Why I Will Pay For Fable

This past week, I had one of the most enjoyable and revelatory experiences of my career. And I do not say that lightly. I started writing software in the early 80’s. Yeah, almost fifty years ago. Getting a computer to do what you wanted felt magical. I still have this felt sense I can remember the smell of my Apple II+ when we unboxed it. It smelled like ozone and magic.

But until I asked Fable 5 to help me study call records for GameTruck, I had never really experienced what I would call magic.

I know, examining sales, and customer service phone calls may not sound that exciting to you, but when your business is getting squeezed by an indifferent economy (higher everything while families are cutting back - feels like everyone gets to raise prices but me), you start looking for every edge you can find to gain more business.

When demand is high you focus on delivery. When demand is low, you focus on sales and marketing. (And why is sales first and not marketing? Doesn’t marketing lead to sales? But I digress¹). Therefore, studying recordings of calls it is.

But how do you process 5,000 phone calls?

Enter Fable

I picked up this particular project, just as Anthropic re-released Fable 5. Sitting down to my Claude Code interface - (I just realized writing this, I want to see if I can make my MacOS terminal look like my old Apple II+, I bet someone has made that skin!) I started with my problem solving process. It’s a variation of the Engineering Problem Solving Process, and it goes like this:

10 REM LEARN LOOP
20 PRINT "DISCOVER"
30 REM -- LOOP START:
40 PRINT "EXPLORE..."
50 PRINT "EXPERIMENT..."
60 PRINT "IMPLEMENT..."
70 INPUT "DONE? (Y/N) ", D$
80 IF D$ = "Y" THEN GOTO 100
90 GOTO 40
100 PRINT "COMPLETE."
110 END

The plain English way of saying this is, start where you are, with what you have, and do what you can. But that means you have to first know where you are, and what you have. Then you have to figure out what you can do with that, then… do it.

I tell you this because I want to make it clear I did not have a plan. I did not have a specification. Most software development immediately pushes for clear specifications with testability baked right in.

That always felt backward to me. It was like you had to know the answer before you started.

But I didn’t have any answers, just questions. Heck, I didn’t even know how to access the files I wanted to study. (They were not in a Dropbox folder, Google Drive, or even a USB stick) The files I wanted were stored in an industry standard S3 cloud thingy. Hell, French is “a standard”, but I don’t speak it. Just because something is standard doesn’t mean I can take advantage of it.

Discovery Mode

And this is where the magic started. With my Claude Code model set to ‘Fable 5’, I basically asked it if it could build me a file browser so I could look at the data. I had all the credentials and techno-jargon I supposedly needed to access a remote cloud drive, I just lacked the experience and the mental framework to know how to do that.

But I told Fable what I wanted, something like the old school Norton Commander² interface, so I could look and see what we were working with. The left pane would be the files, the right pane however would display information about the files.

Within Minutes Fable wired up exactly what I was looking for. It wired up the cloud drive and I had a skeleton of an application that let me browse the files on that drive. I asked Fable to make some enhancements. The files I was looking at were json objects with structured data. Imagine taking the rows of your spreadsheet and scattering them like Lego blocks in a folder. That’s what I was looking at, a bunch of individual scattered records. So I asked Fable to show me all the data for each file as I browsed in the right pane.

BANG! It was done.

Pretty much as fast as I asked for a small enhancement, Fable implemented it.

In less than 15 minutes I was able to get a sense of what kind of data I was looking at. Now it was time to explore.

Exploration

What moved me out of Discovery mode and into exploration mode was that now I had a sense of what was there, I wanted to begin to examine it more closely and lay the foundation for working with the data. Put another way, I knew what I had, but not what it meant. To figure that out, I would need to explore.

I asked Fable to change the way I browsed the files so that I could see them not only by date, but also by Franchise, and staff member. There are multiple ways of looking at things, and your perspective matters. Again, Fable was happy to oblige. Now I had tabs on my left panel. Click Franchise ID and I could see how many calls, per day (or week), were coming in for any franchise. Click an agent, and I could see how many calls per day (or week) an agent handled, or I could go back to looking at the data by date.

Every time I had a new idea on how to sort or parse the data, I could ask Fable to add a feature to my phone call browsing app and it did.

Once I realized I now had the rudimentary tools to quickly manage the data, now I could begin to experiment.

