Leveraging AI

309 | Claude Fable 5: My Business, Transformed — My Playbook, Shared

• Isar Meitis • Season 1 • Episode 309

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 28:35

Could one AI model save you weeks of work—or even reshape how you run your business?

After spending two weeks putting Claude Fable 5 through real-world business challenges, the answer surprised me. This wasn't about writing better prompts or generating content faster. It was about handing complex, multi-step strategic work to AI and getting back solutions that would normally require weeks of research, multiple consultants, and significant investment.

In this episode, I share exactly how I approached Fable 5, which projects produced the highest ROI, where it exceeded expectations, where it failed, and the framework I'll continue using once usage moves to token-based pricing.

In this session, you'll discover:

  • Why Claude Fable 5 feels less like a chatbot and more like a senior strategist.
  • The framework I used to identify the highest-ROI AI projects across my business.
  • How Fable delegated work to less expensive models to dramatically reduce costs.
  • Real examples where hours—or even weeks—of work were completed in under an hour.
  • Why long-running AI workflows are becoming a competitive advantage.
  • The biggest strengths (and surprising weaknesses) I discovered during two weeks of intensive testing.
  • How to decide when premium AI models are actually worth the investment.
  • Why measuring business ROI matters more than chasing the newest AI model.
  • The mindset shift every business leader needs as AI moves from assistant to orchestrator.

If you're evaluating whether advanced AI is worth the investment for your business, this episode provides a practical, experience-based playbook—not hype. Whether you're building workflows, scaling operations, or looking for your next competitive advantage, you'll walk away with ideas you can apply immediately.


