So What? How to Get Started with Paperclip AI, the AI Agency Software

Katie Robbert – 00:00

Well, hey, everyone. Happy Thursday. Welcome to So What, the Marketing Analytics and Insights live show. I’m Katie, joined by Chris. John is traveling, but George is here.

Christopher Penn – 00:40

Yeah, John’s out at Salesforce Connections.

Katie Robbert – 00:42

He is, yeah. So if you’re watching this as of today, which maybe you are, go say hi to John. He’s in Chicago. And if you miss him, then come say hi to us at TrustInsights.ai today. We are talking about how to get started with Paperclip AI, the AI agency software. Earlier this week on the podcast, which you can listen to at the TrustInsights.ai TI podcast, we talked about what Paperclip AI is. Today we’re going to talk about setting it up and actually using it. At a high level, Chris, my understanding is that Paperclip AI is another generative AI piece of software.

The idea with Paperclip is that you can basically create your AI agency, but you can create these versions of people who can actually talk to each other and remember the context of conversation. We’re sort of equating it to a company that isn’t siloed and actually shares conversations and resources, versus the way that a lot of large language models operate today. That does come across as a siloed company, where everyone’s in their lane and nobody shares information unless you really force them in the room together to do that. Is that, at a very high level, a good way to think about it?

Christopher Penn – 02:21

Here’s the thing about Paperclip, because I’ve been thinking about this as we talked on Monday about the podcast. When we did the podcast, Paperclip billed it as the agentic control plane for autonomous AI companies, which is a lot of words.

Katie Robbert – 02:38

That is a lot of words. That’s a lot of buzz.

Christopher Penn – 02:42

It is a lot of buzz. So here’s the buzz-free version that, Katie, I think you’ll wrap your brain around immediately. It’s Jira for AI.

Katie Robbert – 02:54

Got it.

Christopher Penn – 02:55

Yes.

Katie Robbert – 02:56

So for those who may not be familiar, describe Jira.

Christopher Penn – 03:03

Jira is project management software where you set up projects, different things people are working on in your company, and it could be any department. Coding, development, UI, UX, customer service, whatever. People open tickets and assign them to other people in other departments. A customer just called in, and this thing’s not working. You assign that to the development team, it routes it to a developer, the developer works on it and says they solved this problem for this customer, and so on and so forth.

Katie Robbert – 03:32

So at a high level, when we talk about things like Jira, it is project management software. Project management software a lot of times is meant to be the system of record for tasks, deliverables, documents, and flow. Chris, the flow you just described in terms of a ticket going through people, it’s that ticket that allows for the collaboration between you and me and the developer and someone else, because that ticket has all of the information that we all need, and we take from it what we need. I think that’s an interesting way to think about it, which is not the way that I would assume AI agency software is positioned. So should we get into it? What are we doing?

Christopher Penn – 04:23

Yeah, we can get into it. We’re not going to talk about setting it up, because setting it up frankly is a five-hour ordeal.

Katie Robbert – 04:30

Yeah, we’re not going to do that. Not even with the Benny Hill sped up music. At a very high level.

Christopher Penn – 04:38

Yes, at a very high level, you should run it on a VPS or a dedicated box that has to have access to agentic software of some kind. It could be Claude Code, it can be open claw, Hermes Agent, anything. In fact, we did a show on Hermes Agent two weeks ago on the live stream.

Katie Robbert – 04:54

Like a month ago, Chris. But if you’re interested in watching that, you can go to TrustInsights.ai/youtube and go to the So What Livestream playlist. Things move pretty quickly, so that was about a month ago.

Christopher Penn – 05:06

I’ve lost all track of time. Essentially, if you have an agentic tool that can receive a heartbeat—meaning this software can ping it and prompt it—you can use it with this. Those are the requirements to set up Paperclip. I strongly recommend, like we said with Hermes Agent, that you have it on its own little sandbox so that it is not on production systems where other important stuff lives. Please do not run this on your main computer and give it blanket access to everything. You will be sorry and not McKinnon.

