The EA AI Toolkit: 6 Tools That Actually Work Together (2026)

Most EAs trying to build an AI workflow have the same problem: the tools aren't broken, the handoffs are. You have ChatGPT for drafting, maybe Otter for notes, possibly a calendar optimizer — and they don't talk to each other. A meeting summary sits in one app. The action items live somewhere else. The follow-up never gets written.

Here's the upfront answer: the default working stack is ChatGPT Plus + Granola Business + Make Core — $46/month total after testing — and it covers drafting, meeting capture, and cross-app automation in one connected loop. The best free starting point is Make's free tier plus Granola Basic plus Notion Free, which is enough to validate one workflow before spending anything. What to avoid first: buying a second calendar optimizer before your first one has explicit rules for breaks, VIPs, and immovable blocks. That's where AI scheduling destroys EA trust fastest.

The rest of this article shows how the pieces connect and in what order to build them.

The Loop Every EA Runs — and Where It Breaks

Every EA operates the same five-stage cycle: intake and triage, calendar and time protection, meeting capture, follow-up and task routing, then documents and knowledge. Tools fail not because they're individually weak but because no one designs what passes between stages. A notetaker that doesn't export to your task system creates a second inbox. A calendar optimizer that doesn't know your priority rules moves the wrong meetings.

According to a 2026 survey of 600 EAs by Vimcal and CCing My EA, nearly 30% identified scheduling and calendar management as their single biggest time drain. Yet most AI tools were designed for individual email or personal productivity, not the EA-to-executive operating rhythm. Meanwhile, 86% of AI-using EAs are already on ChatGPT — but a tool isn't a system.

The design principle that makes a stack actually work: AI proposes, EA approves. Nothing external goes out, gets booked, or gets published without a human gate. Build that into every workflow from the start.

The 6-Tool Stack and How They Connect

ChatGPT Plus

This is the anchor most EAs already have, and it's the most misused. Its real job in a connected stack isn't open-ended web browsing or general research — it's taking structured input (a meeting summary, a briefing template, an email thread) and producing a reliable draft or classification. That's where it earns its $20/month.

The free tier hits message limits fast on a heavy EA day, so Plus is the functional floor for daily use. One integration worth knowing: ChatGPT can receive context from Granola via MCP, which means you can ask it "what did we decide about X on the March 15 call?" without copy-pasting anything. The honest limitation is security: never feed it board materials, executive compensation data, or personal travel details. Those stay out of consumer AI tools entirely.

Granola

This is the strongest underdog recommendation in the stack. Unlike Otter or Fireflies, Granola doesn't join your call as a bot — no awkward "AI is recording this meeting" announcement. It detects upcoming meetings from your calendar automatically, captures in the background, and exports notes directly to Notion or Slack.

The free Basic tier gives you AI notes and limited history. The Business tier at $14/month unlocks unlimited meeting history, advanced integrations with Notion, Slack, HubSpot, Zapier, and Affinity, plus MCP connectivity so Claude or ChatGPT can query your past meetings by date, topic, or participant. That last feature is where meeting capture becomes meeting memory.

The honest limitation: Granola's desktop experience on macOS and Windows is where the calendar sync works most reliably. Heavy mobile users may find Otter a better fit for their capture habits.

Notion

Notion is where Granola notes land, where briefing templates live, and where action items get tracked across sessions. The free tier handles unlimited pages for a solo EA. The critical integration: Make can write new Notion database items triggered by external events, and ChatGPT can enrich Notion content through Make. If your organization runs on Microsoft 365, SharePoint and Microsoft To Do may already fill this role — don't force a Notion migration on top of a working environment. In that case, Notion stays personal for the EA's own prep rather than becoming a shared team system.

Reclaim

Reclaim earns its place at the calendar stage through its Slack integration, which is genuinely unusual: your agenda, task list, scheduling links, and conflict alerts all live in Slack, so you're not context-switching to manage the executive's day. The free Lite tier — one user, one calendar sync, basic focus time, habits, and Slack — is enough to test whether it actually helps before spending anything.

The essential setup step most people skip: encode your break rules, VIP categories, and immovable blocks before letting Reclaim optimize anything. Without that policy, it will move the wrong meetings, and that's how AI scheduling erodes trust with executives. One market note: Clockwise shut down in March 2026 when its team joined Salesforce, and Reclaim was acquired by Dropbox in 2026 — its long-term independence is worth monitoring.

Reclaim's Lite tier is free; Starter is $10/month.

Make

Make is the connective tissue the rest of the stack depends on. It handles the handoffs no native integration covers: new Granola note → create Notion page → notify task owner in Slack → hold for EA approval before anything goes external.

The free tier allows 1,000 credits per month, with each module action consuming roughly one credit. A four-step workflow running on 50 meeting notes per month uses around 200 credits — well within free limits for testing. The Core tier at $12/month unlocks unlimited active scenarios and minute-level scheduling (free runs every 15 minutes minimum, which matters for time-sensitive follow-ups).

The design rule that prevents most automation disasters: Make moves data; it does not make decisions. Every workflow should end with an "EA Review Required" state before anything reaches the executive or an external contact. Zapier is the functional alternative — slightly broader app catalog, starting at $19.99/month for multi-step workflows — and the right call if your organization is already invested there.

Make's free tier is the place to start.

Grammarly

Two sentences, because that's all it warrants here: Grammarly sits at the end of every external draft as a voice and tone check — not because AI-generated writing is illiterate, but because maintaining the executive's specific communication style across hundreds of messages requires a quality gate. The 44% of AI-using EAs who use it, per ASAP's 2024 survey, are mostly using it for exactly this purpose.

