Managing your AI prompt library

There may be a prompt in your firm that one person is quietly proud of. It took a few goes to get right, and sometimes it is several prompts run in order. The draft it gives back needs little changing.

That prompt tends to live with the person who built it. It might be in a note on the desktop or somewhere in a chat history.

This hidden prompt could lead to productivity gains across a firm.

1What a prompt library is

A prompt library is a shared collection of prompts that have produced good work, written up so the rest of the firm can use them without the author sitting next to them.

Most people in your firm probably already have one, but they may not call it that.

It works a lot like a firm's templates, such as the standard engagement letter or the file note format. They were built carefully over years, and everybody uses them, because they work.

A prompt can also write the draft for a particular client and matter. Take a follow-up email on an unpaid invoice. A good prompt can write it knowing the client has been with the firm for 11 years and the invoice is nine days overdue. A person in accounts can teach it how the firm writes to its clients.

Let's take a look at Sally and John.

Sally in one team has worked out how to turn a meeting recording into notes and actions that need almost no tidying. It took her three or four rounds over a fortnight, mostly in her own time.

John in another team is doing the same job with the same tool. He gets back a wall of text and reformats it by hand. He is two weeks behind her on a problem she has already solved.

To Sally, the method feels like knowing a keyboard shortcut. So how would John find out it exists?

2Work done twice

Atlassian asked about this in its State of Teams 2025 research, a survey of 12,000 knowledge workers and 200 executives. Half said teams at their company are duplicating work without realising it. The same research found about a quarter of the working week goes on searching for information. Those figures cover knowledge workers in general, across industries.

In law, Clio's 2026 Legal Insights Report, a survey of 533 Australian legal professionals, found 59% rely on generic tools with no understanding of their matters, their clients or the law. Clio also reports that 30% of Australian legal professionals say fitting new tools into existing workflows is their biggest challenge. Sally has taught her generic tool how the firm likes its meeting notes, and that teaching sits in her chat history.

3A firm with one library

When a firm notices good methods sitting in chat histories, it will often start a shared folder called prompts.

But what happens after the folder is made? A handful of prompts go in early on, each one a block of text. The next person to open it may read two and decide it is quicker to write their own. Where new prompts need approval before they go in, what we're seeing is that entries slow after the first month and then stop.

What to put in a prompt library

Four decisions set a library up.

1. Choose the work that goes in. It should pass three tests: the work is repeated, more than one person does it, and a person reviews the output before it goes anywhere. Client onboarding letters, file notes, status updates, invoice follow-ups, chronologies of key dates and first drafts of routine advice all pass.

2. Write each entry up. Alongside the prompt, note what job it does, in the words a colleague would use to ask for it. Mark inside the prompt what to change before running it. Add an example of output that came back right, who wrote it, and when it was last checked and against which tool. Client-facing prompts can share one set of instructions about the firm's tone.

3. Put it where your people already go for templates and precedents. That might be SharePoint, or the practice management system that holds your precedents. A shared project in Claude or ChatGPT can work too, where the firm's plan and settings allow everyone who needs it to have access.

4. Let people add to it freely, and give it one owner. The owner looks every month or so at what has been added and tidies anything that hasn't been used in a while. This can be the person who looks after AI across the firm.

With those decisions made, John opens the library and finds Sally's meeting notes method, with an example of good output beside it. Sally's method becomes how the firm does meeting notes. It works when she is on leave, and it stays with the firm if she moves on. The invoice prompt from accounts does the same for invoice follow-ups.

When a new starter joins John's team, they are shown the library in their induction, next to the templates and the file naming rules. They can work the firm's way from the start. Because client-facing prompts carry the same instructions about tone, a reminder from accounts and an update from John's team read as though they came from the same practice. When the firm reviews its tools and subscriptions, the library shows what each one is being used for.

4Where we come in

You may have several Sallys already, each with a method the rest of the firm could use. Our investigation finds those people and maps the work your firm repeats most often. It measures how long those tasks take today, so you start from a baseline of your own. The findings belong to your firm whether or not we go any further together.

If you'd like to know what is already sitting in your firm's chat histories, reach out to our team. The first conversation works out whether your firm is the right fit for the AI Uplift Program, and whether we are the right people to help.

Starting a prompt library