Investor Update Drafter

Founders skip investor updates because they take 2 hours to write each month. Two skipped updates and trust erodes; six and you're a mystery to your cap table. The agent handles the synthesis — pulls metrics from your trackers, summarises the month's wins/losses, drafts the email. You edit and send.

What it does

Maintains a running log of "shipped this month" entries (you tag as you go)
Pulls metrics from configured sources (Stripe MRR, GitHub PRs, user count)
Logs wins, losses, asks throughout the month
Drafts the monthly update from these inputs in your voice
Sends draft 5 days before send-target so you have edit time

What you'll need

Memory for monthly tagged entries
A defined "investor update format" matching your prior style
Configured metric sources (Stripe API, your DB query, etc.)
Cron

Setup

1. Define your format

Send to Flowly

Remember investor update format (tag "iu-format"): Subject: [Company] — <Month YYYY> update Greeting: 1 sentence — month summary in 5 words ## Numbers - MRR: $X (Δ from last month) - Users: N (Δ) - Specific KPI: ... ## Highlights 3-5 bullets, hardest first. Specific, not vague. ## Lowlights 1-3 bullets. Honest. What didn't work, what we learned. ## What's next 3 bullets — concrete next-month focus. ## Asks 1-3 specific intros, expertise, or feedback requests. Skip if nothing genuine. ## Cap table updates Only if changed. Sign off: brief, warm.

2. Capture as you go

Throughout the month, tag entries:

"shipped: plugin marketplace MVP — internal launch Tuesday" → memory tag "iu-month:<YYYY-MM>:shipped"

"learned: cold outbound to enterprise prospects with custom intros converts 3x better than mass email" → memory tag "iu-month:<YYYY-MM>:learned"

"loss: lost the @bigco deal, they went with a competitor. Reason: integrations breadth — they needed Notion + Slack + Salesforce day 1, we have 1.5 of those" → memory tag "iu-month:<YYYY-MM>:lowlight"

3. Configure metric sources

Send to Flowly

Remember metric sources (tag "iu-metrics"): - MRR: query Stripe API at /v1/subscriptions, sum active subs by plan, convert to MRR - Users: query my Supabase: select count(*) from auth.users where not deleted - Active users (7-day): SQL: count distinct user_id from events where ts > now() - 7 days Each metric: source, query, formatter, frequency.

4. Drafting cron

Send to Flowly

Cron "investor-update-draft" 5 days before each month-end at 9 AM: 1. Compute the time window (this month's entries) 2. Fetch metrics from iu-metrics sources, plus the same from previous month for delta 3. Pull from memory: - "iu-month:<this-month>:shipped" entries - "iu-month:<this-month>:learned" entries - "iu-month:<this-month>:lowlight" entries - "iu-month:<this-month>:next" entries 4. Draft the update following iu-format: - Synthesise highlights from "shipped" + "learned" - Lowlights directly from tagged entries - "What's next" from "next" entries plus my known pipeline - "Asks" only if anything was tagged "iu-month:<this-month>:ask" 5. Flag in the draft anywhere I haven't given enough context (e.g. "highlight needs a number — what was the impact?") 6. Save draft to memory "iu-draft:<YYYY-MM>" 7. Send to Telegram with: "Draft for <month> update is ready. 5 days to edit. Reply with edits or 'send to <list>' when ready."

5. Send

When ready:

*"Send the iu-<YYYY-MM> draft to investors-list (with my edits below):

Change MRR delta from +12% to +12.4%

Add to highlights: paying customer #50 closed

Drop the ask about hiring"*

Agent applies edits, sends via your configured email pipeline.

Tips

Tag immediately, not at month-end. "Shipped X" tagged the day it shipped is gold. Re-constructed at end of month from memory is vague.
Honest lowlights build trust. Investors who only hear good news distrust founders. The lowlight section is where credibility compounds.
Specific asks > vague asks. "Intros to PMs at fintech scale-ups" beats "happy to chat with anyone." Focus on something the recipient can actually do.
Don't send to all investors. Different stages care about different things. Pre-seed gets product detail; Series A gets metrics. Configure separate distribution lists if your cap table spans stages.
Voice consistency matters. Have the agent emulate your prior updates' voice. Show it 2-3 examples in memory once; future drafts match.
5-day edit window is the right length. Less and you don't have time; more and you procrastinate.
End the year with a yearly summary. December cron generates the year-in-review by aggregating 12 monthly updates. High-trust artefact for cap table conversations.