Autopilot discovery: how Prio learns your workflows without you building them

Every automation tool on the market starts the same way: a blank canvas. You pick a trigger. You configure conditions. You wire up actions. Then you test, debug, and maintain it when something changes upstream.

This works for obvious, high-volume workflows — "when a form is submitted, add a row to a spreadsheet." But the automations that would save founders the most time are the ones they never think to build. The invoice you forward to your accountant every time it arrives from the same supplier. The follow-up email you draft after every client meeting. The weekly report task you create every Friday morning without realizing it's a pattern.

These micro-workflows cost 3 to 5 hours per week in aggregate. They're too small for Zapier, too contextual for templates, and too invisible to formalize. Until now.

What Autopilot Discovery does

Prio already sits across your email, calendar, tasks, and contacts. Every time you approve an email draft, reject a calendar suggestion, create a task, or ask Prio to triage your inbox — that's a behavioral signal. Autopilot Discovery is the engine that analyzes those signals over time, detects repeating patterns, and surfaces them as suggested automations.

Here's what that looks like in practice.

You've forwarded invoices from acme.com to your accountant 12 times in the last month. Prio notices and surfaces: "You forwarded 12 emails from acme.com to accountant@firm.com. This could be automated." One tap, and it becomes a live automation rule.

You create a task called "Weekly report" almost every Friday. Seven times in 60 days. Prio surfaces it as a recurring task suggestion with the frequency and day of week already filled in.

After meetings with external attendees, you consistently send a follow-up email within two hours. Nine times in the last month. Prio offers to draft the follow-up automatically after each meeting ends.

The key difference from traditional automation: you never designed these workflows. You just worked. Prio observed, measured confidence, and waited until the pattern was clear enough to suggest.

Why observation beats configuration

The fundamental problem with workflow builders is that they require you to know what to automate before you automate it. That sounds reasonable until you consider that the most valuable automations are habits you've never consciously examined.

Zapier has 7,000+ integrations. Make has visual flow builders. n8n gives you self-hosted flexibility. All excellent tools for known, repeatable processes between systems. But none of them can tell you that you email the same person after every board meeting, or that you archive emails from 15 newsletter senders within minutes of receiving them.

The approval queue is what makes this possible. Every interaction with Prio generates a labeled data point. Approved email drafts reveal who you communicate with and how. Rejected suggestions reveal your preferences. Edited calendar events reveal your scheduling constraints. This isn't a side effect of the product — it's the product's data flywheel.

This is also the lesson from the $400M+ AI assistant graveyard. Clara Labs, Fin, x.ai — they all tried to be fully autonomous from day one. They asked users to hand over email access and trust the AI to act correctly without supervision. Users didn't trust them, the products couldn't earn that trust fast enough, and all of them shut down.

Prio's approval model solves this differently. You start by approving everything. Over weeks, Autopilot surfaces the patterns you've been approving repeatedly and offers to handle them without asking. Trust is earned through observation, not demanded through onboarding.

What's coming: full inbox intelligence

Today, Autopilot learns from what happens inside Prio. Your chat interactions, action approvals, task creation, calendar follow-ups. This covers roughly 20% of your work behavior — the part that flows through Prio's interface.

The next phase extends visibility to 100%. Using the Gmail and Calendar access you already granted during setup, Prio will scan your sent folder, archive behavior, forwarding habits, and response times in the background. Thirty days of email history, analyzed once, patterns surfaced immediately.

No new permissions. No browser extension. No desktop app to install. The OAuth tokens from your initial connection are sufficient.

The result: on your first morning after the scan completes, before you type a single message, Prio already knows who you email most frequently, which senders you always archive without reading, and which contacts get replies within five minutes versus five days. Each of these becomes a suggested automation.

The compound effect

This is how Autopilot compounds over time.

Day one, you approve everything manually. Prio is your assistant, and you review every action before it executes. This feels manual, and it should — you're building the behavioral dataset.

By week two, the first patterns surface. Invoice forwarding. Recurring tasks. Morning routines. You activate the ones that feel right and dismiss the rest. Each decision refines the model.

By month two, you have 10 to 15 active automations running. Your Autopilot dashboard shows patterns discovered, hours saved, and automations active — all calculated from real usage, not projections. The more you use Prio, the more patterns it finds, and the more time it saves.

This is the moat that no workflow builder can replicate. Your approval history is a proprietary model of how you work. It's trained on your decisions, your contacts, your schedule, your communication style. It can't be copied by a competitor because it doesn't exist in a template library — it exists in your behavior.

Try it

Prio is live at prio.sh. Connect your Gmail and calendar, use it for a week, and see what Autopilot discovers. The patterns are already there. You just haven't noticed them yet.