have a high performing team of agents under them, each of which do various tasks very well.

**[0:30:02] Daniel Whitenack:** And you're now thinking more at the strategic or outcome-based level, like that team principal, like that strategist, right? So I think, actually, if you think about it this way, it does a couple things. One is it. It kind of empowers people to maybe think about all of those elements in their job, where they might have this team of of agents to to help them and actually operate on their behalf. It helps you understand, maybe, like, the level at which you should be thinking.

**[0:30:32] Daniel Whitenack:** And I think it actually elevates the human in the situation because now they're able to actually apply their domain knowledge, their strategy, their their kind of outcome-based thinking, uh, towards towards the goals that matter for the for the business. Now that's the more positive side of things. Obviously, you could argue, well, not everybody's gonna be operating at that STR strategy outcome level. Then how many like strategists do you need? And do you need less humans? And what is everybody's job?

**[0:31:03] Daniel Whitenack:** And there's a whole series of questions that could come from that, right? But I find it, uh, it, it at least to be a useful, you know, a u- a useful tool to think about how this might operate in terms of your, your immediate surroundings and your business.

**[0:31:19] Chris Benson:** that's right, and and it's it's important to recognize that this is, you know, one of the reasons, like I said at the beginning of the episode, that we're doing some of these episodes addressing some of these, uh, are. This is happening really fast in terms of the availability, and, um, but so is the n- the understanding. Yeah, that that we really wanna bring all people along so everyone can benefit from this capability. Um, but this is changing, uh, this is changing, you know, global economics gradually, and so it's not something it.

**[0:31:53] Chris Benson:** It can seem very esoteric to a lot of people who aren't familiar with, uh, where agentics are today. If I sit down with my extended family at a family outing, um, and and try to have that. And they're like, aware that something's happening, but they don't know it super well. E- Uh, and so we'll, we'll talk a little bit about that. But this is the kind of thing you really wanna get on top of, cause it's moving quickly and it's creating huge competitive advantage in some organizations, and while others, it's diminishing rapidly.

**[0:32:22] Chris Benson:** Um, so I just wanted to draw the importance of kind of of kind of grabbing what the agentic economy looks like as early as possible.

**[0:32:30] Daniel Whitenack:** As you're listening to this episode, you can tell that there's a lot to think about, there's a lot to know when you're scaling up your digital workforce of agents. I'm leading a company called Prediction Guard, which provides an AI control plane to help you actually scale your digital workforce in a governed, controlled way. We help you look at your supply chain, including models, mCP servers, making sure that you know what's in your supply chain, that you can track that.

**[0:33:00] Daniel Whitenack:** We institute runtime governance over all agent interactions, controlling that behavior as agents operating in your environment, and we provide observability around what those agents are doing and any policies that they're violating. So you can be sure that your agents are under control and being scaled in a way that's governed. Please check us out at prediction guard dot com and book a call with myself and my team this week to talk through how you can scale your digital workforce while applying zero trust and maintaining control.

**[0:33:35] Daniel Whitenack:** That's prediction guard dot com. Well, Chris, I, I, uh, of course, uh, w- you know, it's probably no secret to our listeners. We're always using AI to help us, uh, do these interviews or do research or or whatever. One of the things that that I did in prepping for this conversation, Chris, was just get get a few prompts and maybe questions that people are wrestling with right now for for us to consider. And so, m I, I don't know how many of these we'll get through, but I thought that they were interesting prompts to to us.

**[0:34:10] Daniel Whitenack:** Which, uh, one of one of them, which I thought was interesting, is is multi-agent architectures, so these fleets, swarms, whatever you consider. Is that really the future, or are we just compensating for limitations that better models will remove, meaning models seemingly will get better? So do we really need multi-agent systems, or is the better model just gonna be able to solve those problems?

