This website uses cookies

Read our Privacy policy and Terms of use for more information.

Have you noticed your feed this week? Every second post is announcing something called Fable 5... and I couldn't find a single one that explained what it actually is. Everyone announcing, nobody explaining. So let me do the explaining, and then I'm going to give you a job for it that I think will genuinely change how you see this technology.

First, what actually happened

Anthropic, the company behind Claude, put a new engine inside it on Tuesday. That's all a "model" is by the way... the engine inside the chat window. You won't see anything different on your screen, but what it can do has changed quite a lot.

Now the part I find genuinely fascinating, and the part nobody is talking about.

They built this engine months ago. And they decided it was too powerful to give to people. Think about that for a second... a company built something, looked at it, and said no, not yet. So they locked it away, and only cybersecurity teams and companies running critical infrastructure were allowed anywhere near it. What launched on Tuesday is that same engine with safety guardrails wrapped around it so the rest of us can use it.

And if you pay for Claude, it's already sitting in your account. You don't install anything. It costs nothing extra until the 22nd of June, then it moves to a pay as you go arrangement for a while.

So what can it actually do that the old one couldn't? Three things worth knowing.

The first is stamina. You know that thing where you give AI a long messy piece of work and somewhere in the middle it loses the plot? This one doesn't. Stripe, the payments company, told Anthropic that a piece of work they had planned two months for was finished in a day.

The second is sight. It's much better at reading what you show it... screenshots, scans, a photo of a document. You stop typing things out just so it can understand them.

The third is numbers. Its scores for complex data analysis are the best of any Claude version so far, which matters if your working life happens inside spreadsheets, like mine did for 17 years.

One honest warning before the good part. The guardrails are set cautious right now, so every so often it will refuse something completely harmless. Anthropic have admitted this themselves. If it happens to you, rephrase and move on, you've done nothing wrong.

Now, the £5,000 job

Here's the test that changed how I think about all of this. Stop asking whether AI can help you with a task. Ask whether it can do the job you were saving up to pay a professional for... and then judge it by that standard, not by chatbot standards.

A qualified accountant in my cohort was quoted £50,000 to £60,000 for a piece of technical accounting work. She did it herself with Claude in 20 minutes, then checked it the way only a qualified accountant can. That's the standard I mean.

So here's the job. A proper pricing review. The thing consultants charge four or five figures for, and most small businesses never buy, so they just don't review their prices. For years.

You'll need Claude Cowork for this one. It's the version of Claude most people don't know exists, a desktop app where Claude works inside a folder on your computer instead of a chat window. It reads what's in there and creates real files next to your own. With Fable 5 as the engine, it has the stamina to do the whole project, not a slice of it.

Step 1. Make a folder. Into it goes a year of sales data exported from wherever yours lives (Stripe, Xero, your till system, a spreadsheet), your price list, your costs, and the pricing pages of two or three competitors saved as PDFs. Scans and screenshots are fine, that's the sight improvement earning its keep.

Step 2. The honesty bit. Strip customer names out of the sales export, they're not needed for this. If you're doing this on company data as an employee, your firm's AI policy beats my enthusiasm, so check it first. And know that Cowork sits on the paid plans and uses your allowance faster than chat. Worth it for this job, but I'd rather you knew.

Step 3. Point Cowork at the folder, make sure Fable 5 is the selected model, and give it this brief:

"In this folder you'll find a year of sales data, our price list, our costs, and PDFs of three competitors' pricing pages. Work out our margin on every product or service. Then model three pricing scenarios in a spreadsheet I can edit: a careful one, a confident one, and the one you would actually recommend, with the revenue impact of each based on our actual sales mix. Finish with a one page memo telling me the single change you would make first and what it's worth per year. If anything is unclear or missing, ask me before you start."

Then go and do something else. The school run, in my case. This is not a two-minute answer. It's the engine working through a project, asking you questions when something doesn't add up, and leaving files in your folder when it's done.

Step 4. When you come back, pick two products and check the margin maths by hand. If those two are right, work with the rest. And remember what it can't know... it can see your sales mix, but it can't feel your market. The scenarios are its analysis. The decision is still yours.

If you're employed rather than running a business, the same workflow reviews your department's budget or your firm's supplier costs. Same folder, same brief, same standard.

You're done when there's a margin breakdown, an editable three-scenario model, and a one page memo sitting in your folder that didn't exist this morning. Then ask yourself what you would have paid for that a month ago.

If you run it this week, hit reply and tell me what the memo said. I want to know.

One last thing

I've just launched something and wanted to share with you all.

It's called the AI Audit. If you read this issue and thought, fine, I believe it works, but where does it actually start in my work... that is the exact question it answers.

Here's what it is. You and me on a Zoom call for 45 to 60 minutes, talking through how you actually work. No preparation needed. Within 48 hours you get a 7 to 9 page report built around your specific role or business, with four prioritised recommendations, the implementation detail for each, and the ROI worked out... the hours and the money, per week. Then a 30 minute follow-up call where we walk through it together.

One version for professionals, one for business owners.

Thanks for reading,

See you next week with more ways to use AI without losing your mind (or your credibility).

Orgesa Meli

P.S. If this saved you from a future hallucination disaster, forward it to someone who's using ChatGPT/Claude for proposals, reports, or client work. They'll thank you later. Subscribe to my community here.

Keep Reading