The All-Hands Meeting Guide: How to Run One People Actually Watch

The all-hands is the most expensive meeting your company runs, and usually the least examined.

Do the arithmetic once and it becomes hard to unsee. Sixty people, one hour, once a fortnight. That is sixty hours of paid attention every two weeks, roughly a full-time role, spent on a single recurring event. Nobody would approve a new headcount with that little scrutiny of what it produces. Yet most all-hands agendas are inherited rather than designed, and most companies have never asked the only question that matters: what is this meeting for, and would we notice if it stopped?

Some companies would notice immediately. Their all-hands is where strategy gets explained, where decisions get their reasoning attached, and where people can ask the leadership team something uncomfortable and get a real answer. Others would not notice at all, because their all-hands has quietly become a sequence of departmental status updates delivered to an audience of muted rectangles, most of whom are answering email.

This guide is about turning the second kind into the first. It covers what an all-hands is actually for, what belongs in the agenda and what should be a document instead, how to handle questions when the honest answer is bad news, how the format has to change for a distributed team, and what to do about the people who could not attend — which, in any company spanning more than one time zone, is a substantial fraction of the audience.

It assumes you are running the meeting or about to inherit it, and that you would rather fix it than cancel it.

What an all-hands is actually for

The failure mode is almost always the same: an all-hands that has become a broadcast of information. Sales reports their number. Engineering lists what shipped. Marketing shows a campaign. Each department gets its slot, each slot gets filled, and the meeting ends having transmitted a lot of facts that could have been read in four minutes.

Information transfer is a bad reason to gather everyone at once, because documents are strictly better at it. A written update can be read at the reader's pace, skimmed by people it does not concern, searched later, and consumed across time zones without anyone waking at five in the morning. If the entire content of your meeting could be an email, the meeting is a worse email.

What a synchronous gathering does that a document cannot is narrower, and much more valuable:

It creates shared context at the same moment. When the whole company hears the same explanation simultaneously, there is no version drift. Nobody gets the manager's paraphrase of the strategy shift three days later. Everyone can refer back to the same thing, and knows everyone else heard it too. That common knowledge — not just knowing something, but knowing that others know it — is genuinely hard to manufacture any other way.

It lets people see the reasoning, not just the decision. A document can state that the company is deprioritising a market. It struggles to convey what the decision cost, what was argued, and how confident the leadership team actually is. Watching someone explain a hard trade-off, including the part they are uncomfortable about, communicates something a written summary flattens out.

It makes leadership answerable in public. Someone can ask a difficult question and everyone watches the answer. That is a small, real accountability mechanism, and it is the first thing to disappear when the all-hands becomes a series of prepared segments with questions squeezed into the final four minutes.

It marks time. Companies need moments that punctuate the work — a launch shipped, a milestone hit, someone leaving well. Rituals are not decoration; they are how a group builds a sense of continuous shared history rather than an undifferentiated stream of tickets.

Everything in the agenda should earn its place against those four functions. If a segment is not creating shared context, exposing reasoning, enabling accountability, or marking time, it belongs in a document.

The agenda that survives contact with reality

A working all-hands has a shape that is stable enough that people know what to expect and flexible enough that it does not become liturgy. Here is a structure that holds up, sized for roughly forty-five minutes.

Open with the number that matters most (5 minutes)

Start with the single metric that best describes company health, and show it in the same format every single time. Revenue, active customers, runway, retention — whichever genuinely governs your decisions. One number, its trend, and a sentence about what changed.

The consistency is the point. When the same chart appears in the same place every time, people learn to read it, and they notice movement without being told. When the headline metric changes every month, nobody builds a mental model and the number becomes decoration.

Show it even when it is bad. Especially when it is bad. A company that only presents its metrics in good months trains everyone to read silence as catastrophe, which is worse than the truth almost every time.

Explain one thing properly (15 minutes)

This is the heart of the meeting and the segment most often crowded out. Pick one topic — a strategy shift, a market change, why a project was cancelled, what a new competitor means — and explain it in genuine depth. Not a summary. The actual reasoning, the alternatives considered, the trade-off accepted, and what would have to be true for the decision to be wrong.

One topic, fifteen minutes. The temptation is always to cover four things in fifteen minutes, and the result is four things nobody understood well enough to act on. Depth on one subject beats coverage of many, because coverage is what documents are for.

The most valuable version of this segment is the one where leadership shows uncertainty honestly: we think this, we are not sure, here is what would change our minds. Teams are far better at handling acknowledged uncertainty than most executives assume. What they handle badly is false confidence that later collapses.

Show the work, do not report it (10 minutes)

Instead of each department reading a status list, pick two or three pieces of work and let the people who did them show them. A demo of the feature. The actual customer call recording. The design that got rejected and the one that replaced it.

