Think about everything you have learned about your executive over time. You probably know when they want a text instead of an email, how much context they want before a meeting, whether they prefer three options or one recommendation, and whether they want to see an early draft or something nearly finished.

AI is becoming another working preference worth discussing.

Microsoft’s 2024 Work Trend Index found that 75% of knowledge workers were already using generative AI at work, while other workplace research found that many employees still operate without clear organizational guidelines for AI use.

For an EA, that creates very practical questions. 

A short conversation now can answer those questions before you build more AI into your workflow.

1. Speed vs. Polish: Where does your executive want a quick first draft, and where do they expect something researched, verified, and polished? AI can reduce production time, but context, fact-checking, judgment, and refinement still require human attention. Research shows that AI’s effectiveness can vary considerably depending on the task.

2. Human Review: Decide what you can review independently and what should always come back to your executive. External communications, legal or financial commitments, sensitive personnel matters, and other consequential work may require additional review.

3. Information Boundaries: Know which tools your organization approves and what information can and cannot be entered into them. This is especially important for EAs who regularly handle client information, calendars, contracts, financial details, personnel matters, strategy, and private executive correspondence.

4. Verification: Agree on when AI-generated information needs to be checked. AI can produce incorrect facts, statistics, quotations, links, and citations very convincingly, so important factual information should be verified against reliable sources.

5. Transparency: Discuss when your executive wants to know that AI materially contributed to the work. Using AI to refine a routine internal email may be very different from using it to develop research, analysis, or an external-facing deliverable.

6. Permission to Act: Agree on where AI is allowed to take action and where it should stop for human approval. Drafting an email, suggesting a calendar change, or preparing an update is different from sending the email, moving the meeting, or changing a record. Decide which actions AI can take independently and which should always come back to you or your executive first.

You do not need another 20-page policy. Your organization may already have policies governing AI, security, privacy, and confidential information. What you need with your executive is a practical understanding of how those rules translate into your everyday work together.

Bring these questions to your next 1:1. Keep this as a working agreement rather than something you create once and forget. Your tools, company policies, responsibilities, and your executive’s comfort level may change over time.

Once you start this conversation, your executive may have questions of their own. Here are a few ways to frame your answers.

👔 Executive: If you’re using AI, shouldn’t this take much less time?
🤝 EA: AI can help me get to a first draft much faster. I still need time to check the context, verify important information, and make sure the final work reflects what you need. We can identify the routine tasks where AI should genuinely shorten turnaround time.

👔 Executive: Do you need to tell me every time you use AI?
🤝 EA: We can agree on a threshold. Routine assistance may not need to be flagged every time, while research, analysis, sensitive work, or something AI materially helped create may be something you want visibility into.

👔 Executive: How do I know the information I give you is safe in AI?
🤝 EA: I only use company-approved tools and follow our policies on what information can be shared. There are also some types of information I won’t put into AI at all. If I’m unsure, I’ll check before using it.

👔 Executive: Could AI send something or change my calendar without me knowing?
🤝 EA: Only if we intentionally set up a workflow that gives it permission to take those actions. My preference is to start with AI preparing the work and keeping a human approval step before anything consequential happens.

👔 Executive: If AI can do more of this work, what does that free you up to do?
🤝 EA: Ideally, less time gathering information, formatting, and starting from scratch, and more time anticipating what you’ll need, spotting issues, following through, and focusing on the work that needs judgment and context.

Microsoft 365 Copilot just got much more “EA‑friendly.” Delegates with calendar access can now search another user’s meetings directly in Copilot Chat, and Copilot in Outlook can reason over an executive’s full inbox, calendar, and enterprise data - not just a single thread - making it easier to draft replies, prep for meetings, and surface key emails on demand. 

Cowork agents can now run full workflows on your own data. Microsoft’s new Copilot Cowork capability lets you (with IT approval) build multi-step workflows - like “every Friday, pull this SharePoint list, update the Excel dashboard, and draft an email summary to the exec team” - so the AI actually does the admin work, rather than just suggesting text. 

Meta’s AI is becoming more of a true digital assistant. Meta AI can now connect to Google Calendar and Gmail to schedule meetings, read calendar events, draft replies, and manage appointments, pushing consumer assistants closer to an EA-style scheduling and inbox helper

Strong EAs spend time learning how their executives work. You learn what they value, how quickly they expect you to move, where they want to stay involved, what they trust you to handle independently, and what a finished deliverable should look like.

As AI becomes more integrated into that work, those expectations need to include AI too. A ten-minute conversation about speed, review, information boundaries, verification, and transparency can give both of you clearer parameters for using AI confidently and responsibly. It can also make future automation easier because you already know which decisions can move quickly and which ones require human involvement.

Before you automate more, agree on the rules. Then build from there.

Have you talked with your executive about your AI working rules yet? Hit reply and tell us what you agreed on, or which question you still need to ask.

The Base Team