From Prompts to Skills: The Power of AI Skills in Accounting Firms

Most practices start with ChatGPT and prompts. A team member types a question into a general-purpose AI, gets an answer, and copies it into their work. Sometimes they refine the answer. Sometimes they use it as-is. This works exactly once. For every new client, every new situation, they start again. But the problem is that prompts don’t scale, they’re just one-off instructions for one-off tasks. Let’s discuss the problems with prompts, and the power of AI skills in accounting firms.

What happens when you rely on prompts?

Output quality varies. One person writes it slightly differently. New team members don’t know the prompt. Inconsistency kills scalability.

Why? A prompt is a conversation. “Summarise this monthly payroll report and highlight any variances over 10%.” Next month, the same team member writes the same prompt again but they phrase it slightly differently or forget to include the “10%” threshold this time. The output quality varies, the process is inconsistent, and you can’t train someone on “remember the prompt I wrote last year”. It just doesn’t scale.≥

What is a reusable skill, and how is it different?

A skill is a packaged, tested, documented instruction built into your workflow. Your team clicks a button and gets a consistent output. No prompts, no copying, and no wondering if they did it right.

A skill is fundamentally different. A skill is a packaged, tested, documented instruction that your team can run the same way every time, with consistent results, without needing to remember the exact prompt or copy-paste anything. It’s reliable, repeatable, and can be built into your workflow (not separate from it).

Practices that scale AI don’t rely on team members remembering prompts. They build skills that are documented, tested, and produce consistent results.

Which tasks work well as reusable skills?

Look for clear rules, inputs and outputs. They should be something that can be reliably automated.

How do you define the task clearly enough to scale it?

Document exactly what the task is. “Summarise client performance” is too vague. Different people will produce entirely different summaries. “Produce a one-paragraph summary showing: (a) revenue vs. budget (in GBP, with variance %), (b) top three expense categories, (c) one risk flag if any metric variance exceeds 15%” is clear. You could hand that to three different people and get roughly the same summary.

Why do clear inputs matter more than you think?

Define the input. What data does the skill need? For example, “Client revenue figures” is not specific. “Client revenue from monthly management accounts, reported in GBP, for the past 12 months, excluding intercompany transactions and one-off items” is. Clear inputs matter because bad inputs produce bad outputs. Garbage in, garbage out. You can’t AI your way past that.

How do you know what good output looks like?

Your firm defines the success of a good output. What does a good output look like to you? Is it a table? A list? Does it need to be copy-edited, or can it go straight to the client? Is there a tone guide (should it be formal, conversational or advisory)? These aren’t small details. They’re the difference between a skill that scales and one that needs heavy manual work afterward.

How do you test a skill before rolling it out?

Test it thoroughly. Run the skill on historical work that you’ve already completed and verified. Does it produce the same output you did? If not, why not? Does it miss nuance? Oversimplify? Get the numbers wrong? If so, refine the skill, run it again, and each cycle makes it better. You’re calibrating the AI to your practice’s standards.

Why does documenting the skill matter as much as building it?

Documentation trains your team, guides edge cases, and becomes your reference for onboarding. Write down what the skill does, how to use it, what inputs it needs, what output to expect, and what to do if something goes wrong.

This content may have been developed with the assistance of AI tools. Content is reviewed by a member of the Bright team before publication. Guidance on legislation can be obtained from your industry body.