**[0:34:38] Chris Benson:** so I think it's an apples and oranges thing. And and that is, you know, we we've talked A third, about,

**[0:34:44] Daniel Whitenack:** a third route?

**[0:34:45] Chris Benson:** yeah, uh, it's, uh, there you go, um, it, it. So I think there will be, there will be. Models will get better, and there may be some models that can do quite a bit for you without deploying multiple models in your system. Depends on the nature of the system, obviously. Um, there will be other systems, especially, you know, as you're as you're moving away from where you can host giant frontier models of the future, uh, to where your.

**[0:35:10] Chris Benson:** Your resources are more limited, where I think you're going to be dealing with lots of smaller models that are more specific on tasks. Um, and at the end of the day, though, uh, you're still talking about those are functions to your. At the beginning of the episode, you gave us a great way of looking, uh, about, uh, understanding what models are, and at the end of the day, those are functions and they gotta do stuff. And so I personally, uh, am quite convinced that, uh, multi-agent architectures, uh, not only are where things are going, I think they will be absolutely and completely pervasive in the future.

**[0:35:43] Chris Benson:** Uh, to the point where, uh, we won't want to talk about them at all because they'll just be so built into the fabric. And so I'm actually actively writing a book on this topic, so, um, among other things. So, yes, we are going into a world where this is not going away. Uh, I'll leave it there.

**[0:36:00] Daniel Whitenack:** Yeah, I I think that's a great perspective. And another prompt here, uh, not well, another one I I found interesting is, do we think we're moving towards a future of more interchangeable models in these agents? Or are we gonna see a future where most things are very vertically integrated agent stacks? And just to kind of highlight for people maybe that don't, aren't, aren't as familiar with some of these dynamics happening right now. There's one side of things which maybe is lives in the world of sometimes more open, not always open, though.

**[0:36:42] Daniel Whitenack:** U- open source agent, um, harnesses, but things like, uh, for example, LangGraph or Pydantic agents, or Hermes agent, or Open Claw, et cetera, et cetera. Name, name your favorite one, uh, framework or agent harness. And within these frameworks or agent harnesses, you can interchange the model, like the brain, as as you mentioned, so that could be a closed brain model, it could be an open one, you could use multiple, uh, et cetera, et cetera. But there there's a distinction there between that harness and framework layer and the the model layer.

**[0:37:21] Daniel Whitenack:** Then then you can look at other stacks, like maybe you look more towards, uh, inbuilt agent features in Anthropic or OpenAI's platform, but also others, whether it's like IBM Watson agent thingy, I I forget what it's called, or, you know, the hyperscalers. Uh, AWS has that agent core thing, I think, like others. These are very opinionated approaches, right, to say, hey, we're gonna do all the different things of the stack, we're gonna do them in an opinionated way.

**[0:37:54] Daniel Whitenack:** We're vertically integrated. We maybe even have the compute, we have the models, we have the agent harness, we have the the uh, interface, we have the console. Just like, take everything in our vertical stack, right? So we we do see the both of those dynamics very clearly in the market right now. So what is what is that prompt in your mind?

**[0:38:15] Chris Benson:** Well, I I think we're gonna see all of them, uh, because because we're already seeing all of them. And I don't think that this is a winner or loser side, I don't think one falls away. I think that they fill different needs, um, for different purposes. And um, you know, I in in my own, in my own life, I use all of the above. I u- there, there are some, it's. It really depends on what my need is and what is the sensible architecture and strategy around what it is I'm building.

**[0:38:42] Chris Benson:** And some of that is using, uh, closed frontier models that you would find at, uh, you know, whether it's the the geminis and the clauds and the, you know, and the open ai models, um, and and there's a there's a place for them, and then others, uh, you know, I might be going for the gemmas and the other smaller models. That that where you can run inference on a device, for instance, uh, or or maybe an application that just doesn't need.

**[0:39:10] Chris Benson:** I mean, why spend the money on on closed model APIs when you can, uh, you can do a lot of what they do, uh, at much, much, much reduced cost and host it yourself and get the economics, uh, on your side. So it's an interesting question that you pose, because, yes, all of them are gonna happen, but there's such an economic, um, incentive for all the different versions, depending on what it is that you need. Um, and and I think we're what we're starting to see now in in recent days is really a drive.

