Pilotpilot
Ask a question...
How It Works

How Pilot Works

Pilot works in five steps: upload your documents, Pilot organizes them into a knowledge base by topic, you configure your voice, Pilot produces cited content, and you review and deploy — whether that means publishing articles, shipping docs, or making knowledge searchable.

Five Steps, Then a Loop

Pilot turns your existing documents into working knowledge. Here's exactly how that works, from first upload to first deployment.

Step 1: Upload Your Documents

Start with what you have. PDFs, Word documents, markdown files, HTML pages, plain text — Pilot accepts them all. These are your source documents: the raw material everything else builds from.

There's no minimum. You can start with ten documents or ten thousand. There's no required format or structure. Pilot reads the content regardless of how it's packaged. Upload a folder of research PDFs, a collection of old blog posts exported from WordPress, a set of internal memos — whatever represents what your organization knows.

Each document goes through an ingestion queue. Upload is instant — you drag files and they're accepted in under a second. The actual reading and processing happens in the background. Pilot extracts the text, identifies the topics covered, and maps how each document relates to others in your collection.

Step 2: Pilot Organizes Your Knowledge Base

As documents are processed, Pilot builds your knowledge base — a structured representation of everything you've uploaded, organized by topic.

Topics aren't tags you assign. They emerge from the documents themselves. If you upload fifty articles about construction safety, Pilot identifies "fall protection," "scaffolding standards," "OSHA compliance," and other topics within that material. Upload a collection of market research reports, and topics like "consumer sentiment," "pricing trends," and "competitive landscape" appear.

The knowledge base isn't a folder structure or a search index. It's a map of relationships — which documents cover which topics, where topics overlap, and where your knowledge is deep versus shallow. This map is what allows Pilot to produce content that draws from multiple sources and synthesizes information across documents.

Step 3: Configure Your Voice

Before Pilot writes anything, you tell it how to write. This is voice configuration — a set of parameters that control the tone, style, and editorial posture of everything Pilot produces.

The key parameters, in plain terms:

  • Brevity: Should Pilot write concisely or expansively? A social media post needs compression. A research summary needs room to breathe.
  • Formality: Casual and conversational, or precise and academic? The answer depends on your audience.
  • Confidence: Should Pilot hedge ("research suggests...") or assert ("the data shows...")? Different contexts demand different postures.
  • Citation density: Light sourcing with occasional references, or heavy attribution with citations on every claim? A newsletter is different from a policy brief.
  • Tone: Warm, neutral, or authoritative? This shapes whether content feels approachable or commanding.

You can create multiple voice configurations for different channels. The same knowledge base might produce a casual newsletter and a formal research digest — different voices, same source material.

Step 4: Pilot Writes Articles

With a knowledge base and a voice configured, Pilot generates articles. This can happen in three ways:

Pilot Lights. Pilot analyzes your knowledge base and surfaces content opportunities — topics where your documents contain enough material to write something substantive. These appear in the console as suggestions. You decide which to pursue.

AutoPilot. Set a schedule — daily, weekly, whatever cadence fits your editorial rhythm — and Pilot generates articles automatically. In supervised mode, drafts queue for your review. In autonomous mode, Pilot publishes directly (only recommended once you've tuned the voice and trust the output).

Manual generation. Point Pilot at a specific topic or set of source documents and tell it to write. You have full control over what gets written and from which sources.

However an article is generated, the process is the same: Pilot identifies relevant source documents in your knowledge base, synthesizes their content, applies your voice configuration, and produces a draft with inline citations linking every claim to its source.

Step 5: Review and Deploy

Every piece of content Pilot produces goes through your editorial workflow. You see the full draft with citations. You can edit freely — change wording, restructure sections, remove claims you don't want to make. The content is yours to shape.

When you're satisfied, deploy to your channels:

  • Web channel: Content is available through Pilot's headless API for your website or documentation site to pull and display.
  • CMS webhook: Pilot pushes articles as drafts to your existing CMS — WordPress, Contentful, or any system with a webhook receiver.
  • Email: Formatted for newsletter distribution through your email service provider.
  • Social: Adapted for LinkedIn, X, or other platforms — same content, format-appropriate versions.
  • Ask: Your knowledge base becomes searchable — anyone can query it and get cited answers.

One piece of content can go to multiple channels simultaneously, each formatted for its surface.

The Loop

These five steps aren't a one-time setup. They're a continuous editorial process.

Upload new documents → Pilot identifies new topics and content opportunities → generate new content → review and deploy → repeat. Your knowledge base grows over time, and with it, Pilot's ability to produce substantive, well-sourced output.

The result is a content operation that runs at the speed of your knowledge production, not at the speed of your writing staff. When you publish new research, Pilot can turn it into articles within hours instead of weeks. When your archive grows, so does the range of content Pilot can produce.

What This Looks Like in Practice

Builder's Benchmark, a demonstration tenant, was built by feeding Pilot roughly a thousand open-access research articles about building science, construction safety, and housing economics. Pilot organized them into topic clusters, and from that knowledge base, it generated articles with three different voice configurations for three different audiences — an editorial voice for feature-length pieces, an advisory voice for practical guidance, and a briefing voice for quick-read summaries.

Every article on the site traces its claims to specific research papers. The knowledge base is visible. The editorial process is transparent. That's what Pilot does — and it works the same way whether your source material is academic research, corporate archives, or years of published reporting.

The same five steps power every output — articles, documentation sites, and searchable knowledge portals. The architecture is the same; the delivery channel is what changes.

Last updated March 5, 2026

Related