AI Tools for Executive Assistants: What Earns a Place in the Workflow

Most roundups of AI tools for executive assistants read like they were written by someone who has never run a week on behalf of another person. Features get listed. What you never get is the Thursday where four things move at once, the principal is somewhere over the Atlantic, and two of those four involve people senior enough that a mistake costs real money.

I’m Nir Sabato, co-founder of Catch. We build an assistant that takes admin work off executives’ plates, and the executives are not the ones who teach me most about this category. The assistants are. They find out inside a week whether a tool carries work or just relocates it.

So this is written for you, not your principal. You know the job already. What follows is a way to judge the tools, sorted by the work instead of the vendor, with pricing and limits as they stand in 2026.

The three tests any tool has to pass

Every piece of executive assistant software demos beautifully, because a demo runs the happy path. The same trap shows up when comparing AI scheduling tools. One meeting. Two participants, a single time zone, nobody in transit. Your week looks nothing like that. So, three tests before anything gets a seat in your workflow.

Calendar and scheduling

This is where most assistants start, and it’s the most crowded corner of the category.

Booking links

Calendly is still the default. The free plan covers one event type, one calendar connection and unlimited one-on-one meetings, and Standard is $10 per seat per month billed annually, which unlocks unlimited event types and up to six calendar connections. Enterprise starts at $15,000 per year with a 50-seat minimum. There’s an AI assistant too, Callie, which coordinates inside an email thread and reads context to suggest times, plus in-app AI you can ask for action items from past meetings or a summary of the relationship history before a call.

Then the exception test. Calendly’s own documentation is clear that the conversational scheduling requires a pre-configured one-on-one event type and explicitly will not work with shared mailboxes. If you run an assistant@ or office@ address, as a great many of us do, that particular feature is simply off the table for you. Worth knowing before you build a process on it.

Email-CC schedulers

A newer shape: you copy an agent into a thread and it takes the back-and-forth.

Blockit is looped in by CC on any email or Slack thread, manages multiple Google and Outlook calendars at once, and provisions Zoom, Meet or Teams rooms. Two things make it unusually relevant to assistants. It is explicitly built to work alongside a human EA, handling routine scheduling while keeping the human admin’s visibility intact. And its autonomy is configurable along a range, from reviewing everything to full autopilot, which is the dial you actually want. It also picks up the kind of nuance you’d normally hold in your head: walk time between meetings, days reserved for calls, mornings kept clear after a red-eye. It carries SOC 2 Type II, and the trial runs 30 days with no credit card. On price, the ladder we checked on November 1, 2026 ran from a free tier capped at 15 meetings a month, through Lite at $999 a year with a 40-meeting monthly cap, to Infinity at $4,740 a year with no cap, and Enterprise from $20,000 a year. Monthly billing costs more than that on both self-serve tiers, so check the current figure and the cycle it belongs to before you budget.

Skej works the same way, CC on a thread, and sells texting as a named product feature. It connects to Google and Outlook calendars, Slack, Zoom, Google Meet, Teams video, text message, WhatsApp, Gmail, Outlook Mail, Apple Mail and Superhuman, and its booking link can chat live with a visitor to land a time. It keeps following up until the meeting is booked, and it’s unambiguous that there is no human in the loop.

Howie runs over email, iMessage and SMS. The Team plan starts at $25 a month with unlimited teammates, an instant-response iMessage agent, and proactive calendar audits, briefings, follow-ups and reminders. New accounts get $25 in free credits, and a custom assistant email at your own domain is $20 a month on top. Apply the volume test here deliberately: Howie is credit-based on a sliding scale, with monthly tiers running from 1,250 credits upward, so your real cost depends on how much you push through it rather than on a headline price. Its scope is scheduling and calendar work off the inbox.

Time-blocking and task scheduling

Reclaim, now owned by Dropbox, is free for one user on the Lite plan, then runs per seat per month billed annually: Starter at $10, Business at $15 and Enterprise at $22. It has a chat AI Assistant for planning and prioritizing, auto-schedules tasks from Asana, ClickUp, Todoist, Jira, Linear and Google Tasks, and protects travel time and airport travel time around calendar events. It’ll generate a follow-up email from connected Zoom meeting notes and send attendee reminders for scheduling-link bookings. What it won’t do is read, triage or reply from the inbox. Think of it as a layer between the executive and their own calendar rather than something that reaches out to other people.

Motion sits in the task and project management lane, pitched as a replacement for Asana or Monday. Pro AI is $29 per seat per month month-to-month and $19 billed annually; Business AI is $49 and $29. Individual rates run higher: Pro AI $49 monthly or $29 annually, Business AI $69 monthly or $39 annually. No free plan, and it holds SOC 2 Type II. There’s a forwarding address that turns an email into a scheduled task, plus travel-time buffers and an AI notetaker. If your organization’s problem is project tracking, it’s a serious answer. If your problem is the admin queue, it’s adjacent.

