Solutions architect
The data model, the integrations, and the build order are decided before a line of code, so the scope you see is the scope you get.
Plenty of tools let you mock up an AI demo. Far fewer give you a real product with accounts, a database, guardrails, and a deploy. Lab Twelve builds AI MVPs as bounded slices: the model call sits inside a real workflow, outputs are validated against a schema, and a human review gate runs before anything merges. The pricing engine maps complexity to published tiers, usually Business App through MVP Sprint, so the model never invents your quote. You hold the API keys and own the runtime cost. This is an AI-native engineer shipping your product, with security and cost controls handled, not a science project billed by the hour.
One fixed price buys a whole product team in one: architect, designer, and AI-native engineer.
The data model, the integrations, and the build order are decided before a line of code, so the scope you see is the scope you get.
Interface, flow, and brand are designed for this build, not dropped onto a template. The finished app looks like it was meant to ship.
Senior execution at AI speed, in production with the source code in your hands. One builder with the range of a whole product team.
| Scoped AI workflow | Included |
|---|---|
| Structured outputs | Included |
| Accounts + data model | Included |
| Human review gate | Included |
| Production deploy + cost sheet | Included |
Scoped AI workflow: Inside a real product, not a bare chat box
Structured outputs: Validated against a schema
Accounts + data model: Postgres-backed, auth as scoped
Human review gate: Before merge to main
Production deploy + cost sheet: You hold the API keys
| Training custom models | Not in base scope |
|---|---|
| Unmetered token spend | Not in base scope |
| Open-ended research | Not in base scope |
Training custom models: Inference and orchestration only
Unmetered token spend: Runtime is your provider bill
Open-ended research: Scope needs a definition of done
A representative ScopeSpec ticket. Yours is assembled live from the scope chat, priced by the engine, and locked before you pay.
AI job definition
Input, output schema, and failure modes named.
Product around the model
Accounts, data, and the user path.
Guardrails + logging
Rate limits, PII rules, fallbacks, evals.
Deploy + cost sheet
Production URL, runtime estimate for you.
Number of workflows, structured outputs, and integrations map to Business App or MVP Sprint plus the AI feature add-on, all from offers.ts. The model never sets the number.
Provider is scoped per project, often Anthropic for extraction and chat. You own the keys and the runtime cost after handoff.
Yes. New workflows are change orders, or queue them in a dev lane once v1 is shipping.
No agency to manage, no roster to assemble. The whole team, one number, source code in your hands.