Your AI, in your company's voice.
The graph holds the facts. Your AI Persona holds the judgment: how your AI should speak, who it serves, what rules it follows, what it values right now. We call each piece a lens.
Define it once. It's injected automatically into the right conversations and agent runs.
Ideal customer profile
Mid-market B2B SaaS, 200 to 2,000 employees, with a dedicated platform or operations team. Multiple disconnected systems of record. Active AI initiative with executive sponsorship.
Version history
v4 narrowed the profile to mid-market after 12 closed-won accounts. Runs before the change still audit against v3.
Applies when
Deal or account entities are in scope
Mode
Auto · fires on trigger
Your CRM knows the deal stage. Your repo knows the commit. No system records your tone of voice, your ideal customer, your pricing guardrails, or this quarter's priorities. That judgment lives in people's heads and in scattered docs, and your AI never sees it.
A lens captures one piece of that judgment as a first-class, versioned object. It sits beside the graph and is injected at runtime, so the AI reasons through your business's point of view instead of a generic one.
Stop pasting the same guidance into every prompt. Encode it once, version it, and let it apply itself.
Five kinds of lens
Everything your AI should know that no database stores.
Organize lenses the way your business thinks about them.
Voice
Tone, formality, language, and format. How your AI sounds, scoped to a person or a role so the right voice shows up automatically.
Profiles
Ideal customer profiles, buyer personas, and account notes. Who you serve and what matters about them.
Policy
Pricing caps, compliance disclaimers, escalation rules. The guardrails the AI must respect, applied wherever they’re relevant.
Playbook
Positioning, internal frameworks, and glossaries. The way your company explains itself and gets work done.
Priorities
Quarterly goals, active campaigns, live incidents. The time-varying context that should color answers right now.
Derived over time
Lenses can be suggested from real outcomes, like an ICP inferred from closed-won accounts. The platform proposes, a human approves.
How it applies
The right lens, at the right moment, automatically.
Lenses are selected by deterministic triggers and injected at runtime. You don't pick them each time. The relevant ones just show up.
Scoped to a person or role
A tone lens bound to a user or role fires whenever that person is in the loop. Sales sounds like sales. Support sounds like support.
Triggered by intent
A lens tied to entity types or intents applies only when the work calls for it. The pricing policy shows up on a pricing question, not everywhere.
Auto or manual
Auto lenses fire on their triggers. Manual lenses are attached to a specific chat or agent when you want explicit control. Priority resolves overlaps.
Versioned and auditable
Your identity changes. Your history stays honest.
When your ICP shifts or your tone evolves, you publish a new version. Past runs remain auditable against the exact guidance that was live when they happened.
- Immutable versions
Every edit creates a new version. The lens identity stays stable while its content evolves.
- Audited to the version
Injections record which version was used, so you can always trace why the AI said what it said.
- Snapshots and rewind
Capture a known-good AI Persona and roll every lens back to it at once.
- Lifecycle states
Draft, active, archived. Stage changes before they go live, retire what no longer applies.
Design Partner Program · 4 spots left
Give your AI your judgment.
We help you encode your first lenses around your real voice, ICP, and policy. Direct founder access. Locked pricing.