About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

GMT20260713-221515_Recording_as_1512x982

Hello, and welcome to the Leveraging AI podcast, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This is Isar Matis, your host, and today we're going to talk about Fable 5. Fable 5 is the latest and most capable model from Anthropic, and we had access to it. I had access to it for the past two weeks, and Anthropic just extended that access for another week. So we are going to talk about that so you can benefit from everything that Fable has to offer, and I want to share with you, A, what I did with it, what are the best practices, and how you can use it as well while it lasts, because after that it is going to become significantly more expensive. So let's get into it. So let's start by discussing on what the hell is Fable and why does it matter? So those of you who have been following this podcast, especially the Saturday episodes in which we talk about what's happening in the news, we talked about the fact that Anthropic had a new model they called Mythos. It's a new level of model, a new kind of model that didn't exist before, and Mythos was able to generate havoc when it comes to finding vulnerabilities in software and exploiting them in ways that weren't possible before. Instead of releasing it to the world, they told the government that this is what's happening, and they had a short list of companies that had access to it in order to, A, evaluate it, and B, block all the vulnerabilities that Mythos found in their systems. That includes operating system, banking platforms, et cetera, et cetera. More or less any kind of infrastructure level thing it was able to exploit. So that was going on, and then Anthropic in parallel were working on a, let's call it a safer version of the same model. So they were putting more and more guardrails in order to make Mythos acceptable to be released to the general public, and they did. Three days later, the government said, "Hey, you cannot release this. We got a tip that the guardrails could be jailbroken, and then people can get access to the full model, and we're not willing to give this to our adversaries, so you can only allow US users to use it." Anthropic couldn't find a way to do that, so they just pulled the model back, and they released it again back to the public after adding additional guardrails just about two weeks ago. Well, the way they released it is anybody who is a paying member of Anthropic can use Mythos as part of their subscription, but they initially gave one week. That then was extended for a second week, and as of today, it was extended for a third week. I don't know how long they're going to extend it. However, if you are still hearing this this coming week and you have a subscription to Claude, you can use Fable under your subscription. Now, a few things to know, and I'm going to share with you in this episode exactly why you want to do this and how to do this in the most ex- effective way. But when it stops being a part of your subscription, you will start paying by the token, meaning it's going to be very different than the way you're using Claude right now. If you have a subscription, $20, $100, $200 a month, whatever subscription you have, you have X number of tokens that you're using, and then when you run out, it is going to not allow you access anymore. This is not how Fable is going to work. Once this new extension expires, it is going to charge you for the tokens that you're using. So those of you who are not used to these kind of pricing, this is the kind of pricing that is used if you're using the API or if you're on the enterprise plan and using any of the agentic capabilities, and this is paid per millions of tokens. So every million of tokens input, meaning data you're putting in, is gonna cost you $10, and every million tokens of output is going to charge you $50, which is the most expensive model out there in the world right now. Now, $50 doesn't sound like a lot of money, and a million tokens sounds like a lot of tokens, which is true. But if you think about the fact that the model can run independently for very long periods of time, during this time, it is basically writing everything it, quote-unquote, "thinks," it can run through a million tokens in a day easily, and that means that one day costs you $50. If you're running multiple sessions in parallel like I'm doing, it is going to cost you hundreds of dollars per day. Now, this may or may not be valuable enough to you to do that, but you need to know that that's the case. But as of right now, this is not the case. As of right now, you can use up to 50% of your daily token budget under your subscription towards Fable, and if you have the $200 plan, it will allow you to run multiple sessions in parallel most of the day without running out of your tokens. So this is what I'm doing right now because I want to maximize the benefits of enjoying this model before the pricing model changes. And the reason I decided to release this episode now versus, like, in a week or two, which was the original plan, is the fact that they just extended it one more week, and maybe they're gonna continue extending it, but I don't know. Which means if you're listening to this on Tuesday, July 14th, you still have six days to enjoy the usage of Fable under your existing subscription. So what are the biggest differences in why you should care about this? First of all, based on online usage and other people sharing what they're doing, it is dramatically compressing the level of work you would have done manually in a completely different way that was not possible before. As an example, Stripe, the company, the payment platform, reported that Fable 5 did a code base-wide migration in one day that would have taken them two months to do with their team of developers. Now, this is a development task, but it is the same thing that can be done in any other knowledge work. So taking work that would have taken a very long period of time, I will give you a specific example or examples from my universe in the last two weeks on what Fable was able to do for me and compress it into minutes. So the big deal is stamina, if you want. Its ability to run for a very long period of time on its own consistently without diverging and without losing the focus. So in Anthropic's internal own test, the model stayed on task across huge workloads, three X better and longer than Opus 4.8 which, which was the best model they had before that. So think about the fact between an intern that will come back to you every 10 minutes to ask you what to do next versus a senior employee that will just run with the project and will deliver a final outcome. And I have seen that firsthand time and time again using Fable in the last two weeks. It is extremely good at large scale data analysis, including reading entire 10Ks of companies and interpreting charts and going through huge amounts of data, whether financial or other, to make messy inputs into clean, detailed outputs. It is also significantly better at reading visuals. THere's a test I've been running for every single new model that comes out as far as reading floor plans. The reason for that is many of my clients are in that space, and every single model failed miserably. The only model that is not customized to be doing this kind of work that did it close to solid was