AI for Executive Assistants

What makes EA work hard The volume is the problem. A senior EA might manage 2-4 executives, each with 200-500 emails a day, 15-25 meetings a week, and a running list of projects that change hourly. Context switching alone burns 30% of the day. By 3pm most EAs are doing triage on triage, which is exhausting and leaves no space for the judgment-heavy parts of the job. AI helps in three specific places: inbox triage, meeting logistics, and research tasks. The rest of the job, reading the room, managing relationships, knowing which exec needs what when, stays with you because those are judgment calls that no model does well. ## The three high-leverage spots ### Inbox triage Most EAs spend 2-3 hours a day in their exec's inbox. An AI triage layer running on Gmail or Outlook can: - Sort into urgent, needs-reply, FYI, can-delete buckets based on sender, subject, and content

• Draft replies for the top 10-15 emails a day that need quick handling
• Flag anything matching custom rules (board members, key clients, time-sensitive requests)
• Surface threads that have gone quiet and need a nudge
• Catch emails that should have had a reply by now Time saved: typically 60-90 minutes a day per exec. For an EA supporting 3 execs, that is 3-4 hours of pure triage time back. The quality matters as much as the speed. A good triage agent learns your exec's patterns over 2-3 weeks. It notices that emails from a specific VP always need your exec's eyes, that board reminders are time-critical, that certain vendors always need a CYA reply within 24 hours. ### Meeting logistics automation Tools like Clara, x.ai, and Motion promised this but never quite delivered. A custom Claude agent does it better because it has full context: calendar, CRM, email history, travel prefs, room availability. It handles back-and-forth scheduling across time zones, books conference rooms, pulls prep docs, drafts agendas, and sends reminders. You review the outputs once before they go out. For complex scheduling (5+ people, cross-timezone, requires travel), AI cuts the back-and-forth from 6-12 emails to 1-2. Travel logistics is the same story. An agent that knows your exec's seat preference, hotel loyalty programs, dietary needs, and preferred flight times can book trips in 15 minutes that used to take 90. ### Research acceleration Pre-meeting research, competitive landscape scans, candidate backgrounds, travel logistics, vendor comparisons. A research agent with web access can turn a 45-minute task into a 5-minute review of a prepared brief. Works with Perplexity, Claude with web search, or a custom MCP-based setup that hits your internal CRM and external sources at the same time. Typical result: you spend the time you used to spend gathering on actually reviewing and editing the brief. Quality goes up, time goes down. ## Five to seven AI tools EAs use 1. Inbox triage agent running on Gmail or Outlook via API, trained on your exec's response patterns

2. Scheduling agent with calendar + CRM access, handles comms end to end for complex meetings
3. Pre-meeting brief generator pulling from email, LinkedIn, CRM, and news for context docs
4. Travel planner that books hotels and flights based on prefs, sends itinerary, handles changes
5. Expense coder that categorizes receipts from email and credit card feeds into the right buckets
6. Weekly summary bot that drafts the exec's Sunday-night "here's the week" email in their voice
7. Task tracker that watches Slack and email for commitments made and reminds on deadlines ## Tools in the stack - Claude for writing, reasoning, and drafting. Best at tone matching and nuanced replies.

• ChatGPT for quick tasks, image work, and general Q&A.
• Gemini if you are deep in Google Workspace and want tight Gmail and Docs integration.
• Zapier for simple cross-tool automations that do not need reasoning.
• Claude Agent SDK + MCP for the serious builds that touch email, calendar, and CRM with decision logic.
• Superhuman, Shortwave, Notion on the interface side.
• Calendly, Cal.com, Motion for the public-facing scheduling layer. For most EA builds, Claude handles the draft work (emails, briefs, agendas), and a custom agent built on top of the Agent SDK handles the multi-step stuff (book the meeting, update the CRM, send the confirm, add to the brief folder). ## Build vs buy Off-the-shelf EA-focused tools like Reclaim, Motion, and Clockwise handle scheduling well. Inbox tools like Superhuman AI and Shortwave handle triage decently. Starting there is the right move if you are a solo EA and your workflows are standard. Custom builds make sense when an EA is supporting multi-exec workflows, or when the exec has an unusual pattern that generic tools do not handle. Think: a founder who gets 400 cold emails a day and needs a highly specific filter, or a CEO whose board comms need to flow a certain way, or an EA supporting a CFO whose work is 60% numbers-based. Typical builds run -8K setup and -3K per month. For a senior EA supporting a busy exec, payback is usually 1-2 months in time saved. For EA teams supporting C-suite at a M+ company, the math works out even faster. ## Getting started Shane runs a 30-minute call with you and the exec you support. You map the top 3 time drags, agree on what the AI should handle vs escalate, and scope a 2-3 week build for the first piece. Most EAs start with inbox triage because the time savings are immediate and obvious. Within 3 weeks of launch, you should know whether it is worth expanding. The build process is collaborative. You know the exec's patterns better than anyone. Shane builds the system, you calibrate it over the first 2-3 weeks, and it gets better as it learns. By week 6, most EAs report it feels like having a second EA who never sleeps. ## What to measure in the first 90 days If your exec is investing in an AI build for you, here are the metrics to track: - Inbox triage time per day: should drop from 2-3 hours to 30-60 minutes within 30 days
• Meeting scheduling back-and-forth emails: should drop from 6-12 per complex meeting to 1-2
• Pre-meeting brief completeness: should be 100% briefed vs the typical 40-60%
• Research task time: should drop 70-80% for standard research
• After-hours work: should drop 30-50% for the exec These are specific, measurable, and move within 60-90 days if the system is built right. ## Common EA mistakes with AI First: treating AI like a replacement for judgment. It is not. Use it for the mechanical work (sorting, drafting, logistics) and keep the judgment work (knowing which battles to pick, when to push back, when to protect the exec's time) with you. Second: not calibrating enough upfront. The first 2-3 weeks of use are critical. Every time the AI drafts something that is off, you fix it and the system learns. EAs who skip this step end up with a system that is 70% right instead of 95% right. Third: not telling the exec what changed. The exec should know that emails from key people are still being read by you personally, and that AI is handling only the bucket of admin-level emails. Transparency prevents surprises.