Katie Robbert – 05:42

That’s in general just good best practices for software development, which to be fair, a lot of these large language models are software development.

Christopher Penn – 05:51

Exactly.

Katie Robbert – 05:51

The best practice. Yeah.

Christopher Penn – 05:53

What we see in our agentic control plane is four sections. There’s the work itself, there’s projects, agents, which are your virtual people, and then the company, the org chart, the skills, etc. If you think of it like Jira would have projects or departments, Jira would have people in it. Jira has a ticketing system, and then you have some general reporting to see what’s going on in our company in our project management system.

Katie Robbert – 06:27

Honestly, it’s very reminiscent of Microsoft Project connected with SharePoint. The way that Microsoft Project was the most effective was if you could get all of that company information into it. Then you could choose your agents, which are really just people, to say Chris is the chief data scientist. This is his hourly bill rate, these are his responsibilities, and these are the projects he already has. When you’re setting up a project, you have all of that background data to say, is Chris even available for the month of June?

Well, he already has five other projects assigned to him, and this is what it’s going to cost us to get those done. Can we add in a sixth one? Does he have a spare hour here or there, and does the budget of your project support it? I’m looking at this and you’re right, it is project management software.

Christopher Penn – 07:25

Exactly. Which makes it a lot easier to understand. It does take away the sexiness of all the expensive words, but it also means that if you know project management well, you know how to use Paperclip. There’s a bunch of stuff that’s not in here. There are no Gantt charts. There are no workflows like that because you don’t need it for AI agents. AI agents are assigned to go do things and are told to go do things, and they go off and do things. You don’t have to think about whether Hermes Agent has time to work on this today.

That’s irrelevant in this system as it stands today. With the caveat, there have been 12 major updates in the past two weeks to this. So this is a very fast-moving project.

Katie Robbert – 08:12

I don’t know, when I hear something like that, that to me is a huge red flag.

Christopher Penn – 08:17

But anyway, as we said on the podcast, this is not production ready.

Katie Robbert – 08:21

No, no. It’s a shiny object.

Christopher Penn – 08:26

It’s a shiny object and it’s a good idea. It’s a concept that you’re going to start seeing more and more of. Some of what Paperclip does, for example, is just built into Claude Code; it’s called dynamic workflows, where Claude sets up its own project management system internally to execute on a task. What’s different about that versus this is this has persistent memory, so it remembers what happened. Kind of like you would use Jira to say, what have you been doing all week, Claude? Whereas Claude says I’m doing things, and then you turn it on again and it says I have done nothing.

Katie Robbert – 09:03

It’s like having a whole team of people.

Christopher Penn – 09:05

It really is. First things first, as you do with project management software, we’re going to start a project. I’m going to name it Test. It was called Test Bed. You can configure it to connect to a GitHub repo if you want. If you’re doing things like software development or you want to have a repository where humans can look in on things that are centralized, that’s a great way to do it. And then if you have deadlines, you can put them in here, and so on and so forth. So I’m going to create that first project called Test. Next we’re going to create our first employee, which is just an agent.

Katie Robbert – 09:45

Sure.

Christopher Penn – 09:45

There’s a bunch of different ways to do this. Right now there is no CEO agent at all. There’s nothing in here at all, so this is kind of an interface bug.

Katie Robbert – 09:55

I was going to say that’s a weird thing.

Christopher Penn – 09:58

Yes, it’s a weird thing. If you set it up net new, you create a generic CEO agent up front, which is actually a terrible idea. Katie, do you want to review what you said about that on Monday?

Katie Robbert – 10:13

Yeah, absolutely. You can hear the full rant at the TrustInsights.ai TI podcast. My concern is that you’re setting up net new, and maybe you’ve never been a CEO or you don’t have a lot of experience being a CEO, so you’re just making it up. Or you’re grabbing CEO job descriptions from the internet, which are never accurate. It’s not to say you can’t come up with a good proxy for a CEO, but I do think it’s a huge risk to ask someone who is brand new to their career or has always worked in a different vertical to create a CEO.

