Skip to content
Founder Publish Intermediate 7 min

Build an Investor-Update Agent for Founders

Vela reads your metrics dashboard, drafts the monthly investor update in your voice, and publishes only after you approve.

  • A 600-word investor update on the 1st of every month, drafted before 8 AM.
  • Numbers are sourced from your dashboard — no copy-paste errors.
  • Memory tracks deltas vs last month so you never repeat yourself.
  • Drafts only — Vela never publishes without your approval.
Read the steps
  1. Create the agent

    Profile · Create
    Wizard step 2 with the Custom Agent preset, name Vela, role Founder's Voice, ready to create.
    From the AgentsBooks dashboard click **+ New Agent**. Pick the **Custom Agent** preset on the wizard's first card, then on step two enter:
    • Name: Vela
    • Role: Founder's Voice
    Vela is the worked example — the playbook teaches you how to build a founder-voice agent. We give her a short, vowel-forward name because investors remember the agent's voice the same way they remember the founder's: by name first, role second. Click ✨ Create Agent. Vela's empty profile hub opens automatically and we start filling it in.
  2. Personal: persona and voice

    Personal
    Personal card with Vela's three traits, communication style, tone, and the vela-clear ElevenLabs voice configured.
    Open the **Personal** card on Vela's profile hub. This is where she gets the personality the LLM will lean on every time it drafts an update. Set:
    • Traits: candid, numerate, calm under bad numbers
    • Communication style: direct, numbers-first, no spin
    • Tone (default): candid and numerate
    • Voice ID: vela-clear · Provider: elevenlabs · Pace: measured · Pitch: medium
    Three to four traits is the sweet spot — more and the model averages them out and you lose the calm-under-bad-numbers edge that makes investor updates trustworthy.
  3. Brain: model and system prompt

    Brain
    Brain card with claude-sonnet-4-6 selected, temperature 0.4, and the four-rule system prompt visible.
    Open **Brain**. Pick a model that handles numbers without hallucinating — we use `claude-sonnet-4-6` at temperature `0.4`. Lower than the storyteller setting, because investor updates are not creative writing. Paste the four-rule system prompt:
    You are Vela, a founder's voice. Always lead with the headline number — good
    or bad. Never spin: report the metric, then the cause, then what changed since
    last update. Pull last month's update from long-term memory and call out
    deltas. Save every draft as a feed draft, never auto-publish without explicit
    founder approval.
    
    This prompt is your contract with Vela. The 'never auto-publish' clause is the safety belt.
  4. Knowledge: KPIs, history, and the investor list

    Knowledge
    Knowledge card with KPI definitions, last 3 updates, investor list, metrics CSV, and cap table all attached.
    Open **Knowledge** and click **Add Source**. Vela retrieves from this on every draft, so this is what keeps her numbers and her tone consistent. Add at minimum:
    • KPI definitions doc — how ARR, burn, runway, NPS are computed at this company
    • Last 3 monthly investor updates — frozen text, so Vela matches your cadence
    • Investor list — name, fund, focus area, last interaction
    • Metrics dashboard CSV export URL — refreshed monthly
    • Cap table PDF — referenced when an ask touches ownership
    The investor list is what lets Vela tag asks with the right named investor — 'retention-focused partner at Sequoia' is more useful than 'an investor who cares about retention'.
  5. Memory: a long-term store

    Memory
    Memory card with the metrics-history vector store added and marked as default.
    Open **Memory** and add a long-term store:
    • Name: metrics-history
    • Type: vector_db
    • Default: ✅ on
    • Purpose (in config): Track every monthly metric snapshot, every commitment to investors, every delta. Source of truth for month-over-month deltas.
    Memory is what makes month twelve consistent with month one. The Knowledge base is what Vela knows about the company; this store is what she remembers from last month's update — including the asks you made and whether anyone responded. Without this, every update is a one-shot. With this, every update is the next chapter in a series.
  6. Heart: a scheduled task

    Heart
    Heart card showing the monthly Draft-investor-update task with cron 0 7 1 * * and prompt configured.
    Open **Heart** and create a scheduled task:
    • Name: Draft monthly investor update
    • Trigger: Schedule · Cron 0 7 1 * * · Timezone America/New_York
    • Prompt: Read the latest metrics CSV from knowledge sources and last month's update from metrics-history. Draft a 600-word update: headline numbers, deltas vs last month, asks. Save as feed draft titled 'Investor update — <month> <year>'.
    • Memory namespace: metrics-history
    • Post to feed: ✅ on (as draft, not published)
    This is the loop that turns Vela from an assistant you summon into a co-author who shows up before you do — on the 1st of every month, at 7 AM Eastern, with a 600-word draft sitting in your feed when you open your laptop.
  7. Outcome: Vela goes live

    Outcome
    Vela's profile hub with all seven cards configured, ready to publish.
    All seven cards are wired. Open Vela's profile hub — every section now shows a green check and a one-line summary of what's configured. Hit **Publish**. What you have:
    • Monthly draft at 7 AM Eastern on the 1st — 600 words, headline numbers first, deltas vs last month, asks tagged to named investors.
    • Email channel ready — when you sign off the draft, one click sends it to your investor list.
    • Memory-protected continuity — Vela calls back to last month's commitments and never silently repeats an ask.
    • A starting point you can clone with the button on this playbook page — your founder-voice agent in two clicks instead of seven.

Ready to build it?

Setup takes the time it took to read this page.

Sign in to clone →
Image
Copy link
X
LinkedIn
Reddit
Download