Internal Comms & Meetings

How to prepare an all-hands deck with AI

Internal Comms & Meetings3 AI tools7 steps6 friction points

Want an agent to handle this on your live data? See the Starch version →

An all-hands deck is a recurring deliverable that almost every operator underestimates. You need to distill what happened across the business — revenue, headcount, product milestones, roadmap — into a coherent narrative that lands for people ranging from engineers to finance to a part-time contractor. It has to be honest, motivating, and short. Most founders spend a half-day on it the night before.

The workflow feels like an AI problem because so much of it is structural: take a pile of facts and shape them into a story with a clear arc. You already know what you want to say — you just need help organizing it, writing the transitions, and cutting the noise. A good LLM should be able to take bullet points and turn them into readable slides in under an hour.

ChatGPT, Claude, and Gemini are genuinely useful here. They can draft slide outlines, rewrite dense bullet points into cleaner language, suggest narrative structure, and generate speaker notes on demand. If you bring the raw material — metrics, wins, callouts — they'll shape it into something presentable. The friction isn't in the drafting. It's in getting the raw material assembled in the first place.

Skip the manual work → See the Starch version

Internal Comms & Meetings3 AI tools7 steps6 friction points

AI walkthrough

How to do it with AI today

A practical walkthrough using ChatGPT, Claude, and other off-the-shelf LLMs — what they're good at, what you'll have to do by hand.

Tools that work for this

ChatGPTClaudeGemini

Step-by-step

1 Before touching an LLM, pull your source data manually: export last month's revenue from Stripe, grab headcount from your HR system, screenshot your OKR tracker, and collect a handful of customer quotes or product updates from Slack or Notion.

2 Open Claude or ChatGPT and paste in your raw data dump — revenue numbers, product milestones, team updates, anything you've collected. Ask it to identify the five or six most meaningful things that happened this period and suggest a deck structure around them.

3 Ask the LLM to write a first-draft narrative for each section — company performance, product, people, roadmap — keeping each section to three to four bullet points maximum. Paste your raw numbers directly into the prompt so it can write around the actual figures.

4 Prompt the model to generate an 'opening context' slide: one paragraph that sets the tone for the all-hands and frames the period honestly. Give it your company stage, headcount, and one sentence about what the period felt like.

5 Use the LLM to write speaker notes for slides that contain sensitive or nuanced content — for example, slides about a missed target or a team change. Ask it to draft the talking points you'd actually say out loud, not just what's on the slide.

6 Open Google Slides or PowerPoint and manually build the deck using the LLM's outline and copy as your source material. Copy-paste section by section, adjusting formatting as you go.

7 Run a final pass: paste the full deck script back into the LLM and ask it to flag anything that sounds inconsistent, overly negative without context, or missing a key business area you typically cover.

Prompts you can copy

Here are last month's business metrics: [paste data]. Draft a 7-slide all-hands deck outline covering company performance, product progress, team updates, and Q2 priorities. Keep each section to 3 bullets max.

We missed our MRR target by 12% this month. Write a 3-sentence opening for an all-hands that's honest about the miss without being demoralizing, and sets up a positive framing for what we're doing about it.

Write speaker notes for a slide titled 'Product Update' that includes these three milestones: [paste milestones]. The audience is our full 22-person team, mix of technical and non-technical.

Here is my all-hands script draft: [paste draft]. Does this cover the standard areas a company our size should address in a monthly all-hands? What's missing or underexplained?

Turn these raw Slack updates from department leads into clean, readable bullet points for each section of our all-hands deck: [paste updates]. Each section should have no more than 4 bullets.

Reality check

Where this gets hard

The walkthrough above works — until your numbers change, the LLM hallucinates, or you have to re-paste everything next month.

No live connection to your actual data — you manually export Stripe revenue, copy headcount from your HR tool, and paste Slack updates every single time you run this.

The LLM has no memory of last month's deck, so consistency in tone, structure, and framing requires you to paste in the previous version as context on every run.

Metrics drift between drafts — if you paste slightly different numbers or forget an update, the model works with what it has and you catch the error only in the meeting.

Slide formatting is entirely on you. The LLM gives you text; you still spend 30 minutes in Google Slides making it look presentable before the 9 AM call.

Any sensitive context — a departure, a missed target, a pivot — has to be manually flagged and explained. The model can't know what it doesn't know from your raw paste.

Nothing persists after the session. Next month is the same manual assembly process from scratch, with no built-in record of what changed or what you said last time.

Tired of the friction?

Starch runs the whole workflow on live data — no copy-paste, no hallucinated numbers, no re-prompting next month.

See the Starch version →

Starch alternative

The same workflow on Starch

Starch is an agentic operating system. For this workflow, that means an agent builds a persistent app connected to your live business data — so when it's time for all-hands prep, the content is already assembled, not waiting for you to copy-paste it together.

Connect Stripe and Plaid once — Starch syncs your actual revenue, burn, and balance data on a schedule, so when you start building your deck, the numbers are already current, not pulled from last Thursday's export.

The Investor Reporting app pulls live financial metrics, formats them with charts and narrative summaries, and tracks tone consistency with previous updates — the same muscle applies directly to an all-hands build.

Connect Notion or your internal docs from Starch's integration catalog — the agent queries your team wikis, OKR trackers, and project pages live when it's building the deck context, so product and roadmap updates come from the source.

Describe the deck you want in plain English — 'a monthly all-hands covering revenue vs. target, top product wins, headcount updates, and Q2 priorities, formatted for a 30-minute company meeting' — and the agent builds a persistent app that generates it on that cadence.

Meeting Notes captures decisions and updates from your weekly leadership syncs and archives them in searchable history — so when all-hands prep time comes, the past month's key moments are already documented and ready to pull in.

Presentation Agent (currently in development — request beta access) will take the assembled content and produce a polished, on-brand slide deck automatically, so the last manual step in this workflow closes too.

Get closed-beta access →

Toolkit