**[0:39:42] Chris Benson:** Like we saw the explosion of usage, you know, we were talking about this a few months ago on the show, um, about, you know, just, you know, token, uh, token maxing. And and that is already like, K- it came and it went because suddenly the the cost of that caught up to people. So now what we're seeing is we're seeing people going, how do we get there at the most affordable cost for what it is that we need to accomplish?

**[0:40:06] Chris Benson:** Which I think is a lot more sensible. Um, I'm glad we're past the token maxing moment into more of a sensible, like, let's think about this architecturally and strategically, going forward on what we need to do. So, yes, all of the above, with a sensible planning process, uh, guiding you,

**[0:40:24] Daniel Whitenack:** yeah. And the, uh, you know, we when we have agents that could be operating autonomously for hours or days or months. Obviously, then that makes a totally different impact to your your costs of operating that type of system than just a chat back and forth, right? So that that's certainly a part of it. I I do actually think, you know, you mentioned that maybe both of these will exist, but for different reasons in my mind. Just to draw out one of those reasons right now would be if you look at, you know, most businesses utilize some sort of drive workspace, document ecosystem, right?

**[0:41:08] Daniel Whitenack:** mostly Microsoft or Google, um, although there's others, right? So if you're primarily using Google Drive and Google Docs, then you're automatically gonna get one of these very opinionated AI integrations across those various apps. In a way that is incredibly integrated into that, and that's Gemini, but you would not be able to use that opinionated Gemini stack within Microsoft 360 and documents and SharePoint, et cetera. Because the one that is integrated in a very opinionated way, there is, you know, Copilot, et cetera, uh, across Microsoft applications.

**[0:41:46] Daniel Whitenack:** And so there's a suite of applications, and obviously they have their opinionated, very tightly integrated way of doing those things. However, whether you're a Google shop or a Microsoft shop, if you're creating this, what I think of as your digital workforce of agents, right? You're not gonna have that digital workforce of agents be run by the little widget in your Google Doc Gemini widget, right? A And there's a whole variety of reasons that you wouldn't want to do that.

**[0:42:17] Daniel Whitenack:** One one of them being, you know, just maybe cost and the the h- how much it would take to build that out. But I think to use a different, um, may I? I? I'm trying to think through all these examples right now, cause I love, uh, I, I do love teaching, and I'm I'm trying to think of better examples as as I teach more of this agentic stuff. But I think an example that I could use here is that may connect with some people is that most businesses, uh, or many businesses, at some point in the life of their business, engage with a consulting firm, right?

**[0:42:55] Daniel Whitenack:** Whether that's the big consulting firms like, uh, McKenzie or an Accenture, or something like that, or midsize or smaller, or whatever. Those firms have highly, highly opinionated takes on how to do things, they have their own systems, right? If you engage with them, they're gonna do things in a very specific way to accomplish your goals. But m most businesses do not say we're going to staff our entire business and workforce through McKenzie. One reason is no one would have that much money to to do that because it would be an ungodly amount of money.

**[0:43:31] Daniel Whitenack:** And two, like, w- what, what? What is your worth as a busi- like, what is your? What value are you providing other than the business entity? A And that's where you can really build up the value of your business and your agility as a business over time. Um, th- both from an IP perspective, but also your control, your flexibility, your ability to pivot. And under economic pressures, or under, uh, marketplace pressures or pivots, or that sort of thing, you wanna have control of that, of that main part of your workforce, right?

**[0:44:06] Daniel Whitenack:** And I so I think that control and longevity element of being able to build agents within your digital workforce in a way that is definitely more vendor agnostic, less tied into the strict opinion of a vertically integrated agent stack, I think does have advantages. Now. Will everybody be able to do that, or should everybody do that? I'm not necessarily saying that, but I I'm, eh? That's one example that I've been trying to think through as I'm as I'm thinking of what, what some of these dynamics are.