A word on vendor risk, which is the third test in its harshest form. Clockwise was one of the best-known names here for nearly a decade, with more than 8 million hours of focus time created and 23 million meetings moved. It shut down on 27 March 2026 and the team joined Salesforce. Every route on its old site now serves the same farewell notice. Any process you build inside a single-purpose scheduling tool is a process you may have to rebuild.

The inbox

Inbox work isn’t data entry. Contracts, negotiations and client decisions all happen in there, which is exactly why the overhead is so heavy and why it matters who carries it.

Superhuman is a four-product suite now rather than an email client. Mail is $25 per member per month billed annually ($30 monthly) on Starter and $33 ($40 monthly) on Business, with no free Mail tier. The Go, Docs and Suite ladder is separate: Free at $0, Pro at $12 annually ($15 monthly), Business at $33 ($40 monthly). Pro at $12 is likely the tier that matters to you. Its Email Assistant drafts replies, clears clutter and follows up, and runs on your existing Gmail or Outlook rather than requiring Superhuman’s own client. Auto Drafts, which Mail gates to its Business tier, produces a follow-up draft about an hour before a reminder you’ve set; and when an email asks for a meeting, Superhuman AI drafts a reply with suggested times and a booking link.

Fyxer has a Free plan with 400 monthly credits and limited features; automatic labeling and reply drafts start on Pro at $49 per person per month ($37 annually), and Team is $99 ($64 annually) with a two-seat minimum. Paid plans carry monthly credit allowances, and the advertised trial is 14 days. It sorts incoming mail into categories like To Respond, FYI and Marketing, and its notetaker joins video calls and produces shareable summaries.

Spark is built with shared inboxes in mind, which is why assistants keep finding it. Plus is $10 per user per month ($8.25 annually) with an AI assistant and 40 AI meeting notes a month; Pro is $20 ($16.58 annually) with unlimited meeting notes, read statuses, shared inboxes and CRM integrations. Several people can work a common address, assign messages, set deadlines and track progress without passing passwords around. There are also private team comments on a message, email delegation, and a Gatekeeper that lets you screen new senders before they reach the inbox.

Now apply the exception test, because it lands hard here. Superhuman’s AI produces drafts for you to review and send, not auto-sent mail. Fyxer generates responses for review and requires you to send them manually. Both are deliberate design decisions. But a draft isn’t a handled email, and at forty messages a day a review queue is still your queue. When you trial an inbox tool, count what left the queue, not what got written.

Meeting prep and notes

Notetakers are the easiest category to over-buy, because nearly everything ships one now. Fyxer’s joins the call and produces a shareable summary. Motion’s captures summaries, decisions and recap emails. Spark generates notes from calendar events. Lindy records, summarizes and extracts action items. Calendly’s in-app AI will pull action items out of past meetings.

None of that is the job, though. A transcript is a record. What the executive needs walking into the room is a brief: who’s in it, what was agreed last time, what’s outstanding, what decision is being asked of them. That brief lives across your notetaker, the CRM, the project tracker and three email threads, and a transcript on its own doesn’t assemble it.

So judge this category on whether anything closes that gap. If you’re collecting summaries into a folder nobody opens before a call, the tool is costing more than it returns.

Travel

Travel is where the happy-path problem shows up most clearly, because the booking is the easy part.

Navan covers flights, hotels, trains, rental cars and black car service. It stores traveller profiles including legal names, passport details, TSA PreCheck numbers and loyalty data so they populate future bookings, which saves real time if you book for several people. Its AI agent, Ava, works inside Navan’s own booking and support flows, handling refunds, cancellations, seat upgrades and disruption management. The travel side is free and commission-funded; expense is $15 per user per month after the first five monthly expensing users. One detail aimed squarely at assistants: delegates earn up to 1.5% of booking value on in-policy hotel stays booked for their travellers, and employees earn rewards by booking hotels below a displayed “Price to Beat” target, with points redeemable against personal hotel stays booked through Navan, not against flights, expiring a year after they are earned.

TravelPerk, which rebranded to Perk in November 2025, publishes travel-only tiers of Starter at $0 plus 5% per booking, Premium at $99 a month plus 3%, and Pro at $299 a month plus 3%, with per-booking fees subject to minimum and maximum thresholds that its North American pricing does not put a number on. Platform fees are per company per month. As of 29 October 2026 the North America panel also lists a Travel-and-Spend option at $6 per active user on Premium and $8 on Pro under a limited-time banner, so check whether that still stands when you look. For a US reader it’s worth knowing that TravelPerk integrates with the leading expense tools in the region, and its North America panel also lists its own promotional Spend plans from $9 per active user. It also runs an MCP server, so an external AI assistant can query trip status and spend, and prompt bookings, in natural language.

Concur publishes its pricing per expense report rather than per user: from $7 per report on Base and from $11 on Plus, both with unlimited users, with Premium quoted. The structural point to know before you shortlist it is that Concur Travel appears only in the Premium tier, and receipt scanning via ExpenseIt is on Plus and Premium only. No free trial on any tier. It applies travel policy during booking and flags out-of-policy options in real time, routes approvals through configurable hierarchies, and has an AI copilot called Joule plus a Booking Agent for compliant search and booking.

