Intelligence
Rank, compare and find places, from a question or a typed query.
Ask in plain English or send a typed query. Rank neighbourhoods on any combination of signals, compare up to five, find the places most like one you know, catch the outliers or project a trend. The AI only writes the query. The numbers come from the data.
See it run
A question becomes a query. The query becomes an answer you can run again.
A plain-English question is turned into a typed plan, the plan runs against the store, and the ranked places come back with the plan attached.
The AI writes the query. The data gives the answer.
A question becomes one of seven typed operations, the store answers it, and the plan comes back with the result so it can be checked and run again.
Your words become a typed query
A question is turned into one of seven typed operations with explicit parameters. You get the plan back with the answer.
The AI never sets a number
It reads the question and writes the plan. Every figure comes from the store, so nothing is invented.
Every answer carries its plan
Run the plan again, with no AI in the loop, and get the same answer on the same data and version. Wire it into your product as code.
Scoped to what we can answer
The planner only offers signals the store ranks: crime, deprivation, sale prices, schools and transport. A question about anything else is refused, not guessed.
seven operations · typed · replayable
What the planner covers.
Seven operations
Rank places by one signal or several, with value or percentile filters, in a nation or a district. Read one place. Compare two to five. Score a place. Find the places most like one. Screen for outliers against their peers. Project one signal's trend from its monthly history.
Two ways to ask
A plain-English question, or the typed plan sent straight from your app. The same executor runs either, and the response says which way it was asked.
Hard questions about places, answered from the data, with the query attached so the answer can be checked and run again.
Get a sandbox key
The Intelligence workspace in the console
Where it goes in your product.
Search by criteria, suggest similar places, screen a dataset for outliers, or answer inside your own assistant.
Typed answers your product can show, save and run again.
Search by criteria
Let a customer describe what they want and return the places that fit, ranked, with the listings in them. Save the plan as the filter.
Similar places
Suggest the neighbourhoods most like the ones already in a customer's portfolio, on the signals they care about.
Outliers in a dataset
Screen a nation or a district for places that stand out from their peers on one signal, before they surprise your users.
Inside your assistant
Call the same operations from the AI features in your product through the MCP server, with the answer typed and the plan attached.
Frequently asked
Intelligence, answered.
What can I ask it?
Seven things: rank places on one or more signals, read one place, compare two to five, score a place, find the places most like one, screen for outliers against their peers, and project one signal's trend. The signals it can rank are crime, deprivation, sale prices, schools and transport.
What can it not answer?
Questions about amenities or flood risk, which the store does not rank yet. They are refused with a clear reason rather than answered with a substitute. Those values are on every area response.
Does the AI make up the numbers?
No. The AI only works out what your question means and writes the plan. Every number comes from the store.
How do I know the answer is right?
Every answer comes back with the plan that produced it and says whether the plan came from a question or from your code, so you can read it, check it and run it again yourself.
Will the same question give the same answer?
The same plan on the same data and methodology version gives the same answer, with no AI in the loop. When the data refreshes or the method changes, the answer can move, and the response carries the version.
Do I have to write in plain English?
No. Send the typed plan straight from your app. Either way you get the same answer back.
How do I use it?
Ask in the playground, call it with one API request, or call the same operations from your AI features through the MCP server. Or book a call and we run your own questions.
Ask hard questions of place data. Get typed answers.
Rank, compare, match, screen and project, from a question or a query, with the plan attached. Start in the free sandbox.