If you don’t have a good working understanding of what a CEO does, then you could be making a lot of assumptions and inducing a lot of bias. You should first and foremost create an agent, but calling it a CEO is a risky move because it introduces things that may or may not be true.

Christopher Penn – 11:32

It would be equally appropriate to call it a project manager. If this is Jira, you’d have a PMO, and there’s a head of a PMO, and that would be just as good for this. I’m going to create the runtime manually, and we’re going to choose Hermes Agent. You can see all the different tools that are available to you. If you have something else, if you use open code, Cursor, or Claude Code, you can use them in here. I’m going to use Hermes Agent because I know it’s configured, and we’re going to call this Katie the CEO.

Here’s what we’re going to use. We have thinking effort. The model is the default, which is Minimax. Hermes is the system. Hermes is the local agent. There’s the run policy. There are no company skills set up yet, so I’m going to create the agent. These are the default skills. I’m going to add a blank skill, and the skill we’re going to call is Co-CEO.

Katie Robbert – 12:39

We’ve built that Co-CEO on numerous live streams. You can get that at the live stream playlist on our YouTube channel. The Co-CEO has been around for at least a year in our ecosystem.

Christopher Penn – 12:53

Exactly. Rather than me try to make up and invent what a CEO should be, we reuse what we’ve already done. We’ve talked on past live streams and podcasts about the five levels of AI. From level one, which is ChatGPT, to level two, which is gems—which is what the Co-CEO is—to level three, which is agent systems like Claude Code, to level four, systems like Hermes Agent, which are more autonomous, to level five. If you did your homework and you’ve been working up the ladder, not trying to skip ahead as we just did, we fished something out from level two, the Co-CEO gem or skill. I’m making sure that Co-CEO is installed, so now Katie the AI CEO is ready to work.

Katie Robbert – 13:51

Great.

Christopher Penn – 13:51

This one is the master agent that can then go and hire other agents if needed. We’re going to do a very simple test and open our first issue. Like Jira, this is exactly how work gets done, and the single unit of work is the issue. So let’s come up with a simple task. We’re going to create our first issue and call it Meeting Notes Idea. We’re going to make sure it’s assigned to Katie in the test project. This is where we specify what the task is. I’m going to say, write three plain English bullet points explaining what good meeting notes should include.

Something super simple so that we can test if this thing is actually working. We’ll give it a critical priority and make sure it is on to-do. I could create start dates, but I’m not going to. I will create the issue, and now what we should see is Katie the CEO has an indicator that it’s live, and on the dashboard up front, Katie the CEO is working on this very simple task. On the back end, you can’t see it here, but the system has started to route this, turned on Hermes Agent, and is burning away.

You can imagine how you could set up additional agents or have the CEO create additional agents. If you know your overall project plan, you could design really good project plans with multiple stages and decompose them. A lot of the way people are using this is just winging it. I’ve seen an example on the Paperclip subreddit saying, make me a new product that will earn a million dollars in recurring revenue. That’s just the task they’ve given, and then they fling it to the void and hope it works.

Katie Robbert – 16:13

I’ll go ahead and say it: that’s the wrong way to do it. Experimentation, research, development, and innovation are all valid. But when you bring it into a real company where real dollars and resources are involved and at stake, you really want to make sure you have a good, solid plan. A real challenge that we have is that we are a small company. We were talking on a previous meeting about doing promotions, advertising, and marketing for our academy. It’s something we all know needs to get done, but because we’re a small company, it always kind of falls to the wayside.

I could see saying, what of this could we automate? I already have a plan built. I’m at the point where I have to spend the time to copy and paste it into our social scheduling tool. That’s the part that I’m putting off because I don’t want to do it, but it has to get done. Could an agent in here do that for me?

Christopher Penn – 17:29

This is where someone like you is going to be a much better user of Paperclip than someone like me. You will have already thought out that plan, so you’re not relying on AI to wing it. Instead, you’re saying this is how you’re going to do this task. Even in this silly example, you can see here that it has spit back and said good meeting notes should include a clear agenda, key decisions made tied to the person responsible, and action items capturing discrete steps. We’ve now seen it do its first task, which was the whole point.

