Introduction
In Five Cents Help Me Use AI at Work, we’ll turn AI into a practical assistant for everyday tasks, without handing over your judgment. We’ll cover choosing the right job, writing useful instructions, checking the result, and protecting work information. The best starting point is a task where a mistake is easy to catch and fix.
Think of AI as something that drafts
Think of AI as something that drafts, organizes, and suggests, rather than something that approves decisions. Turning rough, non-sensitive notes into a checklist is a reasonable first experiment. Letting a chatbot decide which employee deserves promotion is not. A useful task has a clear purpose, suitable input, and a result you can assess. If you cannot tell whether the answer is right, fluent writing will not make it reliable. Start with work you understand well enough to review.
That also gives you a sensible way to judge productivity claims. Workplace studies have found benefits in some settings, but not a universal improvement. In one customer-support study, average productivity rose by fourteen percent, with larger gains among less-experienced workers and little improvement among the most experienced.
That result describes a particular tool and workplace, not every profession. Your own test should include the time spent preparing instructions and correcting the output. A faster first draft is only useful if the finished work is better or takes less effort overall. Once you have a suitable task, give the tool five things: the task, the context, the material, the constraints, and the format.
Instead of asking it to write a project update, say: draft a concise update for colleagues about this delay, using only the notes below. Explain what changed, the known reason, the next milestone, and unresolved questions. Do not invent dates or assign blame. Keep the tone calm and finish with three actions. Mark anything the notes do not establish.
That instruction defines both the job and its boundaries. Read the draft, then request a targeted revision: shorten the background, make the actions clearer, or separate confirmed details from uncertainties. Asking for everything again usually gives you less control than correcting the specific weakness.
The same approach transfers across roles
The same approach transfers across roles. In customer support, supply an approved policy and ask for a draft response that stays within it. The human agent checks whether it fits the customer’s situation. In research, ask the tool to extract claims from supplied material and show which passage supports each claim. That is more useful than requesting an authoritative answer with references you have not checked.
In technical work, ask for an explanation of unfamiliar code or proposed test cases, then review and run them in an appropriate environment. For teaching, specify the learners’ level and ask for an explanation with practice questions. Check the answers and whether the language suits the audience. These are starting points to test, not automatic permission to use workplace data.
Before sharing anything, review the parts that could change someone’s understanding or action. Check names, numbers, dates, quotations, and policy details against the original material. Compare a summary with its source. Did it leave out an exception or turn a possibility into a commitment? Recalculate important figures independently. For code, test behavior and security, not just whether it runs.
You can ask AI to identify assumptions or missing information, but its self-check is not independent verification. Finally, read the result as the recipient would. Is the tone appropriate? Are the instructions clear? Could the wording unfairly characterize someone? The person approving the work remains accountable, even when the first draft came from a tool.
The easiest mistake is to paste work material into whichever chatbot is available. Check your employer’s approved tools and data-handling rules first. Do not submit confidential, personal, client, patient, employee, financial, or security-sensitive information unless that specific tool and use are authorized.
Removing a name may not remove identifying details
Removing a name may not remove identifying details. Use fictional material for practice, and share only what the task needs. Privacy protections vary by product, account, settings, and contract. A business privacy promise does not necessarily cover your personal account. Also avoid treating a general-purpose chatbot as an authority for hiring decisions, medical care, legal action, lending, or safety instructions. Those require authorized processes and appropriate qualified review.
When verification would take more effort than doing the task directly, choose the simpler method. Start small, give clear boundaries, and verify what matters. Before using AI, ask whether the task is suitable, the information is safe to share, and a human can check the result. To build on those habits, you can generate two new Five Cents: Better AI Prompts for Everyday Tasks, for more practice turning vague requests into useful instructions, and Checking AI Answers Before You Trust Them, for a deeper review routine.
And with that, you're up to speed in a few minutes.