Engine, formerly Hotel Engine, charges nothing to use the booking platform itself: no membership fees, no contracts, no minimum spend, and it covers hotels, flights and rental cars with admin-set policies and budget controls. Direct Bill consolidates everything into one monthly statement, and it can cover hotel incidental holds upfront so they land on the corporate invoice instead of the traveller’s personal card, which removes a genuinely annoying piece of cleanup from your plate. Its FlexPro add-on is $299 a month or $2,999 a year billed annually, company-wide, and lets you cancel or modify eligible hotel bookings until noon on the check-in day. Read that one carefully before you lean on it: the remedy is a one-year travel credit rather than a cash refund, coverage is not universal, and eligibility has to be checked per booking.

The itinerary is the part all four are built to produce. The rest of the trip is the part to measure: whether the hotel is a seven-minute walk from the actual meeting address or just the one that looked closest on a map, who calls the front desk about the early check-in, and who rebuilds the whole thing when the Tuesday meeting slides to Wednesday.

Where a delegation layer fits

Everything above is software you bring into your own workflow, each piece covering one part of the week. Fine for the work that genuinely needs your judgment, and the wrong shape for the mechanical queue underneath, which is most of the volume and almost none of the value.

Catch is built for that queue. It’s an AI Executive Assistant, and it handles the traditional administrative workload itself, end to end: calendar work, email triage and drafting, scheduling, reminders, reservations, briefings and real bookings, done rather than drafted. It also flags what needs the executive’s attention instead of waiting to be asked. Blockit, by contrast, says outright that it is designed to work alongside a human EA. What software does not do is the physical half of the job: being in the building, running the event, the operational work that needs a person on the ground. Nobody needs to be let go for that to be true, and that’s the half the role moves toward.

Concretely, against the three tests:

• Volume. Flat $99 a month. No credits, no usage tiers, no per-call fees, and voice calls are included. There is a 10-day free trial. The subscription doesn’t move when your week gets heavy.
• The exception. Catch applies judgment rather than hard-coded rules. It does not run on “always follow up after two days.” It learns the executive, weighs what matters, resolves calendar conflicts by actually reaching out to the other party, and asks when a situation genuinely needs a decision. On travel, for eligible beta customers, it finds supported flights and hotels from calendar and inbox context, applies learned preferences, and books only after approval. It can call the hotel front desk, and when the trip moves it helps review supported changes, with rebooking also requiring approval.
• Maintenance. No workflow builder, no dashboard to learn. You connect Gmail or Outlook by OAuth, and it works with Google Calendar or Outlook Calendar. You talk to it in Slack, email, text message, iMessage or by phone, including voice notes. It can place outbound calls on your behalf the way you would, and it identifies itself as an AI agent on the call. It will not pick up your principal’s incoming calls, because an assistant doesn’t do that either.

Corporate travel booking is in beta for selected Catch customers. It covers supported business flights and hotels, with approval required before purchase or rebooking. Fares, room charges, and supplier change or cancellation fees are separate from the $99 per person monthly subscription. Ground transport may need separate arrangements. See the travel beta.

It integrates with Asana and Notion rather than replacing them, works with popular notetakers, and reaches HubSpot and Zoho. On security: SOC 2 Type II, CASA Tier 2, and data hosted in the US.

Worth naming what it isn’t, too, because two other shapes get recommended for this job. Builder platforms like Lindy and n8n hand you the components. Lindy is credit-based: the free plan carries $50 in credits that expire after seven days, and paid plans are billed per active user from $29.99 a month for 3,000 credits, topped up at $10 per 1,000, with actions pausing once the pool runs dry. n8n gives you Gmail and Outlook nodes, calendar triggers, Twilio for calls and SMS, a code node for custom JavaScript or Python, and cloud or self-hosted deployment. Both are capable. Both also make you the maintainer, which is the third test failing in the most expensive way an assistant can fail it. Catch ships with opinions and playbooks instead, and the autonomy it has is earned as it learns the executive rather than switched fully on at signup.

How to audit your own stack this month

1. List the last twenty things that cost you more than fifteen minutes. Not the tasks you think you spend time on. The actual ones, pulled from your sent folder and your calendar.
2. Mark each one J or M. J if it needed your judgment or your relationship with someone. M if it was mechanical and only needed doing correctly.
3. Count the M hours. That number is your delegation budget, and it tends to be bigger than people expect.
4. Check what each tool you pay for actually removed from the M column. A tool that produced drafts you still had to read and send did not remove the item.
5. Count your review queues. Every tool that outputs something for approval has created one. Three of those and you’re supervising software for a living.
6. Cancel one thing. The tool that survives the J/M test and the review-queue count is the one worth keeping. Most stacks carry at least one subscription that no longer earns its place.

Frequently Asked Questions