AutoPilot

Scheduled Content Generation

AutoPilot is Pilot's scheduling system. Instead of generating articles one at a time by clicking through the console, you set a cadence — daily, weekly, or a custom schedule — and Pilot generates content automatically. The articles draw from your knowledge base, follow your voice configuration, and publish to your channels.

Two modes determine what happens after generation.

Supervised Mode

In supervised mode, AutoPilot generates articles and places them in your review queue. Nothing publishes without your approval.

This is the default, and it's the right starting point for most teams. The workflow:

1. AutoPilot runs on your schedule
2. Pilot identifies the best content opportunity from Pilot Lights
3. Pilot generates an article — sourced, cited, voice-configured
4. The article appears in your review queue
5. You review, edit if needed, approve or reject
6. Approved articles publish to the assigned channels

Supervised mode gives you the throughput benefits of automated generation — you're not manually triggering each article — while maintaining full editorial control over what publishes. You're reviewing, not writing. The bottleneck shifts from "find time to write" to "find time to review," which is a much lighter editorial load.

Autonomous Mode

In autonomous mode, AutoPilot generates and publishes directly. There's no review step. Articles that meet your quality threshold go live on their configured channels.

Autonomous mode is for teams that have:

• Tuned their voice configuration until the output consistently matches their standards
• Built a knowledge base deep enough that source material is strong across their topic areas
• Set a minimum quality threshold they trust
• Accepted that some articles may need post-publication editing

The quality threshold is a score that Pilot assigns to each generated article based on source coverage, citation density, voice adherence, and content coherence. You set the minimum — articles below the threshold are held for review instead of publishing automatically.

Autonomous mode isn't for everyone, and it's not for day one. Start supervised. Switch to autonomous once you've reviewed enough articles to trust the system's judgment matches yours.

Setting Up AutoPilot

Configuration happens per channel. You might run AutoPilot in supervised mode on your web channel (daily cadence, articles queued for review) and skip it entirely on your social channels (where you prefer manual selection).

Settings include:

• Cadence — how often AutoPilot generates: daily, weekly, or a custom schedule
• Mode — supervised (queue for review) or autonomous (publish directly)
• Min score — the quality threshold. Higher means stricter. In supervised mode, low-scoring articles are flagged but still appear in your queue. In autonomous mode, low-scoring articles are held instead of published.
• Channel — which publishing channel receives the output

Each AutoPilot run is tracked. The console shows run history: when it ran, which article was generated, the quality score, and the cost in tokens. You can see at a glance whether AutoPilot is producing good content and what it's costing.

Cost Awareness

Every AutoPilot run consumes tokens — your tokens, since Pilot uses the BYOK model. The Analytics page shows AutoPilot costs alongside manual generation costs, broken down by channel and time period.

A daily cadence means roughly 30 articles per month. At typical token costs, that's a manageable expense for most organizations, but it's worth monitoring — especially when you first turn on AutoPilot and haven't yet dialed in the voice configuration (articles that get rejected and regenerated cost double).

The console shows a cost estimate for your current AutoPilot configuration before you activate it, so there are no surprises.

Best Practices

Start supervised. Review your first ten to twenty AutoPilot-generated articles carefully. Tune the voice configuration based on what you see. Only consider autonomous mode after you've built confidence in the output.

Set a realistic cadence. If your knowledge base is small, daily generation will exhaust the strongest content opportunities quickly and start producing thinner articles. Match the cadence to your knowledge base depth. Weekly is a reasonable starting point for most teams.

Use min score intentionally. A high threshold means fewer articles make it through, but those that do are consistently strong. A low threshold means more volume but more variance. There's no universal right answer — it depends on your audience's expectations and your tolerance for post-publication editing.

Monitor cost. Check the Analytics page weekly when you first enable AutoPilot. Understand the per-article cost and the monthly run rate. Adjust cadence or voice settings if costs are higher than expected.

Feed the knowledge base. AutoPilot's quality correlates directly with the depth and freshness of your knowledge base. Upload new documents regularly. The best AutoPilot operations are backed by active knowledge bases, not static ones.

When Not to Use AutoPilot

AutoPilot is wrong for:

• Early-stage tenants with a small knowledge base. Generate manually until you have enough source material for consistent quality.
• Highly sensitive content where every word needs legal or compliance review. Use supervised mode and a rigorous review process, or generate manually.
• Low-volume operations where one or two articles a week are sufficient. Manual generation with Pilot Lights is simpler and gives you more control.

AutoPilot is right for content operations that need volume and consistency: trade publications with daily article needs, newsletters with weekly cadences, organizations that want to keep their web presence fresh without dedicating writing staff to every piece.

For how to generate your first article manually before turning on AutoPilot, see Getting Started. For how content reaches its audience after generation, see Channels and Publishing.