Transparency
How we use AI
PlanningLeads is built on Ireland's official construction record. AI is the tool we use to make that record readable at scale - nothing more mysterious than that. Here is exactly what it does, what it sees, and the rules we hold ourselves to, in plain English.
What the AI does
Reading public planning text
Our main use. A large-language model reads the public description of each planning application or consent ("construction of 42 dwellings...") and turns it into structured fields: project type, scale, sector, storeys, renewable-energy signals, unit counts and floor areas where stated. That is what makes the record filterable. The same kind of read picks out the planning conditions on a grant that require a consultant's report (an ecologist, an acoustic survey, an arborist), which is how professional-services leads are found.
Search and drafting help
A model parses natural-language searches ("granted solar farms in Cork") into filters. On Max, a model can also write a short read of a single project (what it is and why it fits your trade) when you ask for one, and draft an optional outreach note for you to review, edit and send yourself. AI-drafted text is labelled as such in the app.
The rules we hold ourselves to
What the models see, and never see
The models are sent public, organisation-level record text: project descriptions, use classes, statutory metadata. When you ask for a project read or an outreach draft, the model also gets the trade and business description you gave us, so the draft fits you. They are never sent your login details, your saved searches, homeowner quote requests, or personal contact details of private individuals (which we do not hold in the first place).
No decisions about people
Our AI classifies construction projects, not humans. It makes no automated decision with legal or similarly significant effect on any individual (no GDPR Article 22 profiling). Lead scores rank projects; a human customer decides what to do with them.
Processor terms, EU transfers, no training
The models are Anthropic's Claude models, provided under a data-processing agreement with EU Standard Contractual Clauses; API inputs are not used to train their models. Our servers are EU-hosted (Hetzner).
The outputs can be wrong - and are labelled
AI classifications are indicative, not authoritative: a use class can be misread, a unit count missed. Modelled figures (values, grant estimates, paybacks) are always labelled as modelled, with the method stated. The underlying public record - which we always link - is the source of truth. Spot something wrong? Tell us: hello@planningleads.ie.
This page is kept in step with the product: if our AI use changes, this page changes with it. The formal detail lives in our privacy notice; our data sources and licensing are on the data page; security practices are at trust & security.