Fable 5 launched on 9 June. Three days later a US export control directive took it offline worldwide, and it stayed dark for 19 days. It came back on 1 July with new safety classifiers and a short free window that has since been extended three times: first to 7 July, then to 12 July, and now to 19 July.
Here are the terms as they stand today. On a paid Claude plan (Pro, Max, Team, or a premium seat on a seat-based Enterprise plan), Fable 5 is included for up to 50% of your weekly usage limits until 11:59pm Pacific time on 19 July 2026. There is nothing to activate. After the window closes, the model moves to usage credits at the highest rate Anthropic has published for a generally available model.
Anthropic says it intends to fold Fable 5 back into standard plans once it has the compute, with no date given. Treat the window as real and the date as movable, in both directions.
KEY POINT
The free-access date has moved three times in five weeks. Check Anthropic's current terms on the day you plan around them, not the day you read this page.
Until 19 July: Included, up to 50% of weekly plan limits
After 19 July, input: $10 per million tokens
After 19 July, output: $50 per million tokens
One thing to be clear about before the prompts, because it decides what you should expect from them.
Fable 5 is a reasoning model. It reads, structures, weighs and drafts, and it is genuinely strong at long, messy thinking jobs: a year of purchase history, a pile of unstructured requests, an approval logic full of exceptions. What it does not do is act. It will not watch your inbox, move a request through Teams, or fire an approval in Power Automate on its own.
That split is the honest shape of AI in a small business right now. The model does the thinking. The tools you already pay for, like Teams, Slack, Power Apps and Power Automate, do the doing. Connecting the two is a build, and the build is the part these prompts deliberately leave out. Each prompt below gives you the thinking in full. Where a section touches the plumbing, it says so and tells you who to talk to. Spoiler: me.
KEY POINT
Fable 5 is a reasoning model, not an automation platform. It thinks; it does not click. Anything that must happen automatically inside Teams, Slack or Power Automate is a build, and the build is the part these prompts leave out.
BEFORE YOU PASTE
Do not feed it anything you are not permitted to share. Remove names, prices, and personal data, or substitute rounded figures and initials. Test the tool on a redacted version first.
Every SMB has the same intake system. It is called chaos. Someone needs software, someone needs a supplier, someone needs a forklift, and it all lands at 4:56 on a Friday through whichever channel found you first.
Prompt 1 takes the whole pile and structures it: what is a real request, what is a duplicate, what is Kevin ordering another monitor. You get a register you can act on instead of an inbox you avoid.
The prompt sorts the last 30 days. It will not catch the next 30 automatically; a request front door inside your existing tools is a build. DM INTAKE to @d1_advisory on Instagram, or bring it to a call, and we will fix your front door.
Turn 30 days of scattered purchase requests into a register you can act on.
Remove staff names and any client-identifying detail before you paste. Initials work fine.
Big companies have demand planning teams. You have a gut feeling and a spreadsheet you are scared to open. The signals already exist, scattered through sales notes, operations requests and every "we always run out of this" said in passing.
Prompt 2 gathers the noise into an actual picture of what the business needs and when, ranked by how much a shortage would hurt.
Turning that picture into a live feed from your Teams channels and your own data is the bridge I build. DM DEMAND to @d1_advisory if you want to see what your business is actually asking for.
Consolidate scattered demand signals into ranked demand lines.
Strip customer names from sales notes before pasting. The pattern matters, not the person.
Most small businesses forecast one of two ways: panic buying, or running out. Both quietly drain cash you never get back.
Fable 5 is built for long, messy reasoning, which is exactly what a forecast is. Prompt 3 reads your purchase history and your patterns, sketches next quarter line by line, and marks each line solid, probable or guesswork, so you negotiate with suppliers from a plan instead of a shortage.
Forecasting on vibes? DM FORECAST to @d1_advisory and we will build you a real one, refreshed on a rhythm instead of in a crisis.
Sketch next quarter's demand from your own history, with honest confidence levels.
Round any commercially sensitive figures before pasting, and check your plan's data retention settings first.
In most SMBs the approval process is "ask Sarah", and Sarah is on leave. Nobody knows the threshold, nobody knows who signs off, and everything stops the second one overworked person goes quiet.
Prompt 4 drafts the logic: under this amount, approve; over it, escalate; new supplier, flag for review. Same rules, every request, no bad moods and no forgotten steps.
Plugging that logic into Teams, Slack or Power Automate so it fires without you is the build, and that part is on me. DM APPROVE to @d1_advisory and let's take the pressure off your Sarah.
Draft plain-English approval rules that work on paper before they touch software.
Describe roles rather than named people where you can. The matrix should survive staff changes anyway.
Here is why your last process died. You asked people to log into somewhere new, and they did not. The system sat empty and the requests went back to hallway conversations.
Prompt 5 maps every procurement step into the tools your team already lives in, then picks the one pilot worth running first. The request happens where people already are, and the approval comes back the same way.
Want procurement that hides inside the tools you already pay for? DM TOOLS to @d1_advisory.
Map procurement into the tools you already pay for, and pick the first pilot.
List tool names and how they are used. Never paste passwords, licence keys or admin details.
Every "I just grabbed it myself" costs more than you think. When there is no easy way to raise a request, people go rogue: off-contract, retail price, a new supplier when you negotiated hard with an existing one. For a small business that leak quietly adds up to thousands.
Prompt 6 reads a redacted spend export, names the leaks, sizes them from your own figures, and gives you the three fixes that matter.
Behaviour follows friction. Making the easy path the right path takes a ten-second request flow, and standing that up is what I do. DM SPEND to @d1_advisory and let's plug the leak.
You do not need a procurement department. The gap between your business and the big players is infrastructure, the kind they can afford and you assumed you could not. Intake, demand, forecasting and approvals, running quietly in the background, is now within reach of a business your size.
Prompt 7 designs the right-sized version: the minimum operating model your business needs, sequenced over 90 days, with clear instructions on what not to build.
The design is free. Making it run inside your tools without a single extra hire is the work. DM PROCURE to @d1_advisory and let's talk about punching above your weight.
Run any prompt above and you will have something real: a register, a forecast, an approval matrix, a leak list, an operating model. What you will not have is a system. The output sits in a chat window until someone connects it to the places work actually happens.
That connection, the front door in Teams, the approval that fires in Power Automate, the demand feed that updates itself, is the build, and it is deliberately not on this page. Every business wires it differently, and pretending one tutorial fits all of them is how SMBs end up buying what they were sold.
One more honest note: these prompts do not expire on 19 July. They run on any capable Claude model; the date only changes what the thinking costs. Use the free window to test the reasoning against your own mess, then decide what deserves to become a system.
If one of these prompts put its finger on your problem, bring the output to a discovery call. We will look at what the reasoning found, what a build would take inside your existing tools, and whether it is worth doing at your size. Honest answers, either way.
Buy what you need, not what you were sold.