Katie Robbert – 18:12

So what else can it do? I look at that and I think that’s a good example, but did I need to set up Paperclip to get the answer to that question? If I already have the Co-CEO set up in Google Gemini and Claude, why wouldn’t I just ask the skill in one of those systems? Why would I need to set it up in Paperclip to do that? Granted, it’s just an example, but I feel like it’s helpful for us to wrap our heads around why use this versus what we already have.

Christopher Penn – 18:44

In a lot of ways, with things like dynamic workflows in Claude, you might not need this at all. A dynamic workflow, as long as you’ve built the infrastructure around it, can do exactly what this has just done without all the extra overhead. That’s why this is such an interesting space in AI. Once you see a project like this spring up and the community run with it, the big vendors realize they can implement something like that in their software. If you are just an early adopter, you can say you’re going to wait and see if Anthropic, OpenAI, or Google implement something like this, which they have.

What would be a complex project where you yourself, if you knew the shape of it, would need lots of sub-agents and multiple steps?

Katie Robbert – 20:02

One of the things that we talk about is live events for ourselves. We all collectively have a working knowledge of what goes into planning a live event. We’ve done enough workshops and attended events as speakers that we understand the amount of work that goes into putting on a good event. I’m not worried about the workshop itself because we could do those in our sleep, but it’s the event itself and the experience for the user. While we know what it entails, nobody on the team has time to execute against it.

If we are looking at the best practices from the lens of an event coordinator, what are the things that we don’t think of? Who on our team has time to do the research and actually talk to vendors? What I would love in a perfect world is to say here’s the budget, here’s the timeline, and here’s the amount of resources we can spend on this. Someone go make a plan that fits all of that criteria and just hand it back to me so I can approve it. That is the kind of thing that at a small company you just can’t get because everybody’s doing everything. That’s the kind of more complex project that would be great to get assistance on.

Christopher Penn – 21:55

To do something like that, we might put up a new project called Workshop Planning and planning a metro Boston area Trust Insights workshop. We’re going to hit create a project to create the plan. We’re going to make sure that our first issue on this should be essentially that miniature project plan that you just talked about. We’re going to assign this to the CEO, because the CEO is going to make hires as it thinks through what needs to happen. We’re going to be planning a workshop in the metro Boston area within 25 miles of downtown Boston, accessible by public transit for 25 to 50 people.

We need to know where we should have the workshop. We need to know what venues are available, what they cost, and if the cost is inclusive or exclusive of things like catering. We need to know what kind of public transit is in the area and how far out we need to book. We need to know what kind of Wi-Fi is on site for a workshop because this requires people with laptops and high-quality internet access. We need to know if the facilities have robust electrical power to make sure that attendees can plug in their laptops for doing AI work. We need to know what kind of insurance a venue might require and what hotels and lodging are nearby.

That’s all the requirements of what we want to know. Next, we have to give it instructions. Consider hiring research agents that will go out and research all these things, and project managers to consolidate and synthesize results. When you get conflicting results, the majority result wins. If you have five reports and three of the five have one fact, that is going to be the threshold for confirmation. Your output is going to be a CSV file of the venues with all the listed issues as columns. It will also be a markdown document with no tables that we can read and review. So already we’ve got a pretty decent issue here. Anything else that you would throw in here, Katie?

Katie Robbert – 24:51

If given more time, I would give it a budget. We could say our profit margin has to be at least a certain percentage.

Christopher Penn – 25:08

Let’s do it even better than that. Let’s have three different types of venues: low budget, middle budget, and high budget. High budget is going to have a cost of $250 to $500 per attendee. Mid-range budget is between $100 and $249 per attendee, and low budget will be $99 per attendee or less. You’ll need to factor this into your research and do the computations so that we have different options. All the venues you research will have that budget information, and if you can’t find budget information for a venue, you must exclude it. What else would you throw in?