The data flow in one paragraph: Calendar event → Granola captures meeting → notes auto-export to Notion → Make creates a task with owner and due date → Slack notifies the owner → EA reviews and approves → ChatGPT drafts the follow-up from the Notion summary → Grammarly checks tone → EA sends.

Three Recipes to Build First

Recipe 1: Meeting reminder Zap. Trigger on a Google Calendar event 10 minutes before start, filter for internal 1:1s or team syncs only (not external client calls), send a Slack DM to the executive. Setup time: under 30 minutes with Make or Zapier free tiers. Viva's EA team documented a 30% improvement in on-time meeting attendance after implementing this exact pattern. If the executive doesn't use Slack, the same pattern runs through Microsoft Teams.

Recipe 2: Meeting-to-knowledge pipeline. Granola finishes a meeting and exports the summary and action items to a Notion page via its native connector. Make then creates a database task with the owner, due date, and source link, and Slack notifies the owner. The EA reviews the Notion page first and corrects any misattributed action items before the notification fires. This is the workflow Viva documented with a reported 50% reduction in follow-up writing time. Setup: 45-60 minutes split between the native Granola-to-Notion connection and a two-to-three-step Make scenario.

Most operators at my level manage tasks. I build the systems that make tasks unnecessary.
— Julliene Bullecer, Chief of Staff & Executive Assistant to CEO, Aureus Financial

Recipe 3: Inbox classification and draft. A new email triggers ChatGPT to classify by topic, urgency, required action, and proposed owner. The classified draft lands in a review folder — never sent automatically. The EA reviews, edits, approves, sends. The Zonos "Inbox Broom" built on this exact pattern and saved three to five hours per week per executive, according to Zapier's June 2025 case study. The non-negotiable safety rule: auto-sending stays off until the EA has reviewed at least 100 classifications and confirmed the error rate is acceptable. Executive-level email is not where you discover the model categorized a board member as "routine inquiry."

What the Stack Costs at Three Budget Levels

Free core ($0/month): ChatGPT Free + Notion Free + Granola Basic + Reclaim Lite + Make Free. This is a validation stack, not a production one. Make Free runs on a 15-minute minimum interval; Reclaim Lite caps at a one-week scheduling range. The purpose of this tier is to prove one recipe genuinely reduces your cycle time before spending anything. Two weeks with Recipe 1 at $0 is a better buying argument than any product review.

Mid-range working system ($34-46/month): ChatGPT Plus ($20) + Granola Business ($14) = $34. Add Make Core ($12) for multi-step automations = $46. This is where the stack becomes an operating system rather than a collection of experiments. Unlimited Granola meeting history means your knowledge base becomes durable. Make's minute-level intervals mean handoffs are actually timely. Your existing email and calendar stay authoritative — no migration required. Annual billing typically saves 20-33% across these tools.

Premium delegation layer ($100+/month): Lindy Plus at $30/month adds a conversational agent layer across Slack, Gmail, Drive, Calendar, and Notion. The specific feature that makes it appropriate for EA work: it requires explicit approval for any action with external impact — sending an email, updating a ticket, posting to a channel. That approval-required design is not incidental; it's the boundary that makes delegation safe. ChatGPT Pro at $100/month is only justified for EAs doing research-heavy executive briefings. Most EAs don't need it.

Where Human Judgment Stays Irreplaceable

AI scheduling cannot know that a recurring one-to-one is politically sensitive. It cannot decide whether an email that looks routine is actually a board member checking in. It cannot repair a strained vendor relationship with the right tone. These aren't AI limitations to wait out — they're permanent features of the EA role that define its value.

The permanent gates: no automatic external sends, no AI-initiated bookings without approval, no confidential data into consumer tools. A 2026 EA community survey found 86% of respondents believe AI will enhance their role rather than replace it — but that outcome depends on EAs governing the handoffs rather than surrendering them.

The counterexample worth keeping in mind: a C-level EA with 20+ years reported trying ChatGPT, Copilot, and Glean and finding all three gimmicky. She could already write emails, build agendas, and schedule without help. That's not a failure of AI — it's evidence that the stack only matters when there's a real bottleneck, not because AI is trendy.

It's not just about the tools, it's about building the systems that make transformation stick.
— Radia Rerhaye, Executive Assistant to the President, PIPSC IPFPC

Connor Sheffield's arc at Zonos is the better model: he started as an EA with zero automation experience, inherited 15 Zaps, learned by comparing failed runs to successful ones, and became Head of Business Automation in under three years without a technical background. The EA who designs and governs the handoffs becomes more valuable, not less.

Where to Start

If scheduling is your bottleneck, build Recipe 1 with Reclaim Lite and Make Free. Encode breaks, VIP categories, and immovable blocks before letting anything optimize the calendar.

If meeting follow-up is your bottleneck, connect Granola Basic to Notion Free to Make Free and build Recipe 2. Get the pipeline working before buying anything.

If inbox volume is your bottleneck, use ChatGPT Free and Make Free for Recipe 3 — but start with a single inbox label and a narrow category list, not the whole mailbox.

If Microsoft 365 is your organization's standard, audit what Outlook rules, Microsoft To Do, Teams, and Copilot already do before adding external tools. The native environment may solve 60% of the problem.

One thing worth watching over the next year: agent tools are moving fast on approval-flow design, and the question to ask any new agent product is "what happens when it wants to send an email?" If the answer is "it sends it," don't deploy it for EA work.

Recommended Tools & Resources