Two rules make this segment work. First, the person who did the work presents it, not their manager — the detail and the pride are both better. Second, rotate deliberately, so that over a quarter the whole company gets seen, including the functions that never have a demo. Finance explaining how pricing actually works, or support walking through the most common reason customers churn, is often the most useful ten minutes anyone spends that month.

Answer questions, honestly (15 minutes)

The single biggest difference between an all-hands people attend and one they merely join is what happens here. This section covered properly is worth more than everything before it, which is why it should never be the segment that gets compressed when you run late.

Collect questions in advance, in writing, with the option to submit anonymously. Advance collection is not about control — it is about answer quality. A question submitted two days early gets a considered answer with real numbers. A question raised live gets whatever the executive can produce under pressure, which is usually a generality.

Then answer the hard ones first. If four people asked about layoffs, address layoffs before the question about the office coffee. Ordering by comfort is transparent and expensive: everyone can see which questions you skipped, and next time they stop asking.

Some answers will be "I do not know" or "I cannot discuss that yet." Both are fine, and both are much better than a non-answer dressed as an answer. "I cannot talk about that until the deal closes, and I will tell you the week it does" preserves trust. Fifteen words of corporate fog destroys it, and people can tell the difference instantly.

Close by marking something (5 minutes)

End on the human material. Welcome the people who joined. Name a piece of work that went well and say who did it. Acknowledge someone leaving, properly, if they are.

This is the segment most likely to be cut for time and the one that does the most quiet work over a year. It is also where the tone of the whole company gets set — whether recognition is specific and earned, or generic and therefore worthless.

What does not belong

Equally important is the list of things that should never take up all-hands time:

Departmental status updates. These are documents. If sales needs to report weekly numbers, write them down where anyone can read them. Reading a list aloud to sixty people is a spectacularly inefficient distribution mechanism.

Detailed project plans. The four people who need the plan should have a meeting. The other fifty-six are being held hostage to a discussion they cannot contribute to.

Anything requiring a decision from a subset. If the meeting reaches a point where six people are debating and everyone else is watching, the meeting has failed. Take it offline — genuinely offline, with a follow-up, not as a euphemism for dropping it.

Slides that are read aloud. If the text is on the slide and the speaker is reading it, delete one of them. Usually the slide.

Compliance and administrative announcements. The expense policy changed. Someone put a new form in the drive. These are notifications, not events. They belong in whatever channel your company uses for notifications.

A useful test: for every agenda item, ask what would break if it were a written update instead. If the honest answer is "nothing, but people would not read it," the problem is that your written updates are not trusted, and the all-hands is being used as a workaround for a distribution problem. Fix the distribution problem. Using the most expensive meeting in the company as a delivery mechanism for things people should be reading is a very costly patch.

Running it for a distributed team

Everything above holds for a company in one building. Distribute the company and several things break simultaneously, and most all-hands formats never adapt.

There is no good time, so stop pretending there is

If your team spans more than about six hours of time zones, there is no slot that works for everyone. Someone is joining at seven in the morning or nine at night, and if that is always the same group, you have quietly made them second-class attendees.

Two workable approaches. Rotate the pain: alternate between a time that suits the western zones and one that suits the eastern, so the inconvenience is shared rather than assigned. Or decide the meeting is optional and make the recording genuinely first-class — which is more honest, and much harder to do well, because it means the recorded version must be as useful as the live one rather than a consolation prize.

What does not work is the default: schedule it for headquarters, tell everyone else the recording is available, and then treat the recording as an afterthought. That arrangement tells people in other regions exactly where they sit in the hierarchy, without anyone having to say it.

Presence has to be deliberate

In a room, presence is free. On a video call, it is a decision. Sixty muted rectangles is not a meeting, it is a broadcast with extra steps, and everyone knows it.

A few things help. Ask presenters to keep cameras on and leave it optional for everyone else — mandatory cameras produce compliance, not engagement. Use the chat deliberately rather than tolerating it: invite reactions, questions and disagreement there, and have someone whose actual job during the meeting is to read it and surface what matters. And call on people by name occasionally, warned in advance, so participation is not left entirely to whoever is most comfortable interrupting.

The stage needs to reflect what is happening

This is where most video-conference all-hands quietly fail. A grid of equal tiles is the right layout for a discussion between eight people and the wrong layout for a company presentation. When one person is explaining strategy, the audience should be looking at them, not at a mosaic that includes a rectangle of someone's ceiling.

What you want is control over the stage: a spotlight layout when a single person is presenting, so their video and their screen dominate; a podcast layout when two or three people are in conversation and equal weight is correct; and the ability to move between them as the meeting changes shape. Presenters should be able to share a screen without a fifteen-second scramble, and hand the stage to someone else without everyone having to work out who is speaking.