Fable. And the test that I run is I give it floor plans with multiple types of doors and multiple doors, and I ask it to tell me what's the count of doors. All the models fail miserably, and Fable was actually able to do this correctly more floor plans than any other model I ever tested before. It is still not one hundred percent there, which means it is still without using customized models specifically for architecture are not there, but it is significantly better than any model I tested so far as far as reading a chart and then understanding what's in it on its own. So now I want to share with you my thought process of how I approached this since I knew I had limited time of using Fable, and I'm suggesting you do the same. And now after I've been using it for two weeks very excessively, I can tell you, that the plan worked and that you should do the same. So the first thing I did is I had Claude review my entire plan and my entire company workflows, and I have 28 active Claude projects in different stages and in different steps, and a few that are just in ideation. And I ask it to review all of it and create a four quadrants based on two different vectors. One is the level of complexity and the other is the business impact, basically what is going to be the ROI, and then identify the projects or the components of projects that are going to be in the top right corner, meaning they are really complex, which means that Fable will provide more value and they will provide the most amount of business value to me from an ROI perspective. That produced a shortlist of six projects out of the 28 that we should focus on with Fable that are the most, if you want, Fable worthy. Now, that was step one. Step two, which I was actually pleasantly surprised that is possible, is that I asked Fable if it can assign subtasks to the cheaper models. That's something that I never thought of doing before because I never had the problem of potentially running out of tokens because I'm paying the $200 a month and it's enough tokens to do everything that I need on daily basis. In this case, it wasn't, and I knew it's not going to be. So I asked Fable to look at the tasks on each and every one of these six projects, and I'll show you in a minute what I mean by look at the tasks, and then identify which ones should be done with each of the levels of the different models. So starting from the cheapest one, which is Haiku, and then Sonnet, Opus, and all the way through Fable. And Fable was able to go over 900 open tasks in these projects and mark each and every one of them for the right model that it can assign as subtasks while the main flow, the main chat is going on with Fable itself, which means Fable becomes the strategist and the organizer and the orchestrator or the project manager or call whatever you want to call it, but it's not going to waste all the tokens to do all the different tasks using Fable. It is actually going to find the cheapest model possible to do this. So let me share a screen with you. And again, those of you who are just listening to this podcast, don't worry about it. I will explain exactly what we're looking at. But those of you who are watching this either on YouTube or on Spotify will be able to see how this look like So What we're looking at is an Excel file that has multiple columns. Column number one is the task ID. Column number two is the phase, which phase of the project it is. cOlumn C is the task name. Column D is the status, completed, pending, blocking, et cetera. Priority, critical, high, medium and low. Depends on, so is the next column, which basically says this task depends on, first of all, completing whatever previous steps. Then it, it has a blocked by column, meaning is it blocked by anything that it's missing, any previous steps that needs to be done. Then there are notes, and then there is the final column that I just added, or Fable just added, that is called Model. And you can see right now that based on the level of the complexity that Fable finds for this task, it assigned the following, and I'm just gonna read them in order without going to what the tasks are. It says, "Sonnet, Opus, Fable, Haiku, Sonnet, Sonnet, Sonnet, Sonnet, Sonnet, Sonnet, Fable, Sonnet, Fable, Opus, Opus, Opus," et cetera, et cetera, et cetera. What you can see is that it is using all the different models. It is focusing mostly on Sonnet, which is probably the best value for money right now. But it also goes down all the way to Haiku for simple tasks. Now, how did this whole task list came to life? This is something that Claude does for me for every single project. It's an ongoing, ever-updated task list that it's working through together with me. This is something that I'm teaching in the multi-agent orchestration course. So if you wanna learn how to build really complex, long processes that can automate literally anything in your business, including this kind of capability, come and join our multi-agent orchestration course. I believe we are sold out of the August cohort, but if you go quickly, you may find the last seat. But we will open the September cohort, which will begin in the beginning of September, as soon as we close the August one. So either way, you can come and sign up right now and learn how to build extremely complex automations using Claude and any other AI using the same concepts in just four weeks and four sessions live with me. But with that behind us, let me go back to the episode. So the first two steps that I did is I had Claude review Fable, review everything that I'm working on, suggest where it would be most helpful in means of complexity and value to the business, and then dove down into the specific task in each and every one of these things and categorize where we can use other models. So this was number one. But what did I actually do with it? What kind of strategic wins I got out of this? And I'm gonna give you a few examples because I've done a lot of things. The first one was a full day of work of reviewing the things that I'm working on right now and figuring out which ones is going to drive the most amount of value to my business, right? That by itself is worthwhile. So this would've taken me probably at least full day on my own to go over all the different things and doing the research across different things and figuring out what is going to do the most value. Fable did more than that. It did nine research streams and researched 12 other companies on other similar tasks in order to figure out what might be the most valuable things to do to similar-- based on actual real data and companies and people who are like me, who are offering similar things. Just that would've taken a very long time for me to do manually, or even more time for me to do with other AI solutions that cannot break this into nine different research streams at the same time and summarize them all together. So going back to what I said in the beginning, as far as two really powerful capabilities of Fable, one is staying consistent on task for a very long time, and two is consuming a huge amount of data and still making sense in it. Both of these happen in this very first task I did with it. The second one that I worked on is I've already built previously a sales follow-up system, meaning every customer that engages with me and asks me for a proposal or anything, it monitors the entire