Katie Robbert – 25:55

When you talk through a lot of the details of the user experience itself, resource and plan, those are the basics. Anyone who knows the amount of work that goes into finding all of that information and putting it together to make a singular decision knows that’s weeks, if not months, worth of work. That’s just to do the research, let alone call up a venue and say, do I have to supply my own forks? There’s a lot of information. We might also add in things about accessibility. How accessible to anyone should this venue be?

Does it have any kind of hearing, sight, or other impairment assistance on site, or do we have to provide those things? We need to make sure that we are thinking through those details, which is why if you’re just winging it, you’re doing it wrong.

Christopher Penn – 27:13

Exactly. Kimberly chimed in about the registration process, marketing, promotion, website updates, and whether there are speakers at this event.

Katie Robbert – 27:23

Kimberly, don’t get me started on how much work this is. This is why I’ve been putting it off and the team is very upset with me.

Christopher Penn – 27:30

This would be okay for a Jira ticket. This would not do okay in Paperclip, and here’s why: you do not get a chance to ask questions. One of the things that we teach in all of our workshops in AI is that one of the top prompts you should have is asking questions. Do you have enough information to successfully complete the task? You cannot do that in here.

Katie Robbert – 28:04

That makes me nervous.

Christopher Penn – 28:06

What we should do is flip over to a regular LLM first and use a prompt optimization framework to take the raw input, refine it, and do the asking questions here first before it goes into Paperclip. Once you kick that off, it’s going to kick off a bunch of agents and use a lot of tokens. It is better to spend more time planning and then kick it off once, knowing that what you put in is bulletproof, as opposed to realizing you forgot something after you basically burned 100 million tokens and have nothing to show for it.

Katie Robbert – 28:54

I cannot stress enough that all of the planning up front makes the execution that much more streamlined and seamless. It’s the same thing with software development. I’ve talked about the arguments I would get into with my software developers where they just wanted to push the buttons and build the thing. I would ask how they know what the thing is, and they would say they just know and will figure it out. This is why we never hit our deadlines. This feels like a really straightforward example, but there’s so much more.

We were just talking about the venue itself. We haven’t even thought about promoting it, which is what I started with in the previous example. We have all these courses in our academy that we don’t have time to promote, so how do we have time to suddenly promote a workshop? I personally can’t think of a single team right now that has people sitting around looking for more stuff to do.

Christopher Penn – 30:00

What we did was we put this through a prompt optimizer, and what it does is it aligns it to the 5P Framework by Trust Insights, which you can get at TrustInsights.ai. This is how we’ll get better performance out of Paperclip by making sure that it has some kind of framework. We can see what is the purpose of doing this research, who are the people, what is the process, our budget information, our triangulation phase, what’s the platform, the outputs that are required, and your definition of done. All of this now fits into a lovely thing, and then this is what goes into Paperclip. If you want to not burn 100 million tokens for no reason, you have to do the prompt optimization outside of it first.

Katie Robbert – 31:01

That is such a big pro tip. If people don’t take away anything else, take away that. Once you start it, it just goes until it’s done. If you have not given it a good defined definition of done, it may go forever. When it starts, can you check and see how it’s going, sort of the same way you can with Claude? It will show you all the steps, but are they going to be meaningful?

Christopher Penn – 31:37

As it works, you’ll be able to see the activity, the related work, and what it’s doing along the way. In the task itself and the project itself, you can select the task and hit pause on it in case it’s going off the rails.

Katie Robbert – 31:58

Which is interesting, provided you’re paying attention, because the whole idea behind these tools is that they work autonomously and you can go do something else. This says you have to be paying attention, which is definitely conflicting in terms of what it is meant to do.

Christopher Penn – 32:19

When you think about Jira, how much of Jira is hands-off?

Katie Robbert – 32:27

Based on how I’ve seen it used, none of it. It’s all very manual and painful.

Christopher Penn – 32:34

If you are the PM, zero percent of Jira is hands-off because you have to be monitoring what’s going on. If you’re the CEO, Jira is 100 percent hands-off because you have a PM who’s doing the thing. In this Paperclip system, the CEO is the PM. Theoretically it is hands-off, but in practicality, if your CEO PM is no good because you don’t know what a PM does, this system is just going to be an expensive disaster.