This sounds cosmetic and is not. The layout is what tells fifty distracted people where to put their attention, and attention is the entire commodity being spent.

Questions need a written channel

In a room, someone raises a hand. On a video call with sixty people, the social cost of interrupting is high enough that only the confident do it, which means you systematically hear from the same handful of people every time.

A written question channel, open before and during the meeting, with anonymous submission available, fixes this more reliably than any facilitation technique. The questions you get will be better and they will come from a much wider slice of the company. Anonymity in particular is worth defending, even though it occasionally produces something unpleasant: the questions people will not put their name to are frequently the ones most worth answering in public.

The recording problem

Every all-hands should be recorded. Almost every company records them badly, and the failure is usually not technical.

The common version: someone hits record, the file lands in a drive folder with a name like all-hands-may-14-final-2.mp4, and it is never watched. Six months later nobody can find the meeting where the pricing change was explained, so the explanation effectively did not happen for anyone who missed it.

A recording is only useful if three things are true.

It is findable. The recording should live with the meeting it came from, not in a general file dump. When someone opens last month's all-hands, the video should be there, alongside the agenda and the questions that were asked. If finding it requires knowing a naming convention, it will not be found.

It captures the right thing. A recording of the meeting is not a recording of one participant's webcam. It should capture the stage as the audience saw it — the presenter, their screen, and whoever else was speaking — which requires the recording to follow the same layout logic the live meeting used. This is the part that most often goes wrong in practice, and it goes wrong silently: nobody discovers that the recording captured the wrong speaker until they try to watch it.

It carries the surrounding material. The questions asked, the answers given and the links shared belong with the video. Much of the value of an all-hands is in the discussion around it, and a bare video file loses all of that.

Get those three right and the recording becomes genuinely first-class: someone in another time zone can watch it the next morning and be as informed as everyone who attended, including seeing what was asked. Get them wrong and the recording is a compliance gesture that lets you tell yourself the meeting was inclusive.

One more habit worth building: publish a written summary alongside the recording, ideally within a day. Five bullet points covering what was decided, what changed and what to expect next. Many people will read the summary and never watch the video, which is a perfectly good outcome — they got the information, at a cost of ninety seconds rather than forty-five minutes. The summary is also what makes the meeting searchable in a year, when the video is a wall of unindexed audio.

Frequency, and why fortnightly usually wins

Weekly all-hands are almost always too frequent. There is rarely enough genuinely company-wide news to fill a weekly slot, so the meeting fills with the departmental status updates it should not contain, and attendance decays into background noise.

Monthly is usually too sparse. A month is long enough for a strategy shift to have been fully absorbed through rumour before it is officially explained, which is precisely the situation an all-hands exists to prevent.

Fortnightly tends to be right for most companies under a few hundred people. It is frequent enough to be the place news lands first, and sparse enough that each one has real content. Beyond a few hundred people the calculus shifts, because the meeting stops being a conversation and becomes a broadcast — at that point many companies move to monthly and lean much harder on written communication in between, which is the correct adaptation.

Whatever cadence you pick, protect it. A meeting that gets cancelled when the leadership team is busy teaches everyone that it was never important, and it is precisely the busy periods — the ones with hard news — when people most need it to happen.

How to tell whether yours is working

Attendance is a weak signal, because attendance is often compulsory in practice even when it is optional in theory. Better indicators:

Are questions being asked, and by how many different people? A healthy all-hands generates more questions than it can answer, from a wide slice of the company. If the same four people ask everything, or nobody asks anything, the meeting is not being experienced as a place where asking is safe or useful.

Do the hard questions get asked in the meeting or in private afterwards? If people bring difficult subjects to their manager one-to-one rather than raising them in the all-hands, that is a direct measurement of how safe the public forum feels.

Does the recording get watched? If almost nobody watches it, either the meeting is not valuable or the recording is not usable. Both are worth knowing, and they call for different fixes.

Can people explain the strategy back? Ask a few people privately, a couple of weeks after you explained something important, what they understood. The gap between what was said and what landed is the real output of the meeting, and it is usually wider than the person who presented believes.

Would people opt in? The most direct test: make attendance genuinely optional for a quarter and see what happens. Some companies find attendance barely moves. Others discover they have been running a mandatory meeting nobody valued, which is painful but considerably better than not knowing.

Common failure modes, and what fixes them

The meeting is a wall of updates. Fix: move every status report to a written document and reclaim the time for the single-topic explanation. Expect resistance from department heads who see the slot as visibility; the answer is that showing real work in the demo segment is better visibility than reading a list.

Questions get four minutes at the end. Fix: move questions earlier, or protect the slot absolutely. If you consistently run out of time for questions, you have too much content, and the content is less valuable than the questions.