communication with them, and it suggests when I'm behind or when I should follow up with them, and so on. It was not working very well, and Fable was able to review the entire thing, find best practices, things that are missing, gaps that existed in the existing solution, both in means of the way the technical implementation was happening, but more importantly, on why I was missing some of the messages that it was sending me and not following up, and it was able to tidy up and mis-- build a much better solution. This will drive more revenue to my business immediately because it will be a better follow-up with any potential prospect or even existing client asking for additional things Now, another thing that I did with Fable that is aligned with the same thing that I said right now, I had several different processes that are already running. They're working, they're working okay, but I felt that they're not working great. And Fable was able to go with a very detailed analysis and go through the entire system, the way it's built. One of these systems has 44 different components, 26 skills, plus connectors, plus N8N processes, plus a lot of other stuff, and he was able to go through all of them, figure out where it wasn't operating. There were multiple components that I built that were not actually connected to the full process. There were a few that were connected to the process but didn't deliver the right outcome or in the right format and all these kind of things. So systems that I thought that were working properly and were not working properly, Fable's able to figure out running through all of them in a single session, suggesting solutions, and then in a session by session, I built these solutions for all of them. I'll give you another mind-blowing example. I want to expand my courses. I want more people to take the courses that I'm delivering right now. Now, why do I do this? Two reasons. It's a big revenue driver for me, but it's also my way to help other people learn AI faster, and that's my biggest driver. It has been since I established this company. I'm a huge believer in helping others figure stuff and be successful, and this is a great way for me to do this. So I had Fable research everything that I'm doing right now and come up with an expansion plan on how can we get more people to join our courses, what other channels, what other deliverables, what other methodologies, literally anything. I said, You have an open check if you want to come up with anything that we can do in order to do this. And in one run that took about 47 minutes, it did the following. It wrote a master expansion plan with three completely separate engine growth strategy. It built a financial model with different options and different projections to tell us where it's going to go. It did a full market research on what my competitors are charging, which channels they're on, how they're performing, what their funnels look like, what platforms, uh, they're using, et cetera, et cetera. It built a master funnel script that will define the 14 different ways that we're going to approach this. It built the lead magnet copies for all the lead magnets that I need. It built three landing pages for the different components that it required. It did a full email and SMS campaign with 34 emails and three text messages across six different sequences with building notes all available and ready to go in my marketing platform. It did a paid ad creative package and a testing playbook across seven different channels, so we can continue to optimize the paid spend. It built the operation docs for all of these different components, including the descriptions and the skills that require to run them and how they're going to approach different situations such as refund and customer service and all these kind of things. It built a podcast growth strategy for me because it figured out this is probably the best channel from a warm lead perspective. And then it ran an adver-adversarial self-reviews using my gotcha skills that caught 14 issues with the original plan. And then it delivered all of that to me on a silver platter without bothering me for one second in those 47 minutes it was running. This would have taken me weeks to build, potentially using some external help that I would have paid a lot of money for. So those of you who are thinking that paying $200 a month for a model, a single model, just to use it for a week before its, uh, current payment method is expiring, I hope now you understand what kind of value this provides. This is a whole game changer in its ability to build really meaningful work that would've-- I would've paid probably tens of thousands of dollars, but definitely thousands of dollars to consultants to help me build that would've requested days of my personal time to work with these consultants to get to these outcomes, and it did it in 47 minutes. Now, is it perfect? Probably not. Will I go through several different iterations with it this week in order to make it much better? 100% before it expires. Do I think where it got me to is worth every cent I invested out of those $200? 100% and probably 10X that Now, I wanna give you another big difference in how it feels to work with Fable versus what it actually does. And what it feels is like a much more mature employee that will actually do everything he or she can in order to get to the final product versus stop and ask you in many different steps along the way. A great example has to do with a workshop I was doing last week. I did a workshop in Chicago, and during that worksh- and when preparing for that workshop, I already had access to Fable, and I asked it with help in preparing for something I wanted to explain in the workshop. And Fable came back with a perfect explanation on how to explain the something I thought about, But in addition, it actually generated the graphics and the slides that I needed. Now, I have a slide generation mechanism, but it actually did not use that because we weren't generating slides. I was just brainstorming with it. But it understood that this will need to become a part of my presentation, and then it went the extra step to go and generate the slides and generate the images. When I asked it how the hell did it generate the images when I didn't connect any mechanism to it, and Claude, as you know, cannot generate graphics, it told me that it figured out that through my N8N connections, one of the N8N connections, which was built for something else, has the API key and the connection to ChatGPT, and ChatGPT API allows it to generate images. So it wrote the prompts, generated the images, and then generated the slides with the images in it This shows you how smart this model is. It knows it has a task, it understands the goal, what I'm actually trying to achieve beyond what I requested in a word-by-word way, right? So beyond what I expressed specifically, it understood what I'm trying to do, and it went the extra step to figure out if it had tools that it can use to deliver that without asking for my assistant, but also without asking for my permission, which now leads us to some of the downsides of using Fable. It will go beyond what you expect it to do, and it will go very, very far to do what it thinks you need to do without necessarily always asking for your permission. On one hand, this is absolutely amazing. On the other hand, in