Katie Robbert – 33:10

A better positioning for when you’re setting up Paperclip would be to build your project manager instead of building your CEO.

Christopher Penn – 33:33

It just doesn’t go into the whole tech bro agency future narrative. If you say you just made Jira, you’re not going to get 100 million dollars in investor funds.

Katie Robbert – 33:45

If you say important person, and this is the project manager, not the CEO.

Christopher Penn – 33:50

Exactly.

Katie Robbert – 33:52

As we’re talking about designing these projects, could you as the human build in checkpoints? Could we say after you finish this part, stop and let me check the work along the way?

Christopher Penn – 34:12

Yes, instead of having it be one big plan as a single issue, you might break it down into 40 different issues. You might say you’re just going to research this, and you issue tickets out for each of those little things. Then you the human can either consolidate them or tell the consolidation agent to review tickets seven through 14 and come up with a consolidated report after the human has inspected the outcomes. Again, just like Jira.

Katie Robbert – 34:50

Depending on the nature of the project and how skilled you are with a system like this, that’s probably a good way to start so that you can see the expected output or where it goes sideways. It’s like learning any new system; you want to take those baby steps before throwing everything into it. Even if you have something that’s really well prompted, you probably still want to build in those checkpoints because all of this costs money and tokens. At some point you’ll wonder where this $6,000 bill came from.

Christopher Penn – 35:28

Or the half a million dollar bill that was in the news this past week.

Katie Robbert – 35:32

Nobody wants that. People like surprises—flowers, balloons, candy. A $6,000 bill or a half a million dollar bill is not a great surprise.

Christopher Penn – 35:43

The other two things in here are goals and routines. Goals don’t work yet.

Katie Robbert – 35:56

Moving on.

Christopher Penn – 35:57

What those are is either the high-level goal of a set of projects or from an architecture perspective, they’re kind of like system instructions. Routines would be scheduled tasks, so if you had your virtual thing assembling and sending you a status email every so often. That’s Paperclip as it is right now. Do you need to use this? Absolutely not. There’s nothing here that you can’t do in Claude Code or Hermes Agent. Arguably in a system like Claude Code you have more natural interfaces for stopping and reviewing.

OpenAI Codex just released a refresh, Google’s Anti Gravity just released a refresh, and all of these give you the ability to do the same thing but with a bit more work on your end of having that consolidated memory system. What makes this and things like Hermes Agent valuable is that consolidated memory, but you can get that with add-ons to the existing systems without having to buy or install anything new.

Katie Robbert – 37:18

Which is a great segue into our buddy Brian’s question. What do you think it will take or how long will it take for these agentic systems to become predictable and safe enough to run in larger organizations other than tech orgs?

Christopher Penn – 37:35

Predictable and safe are the most important words there. Right now, this system is highly unpredictable, so it’s kind of a non-starter, and the fact that I have to install this on a VPS means it’s not safe yet either. It goes back to the 5P Framework in a lot of ways. If the underlying agents don’t have good governance, then the Paperclip routing system on top of those agents isn’t going to enforce that.

Katie Robbert – 38:15

I was going to say something very similar: that the software is never the thing that you should assume is the safe or predictable thing. Software is volatile. It doesn’t matter how long the software has been around. If you don’t have good governance, guardrails, or requirements going into any project, regardless of the software, it can go sideways if you give it the wrong instructions. Don’t wait for these tools to put better guardrails in place. If you have a use case that requires you to use one of these tools, you the human make it predictable and safe. This is where we still need human in the loop and human oversight.