The same people speak every time. Fix: written question channel with anonymous submission, plus deliberate rotation of who presents work.

Remote attendees are second-class. Fix: rotate the schedule, or genuinely invest in the recorded version. Choose one and commit; drifting between them is the worst outcome.

Bad news never appears. Fix: this is a leadership discipline problem, not a format problem. The correction is to present the headline metric every time regardless of direction, and to answer the uncomfortable questions first. Companies where the all-hands only carries good news are companies where people learn the meeting is theatre, and after that no amount of format improvement helps.

Nobody can find what was said three months ago. Fix: keep the recording, the agenda, the questions and the summary together, attached to the meeting rather than scattered across a drive, a calendar invite and a chat thread.

Where Openbook fits

Openbook has a Townhall room built for exactly this meeting, and the design follows the argument above rather than trying to be a general video conferencing tool.

Sessions live in the room with their agenda and their questions attached, so the meeting is a thing with a history rather than a calendar entry and a link. Questions can be submitted in advance and upvoted, so what gets answered first is what the company most wants answered rather than what is easiest.

The live stage supports multiple presenters, with spotlight and podcast layouts you switch between as the meeting changes shape — one person explaining strategy, then three people in conversation, without the audience having to work out where to look. Screen sharing and handing over the stage are part of the flow rather than a scramble.

Only the people you designate as speakers can go live. Guests — a customer, an investor, an advisor — join through a personalised link with an expiry date, and when that link expires the guest is removed from the session automatically. You can revoke access at any time, and the link is visible under the guest's name in the edit screen for the inevitable moment when someone loses it.

Recordings are captured from the stage as the audience saw it, not from one participant's camera, and every past broadcast is listed on the session so someone in another time zone can watch the right thing without hunting through a drive. Because the Townhall room sits in the same workspace as your boards, docs and chat, the summary you write afterwards lives beside the work it refers to.

None of that makes a badly designed all-hands good. A meeting that is fifty minutes of status updates read from slides will be exactly as tedious on better software. But when the format is right, the tooling stops being the thing that gets in the way — and the people who could not attend stop being an afterthought.

The all-hands after bad news

Every company eventually has to run the difficult one: after layoffs, a missed quarter, a failed launch, a founder leaving. This is the meeting that determines whether the format has any credibility, and the usual instincts are all wrong.

Do not delay it. The gap between the news breaking and the meeting explaining it is filled entirely by speculation, and speculation is always worse than the truth. If people found out on Tuesday, Thursday is too late. Same day if you can, next morning at the outside.

Do not lead with reassurance. "I want to start by saying everyone's job is safe" is heard as a preamble to an announcement that not everyone's job is safe, and it burns the first ninety seconds — the only ninety seconds you have everyone's full attention. Lead with what happened, plainly, in the first sentence.

Give the reasoning, including the part that reflects badly on you. People are considerably better at accepting a hard decision than an unexplained one. A leadership team that says "we hired ahead of revenue and that was my call" retains more trust than one that attributes everything to market conditions, because everyone in the room already knows which is true.

Say what you do not know. After bad news, the questions that matter most are usually about the future: is there more coming, what happens to my team, what changes now. Some of those you genuinely cannot answer yet. Saying "I do not know, and I will tell you the moment I do" is a commitment people will hold you to — which is exactly why it works, provided you honour it.

Extend the question section, do not shorten it. The instinct is to keep the meeting brief and let people process. The effect is to signal that questions are unwelcome. Double the time, and stay until they stop coming.

Then follow up in writing the same day. People remember roughly a third of what they hear in an emotionally charged meeting. The written version is not a formality; it is how most of the room will actually absorb what was said, and how the people who were too upset to take it in can revisit it privately.

A company's all-hands is judged on this meeting, not on the ordinary ones. Everything you have built up over a year of well-run fortnightly sessions is either confirmed or spent in a single hour.

The short version

An all-hands is worth running when it does something a document cannot: create shared context at the same moment, expose the reasoning behind decisions, make leadership answerable in public, and mark the passage of time. Everything else in the agenda is a document wearing a meeting costume.

Structure it around one topic explained properly, real work shown by the people who did it, and a question section that is protected rather than compressed. Collect questions in writing and in advance, allow anonymity, and answer the hard ones first. For a distributed team, either rotate the inconvenience or make the recording genuinely first-class — and if you choose the recording, make sure it captures the stage, stays with the meeting, and comes with a written summary.

Then check, honestly, whether it is working: whether questions come from many people or a few, whether the difficult ones surface in public or in private, and whether anyone can explain the strategy back to you two weeks later.

The meeting costs a full-time role's worth of attention every fortnight. It is worth an hour of design.