# Founder focus: simple tasks, maximum autonomy

**ICP:** Founders and tiny teams who need **agents for straightforward jobs** (inbox triage, daily briefs, social drafts, follow-ups, site checks)—and want those agents to **run on their own** as much as safely possible, without babysitting.

This doc aligns product and messaging. Implementation details live in **[FOUNDER-PATH.md](./FOUNDER-PATH.md)** and **[PRODUCTION-TRUST-LOOP.md](./PRODUCTION-TRUST-LOOP.md)**.

---

## What “simple + autonomous” means here

| Simple tasks | Not (for this ICP) |
|--------------|-------------------|
| Summarize inbox, draft replies, schedule blocks | Open-ended “do my whole company strategy” |
| Post/draft social from a brief | Fully autonomous brand risk without guardrails |
| Daily admin / Jamie-style briefings | Unbounded tool chains with no eval |
| Website QA smoke checks | Replacing human judgment on money/legal |

| Maximum autonomy | Safety baseline |
|------------------|-----------------|
| **Schedules** — agent runs on a cadence | **Eval baselines** before you rely on it daily |
| **Webhooks / email-in** — triggers without you in the app | **Approval mode** for send-email / side effects when stakes are high |
| **Embed** — visitors talk to the agent 24/7 | **Outbound allowlists**, PII filters, preflight before publish |
| **Clear prompts + constraints** in the blueprint | **Support bundle + runbooks** when something breaks |

---

## Product principles (founder + autonomy)

1. **One agent, one job** — Name and goal say what it does; founder duplicates agents for another job (Jamie vs Robin vs Buzz).
2. **Prefer “run without me”** — Every founder-facing template should point to **schedule**, **webhook**, or **embed** where it fits, not only Test chat.
3. **Autonomy is earned** — Green eval + optional baseline **before** marketing “set and forget.”
4. **Dangerous actions default to safe** — Email send, calendar writes, external posts: **draft / approval** paths documented in UI copy and Jordan-style patterns.
5. **Fast recovery** — Failed runs visible, runbooks linked, support bundle one click.

---

## Messaging (use everywhere)

- **Headline angle:** *Configure once; your worker bee handles the repeat work—on a schedule, on your site, or from email/automations.*
- **Proof:** Founder packs + golden path + “trust loop” (eval baseline), not feature laundry lists.

---

## Near-term build priorities (suggested order)

1. **Activation UX** — **Shipped:** **Go live → Autopilot** callout on the main builder (`#activation-autopilot`): schedule, embed, webhooks/API/Zapier, email-in, with links to Quick actions and trust settings.
2. **Founder templates** — Each pack agent: short “Autopilot” blurb (what runs automatically, what needs approval).
3. **Eval nudges** — After first N successful runs, soft prompt: *Save a baseline so changes don’t break your autopilot.*
4. **Metrics** — Funnel: pack → embed/schedule set → recurring runs (use **[FOUNDER-ANALYTICS.md](./FOUNDER-ANALYTICS.md)** + `account_registered`).

---

## Out of scope (for now)

- Enterprise SSO, complex RBAC (unless selling upmarket).
- “General super-agent” positioning—confuses ICP and hurts trust.
- Adding tools without runbooks and failure modes documented.

---

**North star (founder + autonomy):** A founder has **at least one agent** doing a **simple, repeated task** on a **schedule or trigger**, with **guardrails** they understand—and they **trust** it because they **tested and baselined** it.