Data in → AI prompts → validated structure out

Make AI produce the exact shape your product needs.

SourceShape is the commercial product layer for the OpenAI Interpreter worker: a governed transformation workbench that turns crawled evidence, source records, and natural-language requests into validated JSON, query intent, UI behavior, delivery payloads, and versioned knowledge.

See the architecture Explore use cases

ProfilesCrawler enrichment, YP search intent, custom account transformations.

GovernedGuidance versions, eval cases, budgets, replay, audit, and promotion.

ValidatedNo free-form magic. Outputs must pass schema and delivery contracts.

Many data sources flowing through SourceShape into structured outputs and dashboards

{
  "display": "comparison_grid",
  "filters": ["location", "service", "confidence"],
  "sort": "best_match",
  "edit": "guided_review",
  "provenance": true
}

More than prompts. A controlled interpretation layer.

SourceShape packages prompt engineering, context scoping, output validation, delivery shaping, and account governance into one repeatable transformation lifecycle.

⌘

Context profiles

Each use case has its own input schemas, guidance, output contracts, delivery targets, retention rules, and budgets. Crawler data and live YP queries do not fight over the same prompt.

▣

Workbench control

Build, test, compare, replay, promote, rollback, and audit guidance bundles. Domain experts can improve transformations without rewriting application code.

✓

Validated delivery

Raw model output is kept separate from delivered output. Every response can be checked against the exact consumer contract before it reaches a website, API, dashboard, or account tenant.

Multi-context AI orchestration with context profiles, prompt packs, schema rules, validation, and delivery contracts — **Multi-context orchestration.**Sources, contexts, guidance, validation, and delivery contracts stay visible and governed.

The GPN product chain, with SourceShape in the middle.

It connects discovery, scraping, interpretation, and the public YP experience while AccountRevenue and operational tools keep it commercially manageable.

Positioning mapGoogle Places → Domain Scraper → SourceShape → YP website

Google Places to Domain Scraper to AI Interpreter to YP Website platform architecture supported by AccountRevenue, SourceAjax, and Google Cost Monitor

Operational position.Google Places → Domain Scraper → AI Interpreter → YP Website, supported by billing, AJAX delivery, and cost monitoring.

Two first use cases. One reusable engine.

The same lifecycle supports offline enrichment and live answer shaping. That is the core commercial advantage.

Crawler evidence to structured knowledgeTurn Google Places, website captures, sitemaps, reviews, photos, and scraper artifacts into a canonical business graph with claims, relations, topics, taxonomy, and provenance.

Natural language to adaptive interface planInterpret user queries into allowed display, edit, filter, sort, explanation, and ranking instructions so the public site can configure itself safely around the data it presents.

Account-owned transformation profilesSell reusable AI transformation workbenches to external accounts while keeping guidance, evals, input sources, usage, and outputs isolated per tenant.

A self-configuring result page is the interesting part.

The user does not just get an answer. The consuming website receives a safe contract that says what data to show, how to explain it, what filters make sense, which sort order is justified, and where edits or review are allowed.

Natural language query turning into a self-configuring interface with tables, cards, charts, maps, filters, sorting, and edit controls

{
  "intent": "compare_local_service_providers",
  "answerPanel": "ranked_explanation",
  "displayMethod": "cards_with_map_and_confidence",
  "filterMethod": ["location", "service", "opening_hours"],
  "sortMethod": "semantic_match_then_distance",
  "editMethod": "operator_guided_claim_review",
  "unsupportedActions": ["sql", "unsafe_component_id"],
  "provenanceRequired": true
}

How SourceShape gets transformations exactly right.

Prompt quality becomes an operational asset, not a hidden string in application code.

Register input

Connect source references, samples, replay seeds, redaction state, and retention policy to a context profile.

Improve guidance

Version prompts, examples, model policy, schemas, confidence rules, and promotion criteria.

Run and compare

Execute model cascades with budget checks, retries, timeouts, deterministic fallback, validation, and cost tracking.

Deliver safely

Shape validated output for the target consumer while keeping raw output, delivered output, and provenance separately versioned.

SourceShape prompt management workbench with prompt versions, tests, JSON output validation, replay history, and deployment preview — **Prompt development becomes a product workflow.**Compare versions, tune quality, replay failures, validate schemas, and promote safely.

Pricing that matches the value: governed transformations.

SourceShape should be priced like infrastructure for repeatable AI output quality, with AccountRevenue metering actual usage.

Builder

€149 / month

For internal teams or customers testing one transformation profile.

1 context profile
Prompt and schema workbench
Replay and comparison basics
Usage metered separately

Start builder

Professional

€495 / month

For production pilots, vertical directories, and repeatable enrichment workflows.

3 context profiles
Guidance lifecycle and eval sets
Run ledger, budget checks, alerts
Delivery inspection and compatibility gaps

Request demo

Business

€1,950 / month

For account-owned profiles, multi-tenant usage, governed review, and commercial rollout.

Multiple accounts and organisations
Account-owned guidance overrides
Advanced retention and audit
Custom usage events via AccountRevenue

Talk to us

Build the AI layer that tells your product how to present its own data.

Brought to you by Mondial IT and CIO at Work. Designed for the GPN platform chain, but reusable for any account that needs controlled data-in → structure-out AI transformation.

Mondial-IT.nl CIOatWork.nl AccountRevenue SourceAjax Google Cost Monitor

Plan strategy session