**[0:44:40] Chris Benson:** I I think that's a great insight that you had, and that is the, you know, the the having specific vertical stacks, cause those are already, you know, developing. You named a couple of them, uh, there are more out there, uh, that I, a, and they own a certain amount of the business. But I think the the the realities of business is that the ability to move across those verticals, uh, as well as into your own domain of your o- own, uh, you know, uh, the the models and agents that you are managing on behalf of your business that are very specific to your business is a reality that everyone's facing at this point.

**[0:45:15] Chris Benson:** And every aspect, uh, of business is going to have, uh, both the different types of models and the different types of agents having to do those interactions. So I think your point is really well founded that you can't think about, uh, just going and, uh, getting. Like, I'm gonna be an all in on Google, or I'm gonna be all in on Microsoft or some stack like that. Because I don't think in most cases it's gonna meet your full need.

**[0:45:42] Chris Benson:** I'm sure those companies will disagree with that, but um, the realities of life are too complex. And so, uh, the key is to be, I think I think this is the role over the next few years. You know, there's a lot of people out there on LinkedIn calling themselves AI strategists and such. Um, and and if you, if you fancy yourself with that title, then you've gotta figure out the how to apply these different resources to the needs of the business, um, given this very complex ecosystem.

**[0:46:11] Chris Benson:** And so I think that that is where things are going. Um, and it's we, I think it's, it's messy, uh, so

**[0:46:19] Daniel Whitenack:** yeah, well, I I do think that, well, life is messy, so, uh, it's everything that we deal with in the future is gonna be messy, I'm sure. But there there is some, uh, there's definitely some patterns emerging that are helpful for people. I hope we've been able to highlight a couple of those today, or even just clarify some terms as we as we kind of get towards the end here, Chris. I think one question that might be interesting to just wrap us up here is how how should people evaluate their agentic AI efforts as they dig in and try to accomplish things with agents?

**[0:46:58] Daniel Whitenack:** Is that is that by, you know, uh, h- how do they? how do they measure s measure success? Maybe I in terms of, uh, successful agentic AI adoption, or that sort of thing? Any any thoughts?

**[0:47:12] Chris Benson:** I, I would start small, uh, with with the expectation of rapid iteration, um, and and just experiment your way into what's working. Um, be thoughtful about, you know, as we've talked about all these different options that are out there, be thoughtful about what makes sense. Um, uh, I see a lot of companies, uh, uh, that are tied very tightly into specific vendors or vertical integrations, uh, and all it takes is one of those vertical integrations to change something.

**[0:47:42] Chris Benson:** Or to take, uh, something that may be a high value thing that your company does. And suddenly they provide that and it can kind of crash a whole business. So, like, there's there are times when looking at verticals might make sense, but maybe you should be thinking about how you can do it in a in a different non, uh, you know, not bound to a particular vendor. So that your business survives, um, as as people are negotiating these things, so start small, experiment your way and have a backup plan, uh, a, and maybe several backup plans as you're experimenting.

**[0:48:16] Chris Benson:** So that when you cause, you will run into problems and you will run into limitations, and you'll run into models. I run into models all the time, uh, on the frontier side, that won't let me do things I wanna do, uh, because I'm trying to do novel things a lot of the time. And I hit guardrails all over the place. And so, um, when I'm doing, and I'll, I'll. The other night, I was sitting here at home, working on a side project of mine, and that happened.

**[0:48:39] Chris Benson:** And I'm like, okay, well, fable, fable won't help me, and I started downgrading. Have backup plans, figure out what will work for what you're trying to accomplish, and, uh, and see.

**[0:48:49] Daniel Whitenack:** yeah, great, great way to close it out. Appreciate your insights, Chris, and it was a fun discussion today. I hope our e- everyone listening engages with us throughout the throughout the week until next episode. Um, we are on YouTube now, along with all the other socials, so go check us out. You you may or may not want to know what we look like, but you can now if you want to. And so, yeah, we'll, we'll talk to you again soon, Chris.

**[0:49:15] Daniel Whitenack:** Have a good one.