Christopher Penn – 39:21

On the predictable side, this goes back to level one stuff, which is in your prompt optimization framework. Whichever one you’re using, it has to be tuned for agentic work. That means you have to have a very clear priority system. A week ago in our Analytics for Marketers Slack group, someone shared a Claude prompt that just kept blowing up. I took a look at it and there were nine different top priorities. Like Syndrome in The Incredibles says, when everyone is super, no one is. When you put a prompt like that into a system like Paperclip or Hermes Agent, the model doesn’t know what the priority is, so it just wings it. A good prompt optimization framework should either ask you what the actual top priority is, or if you provide enough detail, it should be able to infer it. If you read our workshop planning prompt, factual correctness is probably the top priority. No matter what else the agent does, it has to get real data to help us plan a workshop.

Katie Robbert – 40:54

It’s interesting that something I have strongly built into my Claude workflows is saying, I didn’t have this so I took a guess, but you’re going to want to double check this because I don’t know that it’s correct. I will go check that, and if I provide the real information, it will go back and fix everything. That is such a big deal for any of these systems because AI hallucinates with authority and confidence. If you don’t know any better, you’re going to think it sounds right, and it might not be.

You want to make sure, especially if you’re making financial decisions like in our example of planning an in-person workshop, that it’s as accurate as possible. We didn’t add in that there is a cancellation clause that is clear and isn’t going to cost us more than we put in.

Christopher Penn – 42:22

Going back to the prompt optimization framework, this is where stuff from level one and level two still matters. If you’ve never done event planning, you should do a deep research project asking what can go wrong throwing a workshop. Have tools like Google Gemini come up with the laundry list of everything that’s going to go wrong. Do that deep research, build it, consolidate, and then that becomes part of the prompt optimization that you put into Paperclip. Even though Paperclip is a very cool idea of a control plane on top of autonomous agents, it is still all predicated on whether the data you’re giving it is any good.

Katie Robbert – 43:11

We talk about whether we need to use AI for this, and yes, it uses tokens and costs money, but scenario planning AI doesn’t get fatigued the way a person does. If you ask what is everything that could possibly go wrong, you should be seeing things like a natural disaster, a citywide power outage, or a bad Yelp review that takes down the whole workshop. Let the AI come up with all that stuff and problem solve for it. What do we do about it within the budget?

That way you don’t have to worry about that deep research for yourself. The human can then use judgment to go, yes, that’s something I should probably factor in. We probably don’t have to worry about Godzilla coming through downtown Boston, but we might have to worry about a hurricane, torrential downpours, or a strong windstorm that takes out the power.

Christopher Penn – 44:31

Exactly. A lot of what I see as really terrible use of these systems is insufficient planning up front. If you don’t think this stuff through with your level one and level two systems, you’re not going to get good results. The results you get out of them will be either vague or kind of weird. Which is fine if you’re just doing it for R&D, but if you’re doing it for production, take the time to work your way up the ladder so that when you hand off this stuff to your virtual company, you’ve got all these materials and skills built with all the background data it needs.

Katie Robbert – 45:17

It’s funny because the most common question is, will AI take my job? We keep saying, not if you’re using your brain, not if you are developing your critical thinking skills. We just spent an entire episode talking about how important it is to have those critical thinking skills before you touch any of this software.

Christopher Penn – 45:38

Exactly. Go play with this in a sandbox somewhere if you are inclined. It is worth playing with to understand what the technology’s capabilities are today and where it’s probably going. I strongly recommend that you use a local AI agent if you have the budget, something like Gemma or Qwen, so you’re not paying token costs. Your first few tries are going to go wildly awry, and if you’re spending Claude Opus tokens, you are going to be very sorry very quickly.

Katie Robbert – 46:18

Again, it’s the whole reason to do the planning before you even bring anything to the table with the software.

Christopher Penn – 46:27

Exactly. That’s going to do it for this week’s show and for Paperclip. Good luck. If you want to chat about your experiences, pop by our Slack group, say hi to us on LinkedIn, and we will see you all on the next one. Thanks for watching today. Be sure to subscribe to our show wherever you’re watching it. For more resources and to learn more, check out the Trust Insights podcast at TrustInsights.ai/tipodcast and our weekly email newsletter at TrustInsights.ai/newsletter. Got questions about what you saw in today’s episode? Join our free Analytics for Marketers Slack group at TrustInsights.ai/analyticsformarketers. See you next time.