LinearPilot
AI Engine

You know where your data goes. The AI comes to you.

An AI engine wired into your business, with the providers named in writing. LinearPilot masks sensitive fields before the call and restores them in the answer.

How it works

Private AI that actually understands your business.

Three layers of intelligence — each designed to keep your data safe while giving you powerful AI capabilities.

1

Private AI Engine

An AI engine that reads your documents, CRM and data, and answers questions in seconds — with the providers it uses named on our Sub-processors page.

Learn more
LinearPilot AI
What was our revenue last quarter by product line?
Analyzed 1,247 invoices in 2.3 seconds
Enterprise Plan$847K
Pro Plan$523K
Basic Plan$198K
Total$1.57M
2

Use any AI model — without the risk.

LinearPilot holds the provider accounts, so there are no keys for you to obtain or rotate. Whichever model answers, sensitive fields — emails, phone numbers, card and ID numbers, addresses — are swapped for placeholders before your question leaves, then restored in the answer you see.

See how it works
Works with
AnthropicOpenAIGoogleGroqOpenRouter
1. You ask
"How much did Sarah Chen at Acme Corp pay us?"
2. LinearPilot masks sensitive fields
"How much did [CONTACT] at [COMPANY] pay us?"
3. AI answers safely
"[COMPANY] paid $47,200 last quarter."
4. You see the full answer
"Acme Corp paid $47,200 last quarter."
3

Enterprise-grade security

Sensitive fields are encrypted at rest with AES-256-GCM and masked before any request leaves. Data in transit is encrypted with TLS, and every AI interaction is written to an audit trail.

See the security controls
Security Dashboard
Tenant isolationScoped per company
Sensitive fieldsMasked before send
Two-factor sign-inOne-time codes
Data EncryptionAES-256
AI Audit LogsEnabled
100%
You know which providers see your data
3s
Average response time across your entire dataset
Any model
Use ChatGPT, Claude, Gemini — safely and privately
Under the hood

The AI Engine is opinionated on purpose

Named providers under no-training terms, sensitive fields masked before they leave, a strict tool registry, and explicit budgets per agent. Here is what ships in every deployment.

Model tiers
Named models, no training

All three LinearPilot tiers — Spark, Core and Max — are open on every plan, so moving up is a setting, not a rewrite. Every tier runs under provider terms that exclude your data from training, and the provider behind each tier is named on our Sub-processors page.

Claude
OpenAI
Gemini
External providers
Frontier models, on demand

Frontier models from named providers when a task needs more horsepower. LinearPilot holds the accounts, and you pick which tier answers. Every outbound call has sensitive fields masked and is written to the audit trail.

Knowledge Graph
A business ontology of classes and relations

Extendable per plan. Search finds records by meaning rather than by keyword, and every answer can be traced back through the connections it was built from.

System agents
A bench of system agents

Sales, HR, support, docs, research and more — each seeded per tenant with its own tools and budget. Add custom agents on higher plans.

Tool registry
Risk-gated: READ, CREATE, MODIFY

READ tools run on their own. CREATE tools add new records and notify you. MODIFY tools — anything that changes or sends existing data — wait for your explicit approval before they run.

Budgets
Per-agent monthly cap

80% warning, 100% auto-pause. Visible to admins. Resets on the 1st of each month.

Heartbeats
Tasks on a schedule

Agents run proactively on a schedule you set. Example: “Sales · weekdays 9am · surface stale deals.”

Audit
Every call logged

Every tool run and every outbound model call is recorded — what was asked, what came back, how long it took, and whether sensitive fields were masked.

More in the engine

One engine, a whole stack of AI

Chat is just the front door. The same engine runs code, works with images, audio, documents and data, listens and speaks, and works in the background — lightweight, security-sensitive steps run on LinearPilot's own models, and heavier AI goes to named providers with sensitive fields masked.

Media stack

Vision, audio, documents, and data

The engine does far more than chat. Lightweight and security-sensitive work runs on LinearPilot's own models; heavier generation and extraction goes to an external provider — with sensitive fields masked, your consent, and the provider named in writing.

Runs in-platform
Document classificationSearch result rankingSensitive-name maskingFile conversionForecasting
External provider · sensitive fields masked
Image generationDocument text extractionTranscription

Sandboxed code execution

For real analysis, the engine writes and runs code in a network-isolated sandbox — and only after you approve the run. A Pro add-on; included on Max.

Voice in and out

Talk to the engine and hear it answer. Spoken replies are generated on your device and never leave it; dictation now uses external transcription, with sensitive fields masked and only when your workspace has enabled it.

AI Inbox + nightly learning

A background digest lands every six hours with its sources cited. Overnight, a learning loop reviews the day's work and shows you what improved in What's New.

AI that respects your data.

Get the power of AI without the privacy trade-off. Start free today.