Using AI at work: drafting emails and summaries within company policy

TUREFF• •0

AI chatbots can save real time at work: a first draft of an awkward email, a summary of a long thread, a cleaner version of meeting notes. They can also cause real problems if confidential information ends up where it should not, or if an unchecked error goes out under your name. Here is how to get the benefit safely.

Check your company's policy first

Before you paste anything work-related into a chatbot, find out what your employer allows.

Many organisations have an AI policy covering which tools are approved and what data may be used with them.
Some provide an approved business version with contract terms that keep your data out of training and under the company's control. Use that rather than a personal account.
Some ban public chatbots for any work content, or for certain teams such as legal, HR or finance.
Clients may have their own rules. A client contract can forbid sharing its information with third-party services.

If there is no written policy, ask your manager or IT team. Using a personal account for work data without permission can breach confidentiality rules even when nothing goes wrong.

What never to paste in without approval

Customer or employee personal data.
Unreleased financial figures, plans or pricing.
Contracts, legal advice or HR cases.
Passwords, keys or internal system details.
Source code or documents marked confidential.

When in doubt, anonymise: replace names with roles and remove figures that are not needed for the task.

Drafting emails

Chatbots are good at structure and tone. Give them context, not secrets.

Say what the email must achieve: decline a meeting politely, chase an overdue invoice, announce a delay.
Describe the reader and relationship: a senior client, a colleague in another team, a supplier you know well.
Give the key points in bullet form.
Set the tone and length: brief and friendly, formal, no more than five sentences.
Edit the draft yourself. Remove phrases you would never say, check every fact and make sure promises are ones you can keep.

Watch for a tell-tale style: over-enthusiastic openings, repeated thanks and generic filler. Colleagues notice, and it can make you sound insincere. Cutting it back usually improves the email.

Summaries and notes

Summarise long documents you are allowed to share with the tool, then read the original sections behind any point that matters.
Turn rough meeting notes into actions: ask for a list of decisions, owners and next steps.
Ask for the questions a document leaves open as a quick review aid.

Summaries can drop exceptions, conditions and numbers, which are often the important part. Treat them as a map, not the territory.

Other useful tasks

Rewriting technical text for a non-specialist audience.
Drafting a first outline for a presentation or report.
Suggesting spreadsheet formulas and explaining what they do.
Translating rough drafts, then having a fluent colleague check anything important.
Brainstorming questions to ask in an interview or negotiation.

You remain responsible

Whatever the tool produced, you are the one sending it.

Check facts, names, dates and figures before anything leaves your desk.
Do not let it decide on people. Hiring, performance and disciplinary judgements need human review and often have legal rules.
Be transparent where expected. Some workplaces and clients want to know when AI has been used in a deliverable.
Keep copies of final versions, not just the chat history, in the proper company system.

A sensible first step

Pick one repetitive task, such as weekly status emails, and try using an approved tool for it for a month. Note how much time it saves and how much editing it needs. That gives you an honest picture before using it more widely. How does your workplace handle AI tools? Share what works, and what does not, in the comments.