many cases, it gets stuck on stupid things and it doesn't let go. As an example, as part of the new releases from Anthropic, they also released the ability to now run Claude Cowork on the web and mobile versions. This actually broke a lot of things that worked perfectly fine before that, which I'm really, really pissed about because a lot of processes that I had fully figured out that was working with no problem every single day are now not working. One of those is the seamless connection to N8N that I have across the board, which now fails about 50% of the time because of the new implementation by Anthropic. I'm still trying to figure it out. I really hope they will solve this situation, but it is purely related to the fact that it's now running some of the processes on the web and some of the processes on the desktop, and they still didn't figure out how to make it all work correctly. So Fable was trying to do something else with N8N, and it failed, and it tried again, and it failed. And I found it stuck, and every time it fails, it gives you an error message on the screen. Now, I mentioned before, I run five or six or seven of these in parallel at any given time. So I'm not looking at the same chat. Whenever I give it a task, I go to the next one and the next one and the next one until I roll back and circle to whatever is finished. And I went back to this because I saw it was still working after a very long time, and I saw that it failed about 23 times trying to do the thing. But it keeps on trying instead of stopping for a second and saying, "Hey, this is not working. Can we try to figure out an alternative?" Which is what I did in about three seconds. So it wasted a huge amount of tokens and a lot of time on something that if it would have asked me what to do next because it's getting stuck, we could have solved together in about a minute, maybe two. And so there are pros and cons to that. The other big risk is obviously the cost. So again, right now it's a part of my subscription, and I'm willing to pay the $200 a month with no problem because the value is definitely there. But once it switches to the new payment mechanism, it will be very hard for me to understand how much am I going to pay. Now, will I be willing to pay that? I think it will be a lot more selective than I'm doing right now. Right now, like I said, I'm running many different large projects all in parallel across the board, trying to maximize what Anthropic is giving me right now. I think I'm going to choose way more carefully what I'm going to work on with Fable moving forward once I have to pay per token. I will probably keep on using it. I will probably use it in the same concept of having it delegate most of the work to cheaper models and just monitor and do whatever I need to do with that. Once I get to that bridge and cross it, I will report back on what I find So what is the bottom line that I want you all to remember? Number one is if you have a paid Claude account, try to use Fable for the right projects and use Fable to help you figure out what the right projects are. Number two, it is definitely worth from a company ROI perspective to pay the $200 a month right now, especially if you're running multiple things in parallel, and you probably want to do that. If you're not sure, just start with the $20 a month. You're probably gonna get stuck at 10:00 a.m., then you're gonna upgrade to the $100 a month. You'll probably get stuck at noon, and then you'll upgrade to the $200 a month, and it will allow you to run the full day with Fable with multiple things in parallel. If you're not running multiple things in parallel, then probably the $100 a month will be good enough for you. So this is number one. Number two, which is a big deal that relates to what I just said, the best model is not always the right model, and this is something that we are gonna hear a lot moving forward. We talked about this in the last two or three weeks of the news episodes. We are seeing a very big shift from token maxing on the best models to let's figure out what is the cheapest model that will get the work done consistently every single time, and you can do that together with Fable inside of Claude right now, as I mentioned in this episode. The next one is for every project, you wanna measure the results of what you're doing. So for everything that I'm doing that's not related specifically to Fable, but the Fable ROI question once it goes to the new pricing model will make it very, very important. For everything that we're doing, we first of all establishing a baseline. What are we achieving right now before we build this thing? What are we achieving right now before we fix this thing? And then you can measure the value of whatever it is that you built or that you fixed or upgraded, and you'll be able to see if the tokens and the cost of the tokens that you invested in it makes sense, which will allow you in the future to tweak better and fine-tune which models are best for which use cases. Now, the other thing that I found that is also very helpful is in many cases, Fable was able to talk me out of building more complex solutions. In at least two cases, I already had a plan figured out. I already had my gotcha skill that is my devil advocate give me the green light after fixing a bunch of things. When Fable saw it, it actually came up with significantly simpler, much more elegant, and hence more easy to maintain solutions that will achieve the same, and in some cases, even better results. So it's not just for building more complex things. In some cases, it's identifying how complexity could be simplified in something much more elegant that will give you more value longer with less things to fix And the last component that goes back to, again, the course and some of the things that I'm teaching and how I work, the bottleneck now is not the model capability. This model is really mind-blowing in its ability to do more or less anything across the board. The bottleneck becomes how well can you define what it is that you want to do, and building an AI mechanism that helps you with that becomes one of the most important skills we are all going to have in the future, definitely in business capacities, but also in our personal lives. That is it for today. I hope you found this episode valuable. I hope you're going to push Fable to the max in the next week while it still lasts, and potentially beyond if you still find it valuable or if Anthropic stays as giving as they are right now. And they might stay giving because of the crazy competition that they have right now from OpenAI that is just releasing Werk and releasing a lot of other stuff. I will record a whole separate episode about the comparison between ChatGPT Werk and everything they've released, uh, with the app and the connectivity to Codex and so on. I think they're making the right steps to close the gap on Anthropic. I still think the gap is pretty significant, but by the time I record this episode, which might be in the next week or two, they may close the entire gap. I'm sure they're working on it di-diligently. But for today, that's it. Go play with Fable. Share what you find out. Share with me on LinkedIn and with other people on LinkedIn what you find. Uh, if you have any tips or tricks, I would love to learn that, and have an amazing rest of your week.

Head Shot

Isar Meitis

Host