An AI chief of staff for CEOs who would rather decide than sort.

What an AI chief of staff is

A chief of staff protects a CEO’s attention. They read what comes in, decide what actually needs the CEO, prepare the context before each meeting, chase what is overdue and make sure decisions turn into work. Most founders never hire one: the role costs a senior salary and takes months to learn how the company runs.

An AI chief of staff does the same coordination in software. It is not a chatbot you visit when you remember to. It reads your mail, messages and projects on its own, keeps the list of what needs you current, and does the preparation and first drafts, so the part left for you is the decision. Agently is built around that job, with one rule underneath it: anything consequential waits for your approval.

What Agently does for you each day

Keeps one to-do list that manages itself

Agently sweeps the places work reaches you: your Microsoft 365 inbox, Slack mentions, Linear and Jira issues assigned to you, approvals waiting on you, tasks due or stalled, and routines that failed. It ranks the result by who is waiting, for how long, deadlines, money and urgency, and writes one line on why each item is there. Drag an item higher and it learns what you put first. When you reply outside Agently, or the issue closes at the source, the item closes too.

Triages the inbox

Newsletters, automated senders, bulk mail and marketing copy are removed before anything else happens. Mail from people you already correspond with is kept. What is left is judged once, in a batch, against your role and the company’s context: is this a request, a question, a cold pitch, an FYI, and does it need you or someone else? Dismiss a sender twice and they stay out. The method is the one in our guide to triaging an inbox with AI, run for you on every sweep.

Drafts replies you approve

When a real person asked you something, Agently drafts the answer in the thread, as an Outlook reply or a Slack reply, using what the company already knows about the sender and the subject. The draft waits as an approval card. Edit it, ask for a rewrite, approve it or skip it. Approving sends exactly the text you read. Nothing sends on its own.

Briefs you every morning

A morning brief is set up for you when you join. Each day it reads your connected mail, calendar, Slack, tickets, CRM and documents, then writes one page: today’s meetings with context on each, what is due, what changed overnight and what needs you. It never contacts anyone.

Prepares every meeting

With Google Calendar or Outlook connected, ask “prep me for today” or schedule it: who you are meeting, what you last discussed, open threads with them and what to bring up, from everything the company brain holds about each attendee.

Follows up on what is waiting

Every item on the list carries how long someone has been waiting on you, and stale threads rise until they are handled. Routines such as pipeline review and stale deals catch what has gone quiet outside your inbox, and any nudge they draft waits for your approval.

Runs routines on a schedule

Routines are standing questions with an answer delivered on time: an end-of-day wrap, Monday priorities, a KPI scorecard, a revenue report, an investor update draft. Each one returns a page, or an action held for approval.

Answers by voice

Jarvis, the agent behind all of this, takes voice requests: push to talk in the browser, or a wake word and shortcut in the Mac app. Ask what needs you or hand off a task while you walk between meetings. Delegated work runs in the background and the answer is read back when it is done.

AI chief of staff vs a human chief of staff vs a chatbot

The honest split: a human chief of staff is still the right hire for running a reorg, managing the board relationship or sitting in the room for you. Software is the better fit for the hours of sorting, reading, drafting and chasing that sit underneath those jobs. If you are choosing between tools rather than people, our comparison of AI executive assistants covers alfred_, Lindy, Motion, Reclaim and others.

How to set it up

1. Start free. Create a workspace at beta.agently.dev. The free tier needs no card.
2. Tell it your role. Onboarding asks what you do, reads your company’s site and starts a company profile, so ranking and drafts begin from how a CEO works.
3. Connect your tools. Outlook, Slack, your calendar, Linear or Jira, and whatever holds your documents and pipeline. Each connection is OAuth on your own account.
4. Set your line. Decide what it may do without asking. Everything past that line waits in approvals with the reasoning attached.
5. Work the list. Open the Canvas each morning. Approve, edit, dismiss or drag. Every action teaches the ranking and the sender rules, so the list gets quieter over the first weeks.

Data access and approvals

Your chief of staff reads a lot, so the boundaries matter more than the features.

• Scoped to your workspace. Workspaces are isolated from one another, and every read and search is scoped to yours. Workspace data is never used to train AI models.
• Your mail stays yours. Your inbox and calendar are read only through your own connection, and personal notes are never shared with teammates or with shared routines.
• It acts only with approval. Replies, sends and anything else past your line become proposals. Approving replays the action exactly as proposed; rejecting drops it. Every action it takes in a connected tool leaves an audit record.

The details are in security and privacy and approvals.

Pricing

Every plan includes Jarvis, the company brain, routines and MCP access. The free tier gives you 100 credits and one connected app. Starter is $39 a month with three connected apps, Pro is $99 with ten, and Enterprise is $199 with unlimited apps and members. Yearly billing gives two months free. See pricing for credits, storage and seats on each plan.