Internal Comms & Meetings

How to write a weekly team update with AI

Internal Comms & Meetings3 AI tools6 steps6 friction points

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

A weekly team update is the recurring communication that tells your team — and often your investors or advisors — what happened last week, what's happening this week, and where things stand on the things that matter. For most operators running small teams, it's a 30-to-60-minute ritual of hunting through Slack threads, project trackers, and their own memory to reconstruct a coherent picture of the week before writing anything at all.

The workflow feels like an AI problem because most of the work is assembly, not judgment. You already know what happened — it's scattered across five tools and a dozen conversations. The hard part is pulling those fragments into a single structured narrative quickly enough that the update doesn't turn into a Sunday-night chore. Summarization, structure, and consistent formatting are exactly the kinds of tasks where LLMs visibly earn their keep.

ChatGPT, Claude, and Gemini can meaningfully help here today. Give any of them a raw dump of Slack messages, meeting notes, or bullet points and they'll produce a clean, readable update draft in under a minute. They're good at applying a consistent structure, adjusting tone, and flagging what's missing. The gap is everything upstream: the gathering, the connecting to live project data, and making the whole thing run without you manually assembling the inputs every week.

Skip the manual work → See the Starch version

Internal Comms & Meetings3 AI tools6 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 opening any AI tool, manually gather your raw material: skim your project tracker for tasks closed or moved this week, check Slack for notable decisions or blockers, pull any relevant metrics (revenue, pipeline, shipped features) from wherever they live, and paste everything into a single scratch document.

2 Open Claude or ChatGPT and paste in your raw material along with a structure prompt. Tell the model the audience (internal team vs. investors vs. both), the sections you want (wins, blockers, priorities for next week, metrics snapshot), and any tone guidance — e.g., 'direct, not corporate, under 400 words.'

3 Review the draft output and identify what's thin or missing. LLMs will generate plausible-sounding content for sections where you gave them nothing — watch for this. Go back to your sources, fill in the gaps, and run a second pass with the additions included.

4 If you have a standard template your team expects, paste that template into the prompt explicitly. Something like: 'Follow this format exactly: [paste your template].' This cuts the reformatting step after you get the draft back.

5 For the metrics section, paste the numbers in manually — current week's revenue, open deals, tasks completed, whatever your team tracks. Don't ask the LLM to retrieve them; it can't. Give it the numbers and ask it to write one or two sentences of context around each.

6 Once you have a draft you're happy with, paste it into your delivery channel — Slack, Notion, email, wherever your team expects it — and send. Nothing from this session carries forward; next week you start from scratch.

Prompts you can copy

Here's my raw notes from this week: [paste]. Write a weekly team update with four sections: Wins, Blockers, This Week's Priorities, and Metrics. Keep it under 350 words, direct tone, no corporate filler.

Turn these Slack messages and task completions into a brief weekly update for a 6-person early-stage team. Audience is internal. Highlight the two or three most important things and be honest about what's behind schedule.

I run a weekly update every Monday. Here's last week's update as a format example: [paste]. Here's this week's raw notes: [paste]. Write this week's update in the same structure and tone.

We shipped X, closed Y, and are blocked on Z. Metrics: MRR $48k (up $3k WoW), 4 deals in final stage, 2 critical bugs open. Write a 250-word team update that leads with the revenue number and ends with the two priorities for next week.

Here are my meeting notes from three calls this week: [paste]. Extract the key decisions made, action items assigned, and any blockers raised, then format them as the 'What happened' section of a weekly team update.

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 project tracker, Slack, or calendar — every run requires you to manually copy-paste the raw material, which is most of the actual work.

The model has no memory of last week's update. Maintaining consistent structure, recurring context, and voice across months means re-teaching the format every single time.

Hallucination risk is real in thin sections. If your notes are sparse on blockers or metrics, the model fills the gap with plausible-sounding placeholder content that you have to actively catch and delete.

Metrics have to come from you. Revenue, pipeline, task counts, deployment frequency — none of it is accessible to the LLM, so the numbers section is always a manual data entry step.

Nothing persists between runs. The template you carefully refined, the tone adjustments you made, the sections your team prefers — you're either saving prompts yourself or reconstructing them from memory each week.

Delivery is a separate step entirely. The LLM produces text; getting it into Slack, Notion, or an email thread is your problem, and it's not automated.

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 weekly update app that pulls from your live project data, meeting notes, and communications on a schedule, drafts the update automatically, and delivers it without you assembling the inputs by hand.

Connect Slack and Google Calendar as scheduled-sync providers — Starch pulls this week's events, decisions, and channel activity automatically, so the raw material is already assembled when the update runs.

Use the Meeting Notes starter app to capture and archive what happened in every call this week. When your weekly update runs, it pulls from that searchable history rather than from your memory.

Use the Project Management starter app to track tasks, priorities, and blockers in real time. The weekly update agent reads actual task status from your live board — closed, in-progress, and blocked — not a manual export.

Describe your update format once in plain English — 'Every Monday morning, draft a team update with sections for Wins, Blockers, Next Week's Priorities, and a metrics snapshot, then post it to the #team-updates Slack channel' — and Starch builds and schedules the automation.

The update runs on the same format, pulling from the same live sources, every week. You review and send rather than gather, paste, prompt, and format from scratch.

Connect Gmail or Outlook from Starch's integration catalog so external-facing updates — investor notes, advisor check-ins — can pull from the same weekly summary and be drafted and queued in your inbox automatically.

Get closed-beta access →

Toolkit