We came across a discussion thread recently where someone asked Executive Assistants a simple question: Are you using AI heavily at work, and what are you using it for? The original question mentioned email and scheduling, but the discussion quickly expanded into how EAs feel about AI more broadly.
The responses were all over the map. Some EAs use AI every day. Others use it occasionally for specific tasks. Some are learning how to use it because they expect the skill to become more important in their careers, even while remaining cautious about the technology. A few commenters were very clear that they would prefer not to use AI at all.
What made the discussion interesting was how specific people were about the boundaries they had created.
One EA described using AI to help with a complicated financial-data project that reportedly reduced months of work to a matter of weeks, yet she still wouldn’t use it to write an email because personal communication was something she wanted to own. Another EA uses AI for email polishing, research, and newsletters because it saves time during busy days. Someone else avoids using AI for factual work because checking the output for mistakes can sometimes create more work than it saves.
Those examples raise a useful question for EAs: Where does AI actually make your work easier?
For some people in the thread, the answer went well beyond drafting emails. EAs described using AI to plan offsites, edit presentations, organize complicated travel, capture meeting notes and action items, track down missing receipts, create workflows, and review calendars for scheduling conflicts.
These are also tasks where an EA already brings considerable context to the process. A calendar, for example, may show an open 30-minute window, but an experienced EA may know that his or her executive needs travel time, prep time, a break between meetings, or that a seemingly movable meeting carries more importance than its title suggests.
AI can surface information quickly. The EA still has to understand the context, weigh the options, and decide what to do with it. And that brings us to a term we came across this week.
AI Term of the Week: Meat Proxy
Meat proxy is a deliberately cheeky piece of AI-era slang for someone who takes an AI-generated answer and passes it along without really reading it, checking it, or adding any judgment of their own.
Think: someone asks a question in Slack, you paste it into ChatGPT, then paste the answer straight back into Slack. At that point, you haven’t really added much besides an extra step between the AI and the person who asked the question
Yes it sounds strange, but the idea behind it is useful for EAs.
AI should help you get to a better answer or complete work more efficiently. Your value isn’t simply moving AI output from one place to another. It is knowing the context, spotting what is missing, checking whether something makes sense, and deciding what actually needs to happen next.
The same principle is useful when evaluating any AI workflow
Suppose AI summarizes your inbox every morning. That sounds useful. But if you still need to open every important email, check whether the summary is accurate, reconstruct missing context, decide what requires action, rewrite the suggested responses, and manually complete every follow-up, how much time did the summary actually save?
One simple test: look at the work AI creates after it gives you an answer. If ten minutes saved by AI creates twenty minutes of checking, correcting, copying, and clicking for you, the workflow probably needs some work.
Where do you draw the line?
People can understand AI quite well and still have very different opinions about where it belongs in their work.
Some EAs are comfortable using AI for writing. Others only want help polishing something already written. Some use it for scheduling and calendar management, while others would rather keep those responsibilities manual.
Some are comfortable connecting approved AI systems to workplace tools, while others are understandably cautious about giving an AI system access to email, documents, calendars, or confidential information.
Security came up several times in the discussion, including reminders to use employer-approved tools and understand what type of account or data protections are in place before connecting workplace systems.
There probably isn't one workflow that makes sense for every EA.
Your executive, organization, industry, access level, responsibilities, and comfort with the technology all affect where AI can reasonably fit into your day. Knowing where you want AI involved, where you want human review, and where you would rather keep the work entirely in your hands gives you a useful framework as more AI features appear in the tools you already use.
AI Updates of the Week
Meta launched "Muse," a new personal AI agent that can do things like send emails, book travel, manage schedules, fill out forms, and make purchases on a user’s behalf. It is initially available in the U.S. through WhatsApp and a dedicated Muse app.
Microsoft 365 Copilot Cowork added new frontier model options: Anthropic’s Claude Fable 5.1 and OpenAI’sGPT‑6 Astra). Microsoft is also testing an ‘App skill’ that lets users create lightweight interactive apps by describing what they want in chat, although that feature is currently limited to its Frontier program. It should make it easier for EAs to create quick internal tools (trackers, intake forms, dashboards) without any coding.
OpenAI continued the rollout of GPT‑6 Astra to all ChatGPT Pro, Business, and Enterprise users, its most capable model yet, alongside a faster ChatGPT Images 2.5 tool with a new sketch feature for drafting visuals like posters and flyers.
Final Thoughts
The conversation around AI among EAs is clearly far from settled, and that may be a good thing.
Some people are using it every day, some are experimenting carefully, and others have decided there are parts of their work where AI simply doesn’t belong. Wherever you fall on that spectrum, being thoughtful about how you use AI matters. Pay attention to what actually saves you time, what creates more work, what requires your judgment, and what you would rather continue doing yourself.
You don’t have to love AI. But understanding what it can do, where it can help, and where you don’t want it involved puts you in a much better position to make those decisions for yourself.
We’d still love to hear from you!
Last week, we asked our EA community to share how you’re using AI, where it’s helping, where it’s falling short, and what you’d like to learn more about. Conversations like the one above are a good reminder that there is no single EA experience with AI, which is why we want to keep hearing directly from you.
If you haven’t had a chance to participate yet, we’d love your feedback and insights through our 2-minute survey. Your responses help us understand what EAs actually need from AI, rather than assuming which tools, resources, or skills will be most useful.
See you next week!
And please keep sharing what you’re learning with us. The more we hear from EAs doing this work every day, the more useful these conversations become for everyone.
The Base Team
