05/05/2026
There is a massive difference between using AI to work faster and using it recklessly.
As finance professionals, our entire career is built on accuracy, compliance, and confidentiality.
The tools are incredibly powerful, but if you use them without boundaries, you risk your professional reputation.
If you are integrating AI into your workflows, here are the 10 absolute "Don'ts" you must follow:
1. Do not upload raw, sensitive data.
This is the fastest way to breach professional confidentiality.
Never put real client names, TINs, or exact bank account numbers into a public AI tool. Always sanitize your data first.
Replace "Company X" with "Client A." Protect the client, then process the data.
2. Do not treat AI like a search engine.
AI is not Google. If you type, "How do I do a month-end close?" you will get a generic textbook answer.
You must build a command/prompt. Give the machine strict context, constraints, and a specific professional role.
3. Do not blindly trust the math.
Language models are built to predict words, not to act as calculators.
While their data analysis capabilities are improving rapidly, you must always cross-check the final figures.
I agree that Claude and GPT are heavy on computations. You still have to ensure accuracy of outputs.
4. Do not copy and paste without reviewing.
If left unchecked, AI has a very distinct, robotic tone.
It loves to use overly complex words that real humans rarely say, except for human dictionaries.
Always read the output and adjust it to sound like a natural, firm professional before sending it to a client or your CFO.
Better still, guide it to write like you.
5. Do not start with your most complex process.
Trying to automate an entire group consolidation on day one will frustrate you.
Pick one small, repetitive task first, like drafting a variance explanation or categorizing a large dataset of petty cash receipts. Master the small workflows first.
Always ensure that you protect yourself and your clients when using AI.