Chival is the quoting system of record for commercial service. A tech captures the failed unit on their phone; Chival drafts a priced Good/Better/Best proposal against your price book; a manager approves it before the tech leaves the roof.
The demo runs on sample data in your browser. No login, nothing saved, no real quote.
A live walkthrough of one real visit: what the tech fills in on the roof, and what the manager reviews and approves. Drag the panels around — this is the actual interface, not a screenshot of it.
Interactive walkthrough · sample data · illustrative pricing until you import your price book.
Techs never see cost or margin here — enforced server-side, not hidden with CSS.
Waiting on a quote from the field…
10-ton Trane rooftop unit, 18 years old, R-22 — age and refrigerant guardrail
Extraction has to work one-handed, in bad light, on a unit that hasn't been cleaned since install, with a signal that comes and goes on a commercial roof — and it has to feed straight into a price book that protects your margin, not just spit out text. That's the harder problem, and it's the one Chival is built to solve: real roles, org-scoped data, offline-safe capture, and a software agent that has to earn the right to skip your review. That is what Chival sells.
No new hardware and no separate takeoff session back at the office. The quote is built out of what the tech already does on every call.
A photo of the data plate plus what they found, dictated or typed. Chival pulls make, model, serial, capacity, refrigerant and voltage, and flags every field it is unsure about so the tech can correct it in seconds. The tech never sees cost or margin — that is enforced on the server, not by hiding a column.
Findings become a structured scope — urgency, labor hours, parts — priced into three options against your own catalog and labor rates. Until you import real prices, seed pricing is badged as illustrative in the app and on every proposal — and replacing it takes minutes: drop in a CSV, or paste a raw supplier sheet and let AI clean it up.
The manager reviews the draft, fixes what is off, approves, and sends. Nothing leaves Chival without a human action. Every correction is recorded against the field it changed.
Chival measures its own error rate against your corrections, per equipment type and job kind, and graduates one bucket at a time. That is not a hedge — it's how you know exactly when to stop double-checking it, bucket by bucket, instead of guessing. You set the thresholds. Chival never promotes itself.
Every quote is reviewed. Every new bucket starts here.
Gate: none — this is the default and it never expires on its own.Buckets with a low measured edit rate get flagged low-risk: one-click approve.
Gate: measured edit rate under your threshold across your minimum sample.You set the ceiling. Anything above it still routes to a human.
Gate: the ceiling is yours to set, and only you can raise it.Opt-in per bucket, always logged, always reversible.
Gate: someone at your shop turns it on, per bucket. Revoke any time.A bucket only advances when the measured edit rate stays under your threshold across a minimum sample. Readiness is computed from your own review history in your own workspace — it is not a benchmark, an industry figure, or a number we made up.
A tech creating a quote from a roof needs the scope, the equipment and the customer-facing price. They do not need your unit cost, your catalog cost or your gross margin — and in a commercial shop, that information walking around on twenty phones is a real problem, not a theoretical one.
In Chival this is enforced where it counts: a tech's API response does not contain cost or margin fields at all. It is not a hidden column, a permission flag on a widget, or a CSS rule. There is nothing to reveal with a dev-tools panel because the data never leaves the server.
Commercial contractors run maintenance-agreement portfolios: the same techs, in front of the same equipment, on a schedule. Chival keeps every unit at every site across every visit, so a deficiency found on a PM in March is still attached to that RTU when it fails in July — with the history, the prior quote and what the customer declined.
Context for why this matters: 63% of commercial contractors report that more than half their customer base is secured by preventative maintenance agreements — ServiceTitan 2025 Commercial Service Market Report, n = 1,014, fielded May 2025. https://www.servicetitan.com/blog/commercial-service-market-report-takeaways
Residential quoting tools assume one operator. A commercial service contractor is a company, and Chival is built like one.
| Role | Can | Cannot |
|---|---|---|
| Owner | Everything, including billing | — |
| Admin | Members, roles, price book, settings | Billing, delete org |
| Manager | Review, edit, approve, send, see cost & metrics | Manage members |
| Estimator | Create and edit quotes, see cost | Approve, send |
| Tech | Create quotes from the field, see their own | See cost or margin, approve, send, see other techs' quotes |
| Viewer | Read-only | Any write |
Every row of data carries an organization id and every query is scoped to the caller's organization. There is no endpoint that can return another organization's data.
Measured time from tech submit to sent, approval latency, edits per quote, win rate by tier, and autonomy level per bucket — produced by your own usage. At onboarding we ask for your current turnaround and monthly volume and store it, clearly labeled as your self-reported baseline, so any before/after is against your own stated starting point.
No customer logos we haven't earned, no testimonials, no case studies, no "typical results". If a metric has no data yet, Chival says so — "insufficient data," not a placeholder number dressed up to look real. That's the same rule your price book runs on, and the same one your own reported metrics will run on once you're live: if we won't fabricate a number for ourselves, we won't fabricate one for you.
We publish where BuildOps, ServiceTitan, QuoteIQ and BuildFolio are stronger than us, which competitor facts we have actually verified, and which cells say "not verified" because we will not guess.
Create your workspace, load a price book, run a real visit through it.