Experimentation

What’s in each call? How to find out? Here is where Fable really earned its keep. I knew I did not want to listen to every call. I was going to have AI do that for me. But what model to use? Definitely not Fable (too expensive), but what about Opus? or Sonnet? or maybe ChatGPT? How about an open source model like Qwen3.6, or something from Google? Fable quickly generated a list of likely candidates for the task at hand. We would take a sample of calls, send them to a variety of models and compare the results and the price.

Then Fable proceeded to build:

1. a method to send a call transcript (and later a recording) to a model for analysis.
2. calculate the cost of that request
3. Make it easy for me to compare the results from each model

Suddenly I had gone from asking, “What is all this stuff?” to “How can I, and should I, work with these calls?” The answers came back lightning quick. Google’s Gemini 2.5 Flash was not only fast and cheap, it could also do something none of the other models could - analyze the tone of voice on the calls.

There’s a big difference between a happy “I’m fine” and an angry one. Gemini could mostly tell the difference. Other models couldn’t do it at all.

Empowered with this level of tooling, I spun back around to Discovery mode. Now that I had the files, I needed to find out what was in them.

Implementation

Here’s the lie about my model. Implementation is not something that happens after experimentation, it happens during experimentation. I think of it as build as you play.

With each step, as Fable and I explored and experimented with the data, we implemented new functionality. The application grew organically. Each new understanding led to better insight into what the tool needed to do.

What I ended up with, after multiple iterations was a tool that could analyze all of yesterday’s phone calls and reveal patterns.

If my goal was to sell more out of what we had, we needed to make it easier for the customer, and the sales team. Make it easier for the customer to get what they wanted, and take the sand out of the gears for the sales team.

One point of friction I noticed early on was when a sales call ground to a halt because the agent needed to check with a franchise owner. I instructed Fable to tell the AI analyzing the phone call to call out any time, on any call this happened. What’s more, I wanted the reason made clear, in big bright letters. Why on earth were we stalling sales by forcing agents to talk to the owner?

Remember when I said you could look at calls by Date, Agent, or Franchise? This view let me see:

1. Which agents needed the most help and
2. Which Franchises were the hardest to work with.

Shockingly, it was one of my corporate locations that generated the most “Check with Owner” errors³. In one two-day span eight calls in a row triggered the check with owner flag.

Immediately I could see the payoff. My new goal was to get that number to zero. No check with owners. We needed to empower the team with the right tools and support to help the customer immediately.

But I could also look at calls from the human point of view. Which staff needed more training? With Fable’s help, we created a coaching report - something that analyzed a week’s worth of calls and generated real, actionable feedback the manager could use to help each member of the team improve. It was personalized, custom feedback grounded in real calls with real work.

And by adding an audio player in the app, the manager could listen to the call themselves and confirm what the tool told them.

All In One Day

Did I tell you I did all of this in one day? One single day. That was it. I woke up in the morning, not understanding exactly what I had to work with, and by the end of the day, not only did I have a tool for exploring, examining, and using this information, I gained valuable insights into my business and I was able to hand my sales team manager a new tool she could use to coach up her team every week.

From zero understanding to process improvement tool and business insights in 8 hours.

Are you kidding me?

But here’s the kicker. I blew through my Fable tokens. I hit the cap. No more “free” Fable for me. So I went back to using Opus to try and tweak the app. I built it on my Mac, but I also needed a Windows version.

It took Opus 4.8 four hours simply to convert the app from macOS to Windows. Four Hours. I am not making that up. Opus kept getting lost, wasting time chasing rabbits down holes. It clearly needed a lot more scaffolding and hands on direction. It needed constant approval, where Fable… just built what I wanted, and did it fast.

Summary

This left me with two overwhelming impressions. Fable will enable a whole new paradigm of software development. I call it Minecraft mode. You can go out into the world of your business and look for opportunities to build cool things. You don’t need a detailed design spec, or lots of documentation. You need curiosity and intent. That’s it.

Design specs were made because development resources are expensive and slow. Fable removes that friction. It is fast, and compared to a team of $20K a month engineers, it feels damn near free (it’s NOT!)⁴ If I factor in my costs, like the cost not only of my time, but the impact of rolling out better tools to improve my business quicker? How can I not pay for Fable?