QuoteVoice

Turn a plumber’s site voice note and photos into a checked, editable quote before the job details go stale.

Comment

Business Viability

Solo plumbers often leave a survey with the information needed to quote, but not with a usable record. Photos sit in a camera roll, measurements live in notes, and important qualifications remain in memory. The quote is reconstructed after working hours, delayed, or never sent. QuoteVoice starts at that exact moment: the plumber records a spoken walkthrough, adds site photos, reviews a structured draft, controls every price and line item, and sends a branded customer-ready quote.

We started QuoteVoice on 26 July 2026, late in the competition window. That compressed history explains both the focus and the evidence position: we prioritised a working production workflow, safety boundaries, and auditable Gemini operations, and we report £0 revenue and 0 paying customers rather than implying a longer commercial track record.

This is deliberately narrower than a field-service suite. The first buyer is a solo UK plumber or two-person firm that quotes domestic work from a phone and has no office support. The commercial model is a free first quote followed by a £49 monthly subscription. Recurring value comes from weekly quoting, while later acceptance, deposit, and follow-up increments can improve retention without changing the core product. Acquisition tests will use short phone-recorded demonstrations, trade communities, merchant or bookkeeper referrals, and narrowly attributed search campaigns. The activation metric is a reviewed first quote; retention is measured by reviewed and sent quotes per account, not by page views.

The production system already has self-serve accounts, persistent business profiles and quotes, a live Stripe Checkout path, and a clear trial boundary. It currently has no verified independent product revenue or paying customers in the evidence pack, so this submission does not manufacture traction. The next commercial proof is a small cohort of working plumbers completing real quote journeys, followed by transparent paid conversion and cancellation evidence. Target economics are software-like margins with bounded media volume and measured Gemini cost per completed quote.

The official cash-basis P&L records $0 revenue and $0 product-specific cash expenses. Development used founder time and shared Yensi hosting, subscriptions, and product-factory resources that existed before this entry; they are disclosed here and are not presented as QuoteVoice-specific paid expenses.

AI-Native Operations

Gemini is load-bearing, but constrained. QuoteVoice sends the plumber’s voice/photo evidence through a signed adapter running on Google Cloud Run. Gemini 2.5 Flash extracts bounded work scope, structures draft line items, and flags missing coverage. The production response retains request ID, provider, model, latency, and token evidence. If Gemini is not configured or returns an unsafe contract, drafting fails closed instead of silently fabricating a result.

The model never sets final prices, grants entitlement, sends a quote, or changes payment state. Deterministic product logic applies the plumber’s configured rate, owns totals and state transitions, protects human overrides from later AI refreshes, and requires explicit human review plus business branding before approval and send. This is an AI-native operating model rather than a chat wrapper: Gemini converts unstructured field evidence into the structured scope that drives the product workflow, while the application owns money, authority, audit, and recovery.

The architecture also preserves trust. Raw media has bounded retention; product copy states that customer and property data are not used to train general-purpose models. Accounts and quote state persist across logout and deployment. Automated tests cover consent, trial gating, media validation, invalid Gemini output, human approval, branding, override preservation, and follow-up rules. A deadline-sensitive clock defect found in the audit was corrected with an injected deterministic clock; the suite now reports 39 passed and one optional local-auth check skipped.

Category Impact

For a one-person trade business, evening admin competes directly with paid work and family time. QuoteVoice aims to turn the most perishable input—the explanation in the plumber’s head at the property—into a controlled business record while it is still fresh. The impact is not “AI writes faster” in the abstract. It is fewer lost details, earlier customer-ready quotes, consistent branding, a visible follow-up state, and a path from site evidence to acceptance and money.

The theory of change is measurable: structured capture reduces reconstruction work; scope flags reduce avoidable omissions; a faster reviewed quote increases the chance that a customer receives something actionable; persistent sent/follow-up state reduces forgotten opportunities. The product will measure time to reviewed quote, unresolved-scope rate, human override rate, sent-quote rate, customer response, and repeat quoting by cohort. It will not claim revenue uplift or accuracy until those outcomes are observed with real businesses.

Shared Yensi product-factory contracts and runtime rails existed before this entry. The QuoteVoice workflow, UI, domain rules, persistence, Gemini Cloud Run adapter, and billing integration were built during the competition window. That separation is disclosed in the repository and submission.

Built With

Share this project:

Updates

Submission history