Fable 5.1 in Claude Desktop's Cowork mode can read your company folder and generate a personalized week-by-week revenue plan based on your actual clients, projects, and priorities. Not a generic AI output. A plan that names your real accounts and maps out every week through December. This guide shows the exact setup and the exact prompt.
Most AI planning tools have the same problem: they give you a generic framework.
"Focus on your top clients." "Cut low-value activities." "Review your pipeline weekly." The advice is technically correct. It is also useless because it does not know who your clients are, what you are actually working on, or where your real revenue is stuck.
Fable 5.1, connected to your company folder via Cowork mode, does not have that problem. It reads your actual files. It knows your clients by name. When it builds a plan, it builds one for your business, not a placeholder business.
This guide covers the setup, the prompt, and what to do with the output.
What Cowork Mode Is and Why It Matters
Claude Desktop has a Cowork mode that connects the AI to a folder on your computer. When Cowork is active, the AI can read any file in that folder: client contracts, project notes, pipeline documents, invoices, proposals.
The model you run in Cowork mode matters. Fable 5.1 is Anthropic's most capable reasoning model as of September 2026. It is not the default in Cowork mode. You have to select it manually.
The combination of Fable 5.1 and a connected company folder produces output that is qualitatively different from anything a standard AI chat session can give you. The model reasons against your actual situation, not a generic one.
What You Need Before You Start
Three things:
- Claude Desktop installed on your computer
- A company folder: a single folder containing your key business documents (client notes, project files, active proposals, invoices, pipeline tracking)
- Ten minutes
The company folder does not need to be exhaustive. It needs to contain enough that the AI can identify your active clients and your current work. A folder with 10 to 20 core documents is enough.
Step One: Open Claude Desktop and Create a Project
Open Claude Desktop.
In the sidebar, click "New Project." Give it a name. Something that reflects its purpose: "Ultimate Productivity Agent" works. "Q4 Revenue Planning" works. The name helps you find it later and gives the AI a working context from the start.
This project will hold your Cowork session and its memory of the folder you connect.
Step Two: Connect Your Company Folder
Inside the new project, look for the Cowork or folder connection option. It is typically in the project settings or the session header.
Click it and navigate to your company folder. Select the folder, not an individual file. The AI will have access to everything inside it.
Wait for the connection to confirm. Claude will show a message indicating the folder is accessible. This is the moment the session becomes personalized.
Step Three: Select Fable 5.1
This step most people miss.
In the model selector (usually a dropdown near the prompt input), find Fable 5.1. It is not the default model, even with the Cowork feature active. The default is typically an earlier model version.
Select Fable 5.1 explicitly.
The difference in output quality between the default and Fable 5.1 is significant for planning tasks that require reasoning across multiple documents. The model's reasoning capability is what allows it to prioritize across your actual client list, not just list your files back to you.
Step Four: Paste the Prompt
Once the folder is connected and Fable 5.1 is selected, paste this prompt:
Copy this.
Copy This Prompt Here's everything I'm working on right now. Build me a plan to increase revenue by 30% by Christmas, focused on the smallest number of changes with the biggest impact. For anything low value, tell me whether to cut it, delegate it, or fix it. Then give me a week-by-week breakdown from now until Christmas so I know exactly what to work on and when.
Hit send. Let the model finish before you read the output.
Your First Real Run: Do not edit the prompt on your first run. Use it verbatim. See what comes back before you decide whether to refine the inputs.
The Mistake That Makes It Fail: Connecting a folder that is either empty or too broad. If the folder has 2 files, the AI does not have enough to personalize on. If the folder has your entire 2TB hard drive accessible, the signal is buried in noise. A focused folder of 10 to 30 working documents is the right size.
What the Output Looks Like
When the prompt runs against a connected company folder with Fable 5.1, the output includes:
Prioritized client focus. The model identifies your highest-leverage accounts by name, based on what it reads in your files. Not categories. Actual clients.
Cut or fix recommendations. For activities and obligations it identifies as low value, it makes a specific recommendation: cut it, delegate it, or restructure it. The recommendation references what it found in your documents.
A week-by-week plan. Starting from the current week through the end of December, each week has a primary focus, a secondary task, and a metric to check. The weeks reflect your actual situation, not a template.
Revenue logic. The plan explains why the prioritized actions lead to 30% growth, based on what it sees in your client data. You can interrogate that logic in follow-up messages.
How to Work With the Output
The first output is a starting point, not the final version.
Read it once. Then ask three follow-up questions:
- "What are you most uncertain about in this plan? Where is your confidence lowest?"
- "Which of the clients you named are you relying on most for the 30% target? What assumptions are you making about them?"
- "Is there anything in the folder you did not use that you think I should know about?"
These three questions surface the model's assumptions and fill the gaps that the initial prompt leaves open. The plan improves significantly after one round of follow-up.
Save the final version as a document. Export it from Claude Desktop or copy it into a working document. Revisit it weekly and update the folder with new notes so future sessions stay current.
What Else This Prompt Structure Works For
The same setup (connected folder, Fable 5.1, specific planning prompt) works for other high-stakes planning tasks:
Hiring. "Based on what I am currently working on, what type of person would have the most leverage in our operations? Give me the three most critical skills and one red flag to screen for."
Client retention. "Which clients in this folder show signals of disengagement, delayed responses, or reduced scope? What should I do about each one in the next 30 days?"
Capacity planning. "What is the highest-risk commitment I have between now and the end of the year? What breaks first if things get busier?"
The model reads the same folder each time. The prompt changes. The output personalizes to your actual situation.
Honest: The Limits You Should Know
Fable 5.1 is not the default. You have to select it manually every time you start a new session, or check that the prior session preserved your selection. If you start a session without checking and the default model is active, the output quality drops noticeably on complex planning tasks.
The folder is a snapshot. The AI reads what is in the folder at the time of the session. If a client situation changed last week and you have not updated your notes, the model works with the old information. Keep the folder current.
30% is a number, not a guarantee. The prompt asks for a plan targeting 30% growth. The model builds a plan that it believes supports that target based on what it reads. Whether that target is achievable depends on factors outside the folder. Treat the output as a structured prioritization, not a forecast.
Context window limits. If your folder contains hundreds of large documents, the model may not be able to read all of them in a single session. For very large folders, create a focused sub-folder of 20 to 30 key documents for planning sessions.
The plan is a starting point. One session produces a draft. The follow-up questions, the weekly reviews, and the folder updates are what make it a working system.
The Close
Open Claude Desktop right now. Create a project. Connect your company folder. Switch the model to Fable 5.1.
Paste this prompt:
Copy this.
Here's everything I'm working on right now. Build me a plan to increase revenue by 30% by Christmas, focused on the smallest number of changes with the biggest impact. For anything low value, tell me whether to cut it, delegate it, or fix it. Then give me a week-by-week breakdown from now until Christmas so I know exactly what to work on and when.
Read the output. Ask the three follow-up questions. Save the result.
That is the whole setup. It takes under ten minutes.