Vertical-aware AI - sounds like someone in your business
AI output written for your kind of business. A roofer's follow-up doesn't read like a recruiter's, and neither does the recap of their walkthrough. RadiusOS reads which template your workspace is built on and matches the voice, vocabulary, and scenarios that vertical actually uses - across drafts, walkthrough recaps, and Snap-to-Quote line items.
How it works
Every workspace is built on a template - Roofing, HVAC, Real Estate, Therapist, Recruiter, and so on. When the AI generates anything for you (a draft, a walkthrough recap, a quote from a photo), it layers a vertical-specific context block on top of the base prompt. That second layer carries the trade language, the typical scenarios, and the trust signals that vertical actually leans on.
A roofer's draft mentions tear-off, ridge cap, and the storm date. A real estate draft mentions comps, contingencies, and showings. A roofer's Walk & Talk recap mentions deck condition, flashing, and the inspection findings. A plumbing recap mentions the fixture, the shut-off, and the access path. A therapist's draft stays strictly administrative - intake logistics, scheduling, fee questions - and routes anything clinical back to the EHR.
There's no setting to toggle. The vertical context is picked automatically from your workspace's template. If you want a different voice, change the template (Settings → Pipeline) or build a custom one.
Which templates are covered
All 20 marketplace templates ship with vertical context:
| Group | Templates |
|---|---|
| Trades | Roofing, HVAC, Plumbing, Electrician, Painting, General Contractor, Landscaping & Lawn Care |
| Service / Field | Solar Installation, Insurance Agent |
| Real Estate | Real Estate |
| Sales / B2B | B2B Sales, Recruiting |
| Creator / Freelance | Creator Outreach, Freelance Clients |
| Personal | Job Search, Photography, Fitness Trainer |
| Founder | Startup Fundraising |
| Compliance (under-claim) | Therapist (Private Practice), Financial Advisor |
Custom workspaces (built from scratch in the wizard, no marketplace template) fall back to a generic professional voice. The base prompt's USER vs CONTACT separation still applies - drafts will never confuse the user with the recipient.
Compliance verticals - what's allowed and what isn't
RadiusOS is not HIPAA, FINRA, SEC, or BAA-compliant. For Therapist (Private Practice) and Financial Advisor templates, the AI is locked to administrative outreach only.
| Allowed | Not allowed |
|---|---|
| Initial inquiry replies (fees, openings, sliding scale) | Clinical content, treatment notes, symptom or diagnosis references |
| Intake / scheduling / paperwork links | Specific investment advice, recommendations, performance discussion |
| Appointment confirmations and reminders | Crisis-response language (route to 988 / 911 / EHR) |
| Polite reactivation after a gap | PHI in subject lines or body |
| Generic seasonal touches (year-end, tax-season heads-up) | Promises, guarantees, or specific account/balance references |
If a request asks for clinical or advice content, the AI will draft a short administrative reply that politely routes the client to the EHR / patient portal / phone for substantive matters. By design - we'd rather lose a click than help you publish something a regulator can flag.
Where vertical voice shows up
Anywhere RadiusOS generates AI output for you:
- The on-demand Draft Follow-up button on the contact detail page
- The Morning Ritual approval flow
- The overnight Morning Digest cron that pre-writes follow-ups for stale deals
- The Ask RadiusOS in-product chat when it drafts on your behalf
- Walk & Talk recap extraction (title, summary, suggested tasks, reminders, tags)
- Snap-to-Quote line-item extraction from walkthrough photos
Every one of those surfaces shares the same vertical-context injection pattern, so a roofer's walkthrough recap reads like roofing language and the line items on the quote it drafts use the right vocabulary too. Consistent voice across drafts, recaps, and quotes.
Cost
The vertical context block is ~120-200 tokens layered onto the base prompt as a second cached system block. Cached input rates apply, so the marginal cost is roughly $0.00002 per call - well below the noise floor of a typical workspace's monthly AI usage. Same pattern across drafts, Walk & Talk extraction, and Snap-to-Quote.
Drafts still consume one AI credit each on Free and Pro tiers, same as before. Walk & Talk + Snap-to-Quote runs are gated by the per-plan walkthrough cap (Free 1/mo preview, Pro 30/mo, Business+ unlimited fair-use), not the credit pool.