A survey project designed in minutes, with AI if you like

A survey goes well or badly long before anybody walks out of the van. If the fields are wrong, everybody collects the wrong thing; if the values are free text, nobody spells them the same way twice. The Project designer is where the admin and managers get that right once, with the structure checked as they go, and then hand the same project to everybody.

A design is a draft. It saves as you work, every admin and manager sees the same draft, and no crew sees anything until it is published.

• New project: an empty draft, designed by hand.
• Start from an existing project: a copy of how one of your cloud projects is built, its layers, fields and pick lists. Nothing anybody collected is copied, and the original is not changed.
• A survey template: a ready-made design, used exactly as it is, then changed as the job needs.

Whichever you start from, the draft opens in the same editor: layers of points, lines or polygons, and fields of six types (text, whole number, decimal, yes or no, date, date and time), each one required or not, with a default if you want one. Every layer already carries a name and the dates the app keeps itself, so there is nothing to add for them.

The Survey Template Catalogue holds ready-made surveys drawn from public standards and field practice: street lighting, trees, utility poles, ecology and habitats, marine, engineering, cadastral, energy and more. Search them, or filter by sector and region. Each template shows its layers and fields, the standards it follows, how far it has been checked against them, and where it came from.

A template is a starting point, not a certificate. Check it against your own copy of the standard before relying on it.

On a draft, Design with AI lets you describe the survey in your own words. The designer asks a few questions, with answers to tap, then proposes layers and fields. Nothing changes until you press Use this structure; after that it offers pick lists for the fields that have choices, and you add them with one press. Ask it for changes at any point, such as another field, and it asks what kind of data that field holds.

It sees what you type and the draft’s structure, never anything your crews collect, and it runs in the EU. Whether an organisation uses it is its admin’s decision: one switch, AI tools, turns it off for everybody. Its calls use the organisation’s processing credits, and a yearly allowance comes with the subscription.

Every proposal is checked by the same rules as a design made by hand, so the AI cannot propose what the editor would refuse.

A pick list gives a field a fixed set of values on the phone, so a condition is “poor”, never “Poor”, “bad” or “pr”. In a draft each list goes inside the project when you publish, already linked to its fields, so the crew links nothing by hand. Keep a list for other projects by saving it as one of the organisation’s attribute sets.

Once the organisation’s admin allows it, the admin and managers can connect Claude or ChatGPT to Mapit GIS Cloud. In the chat they can read how the organisation’s projects are built, search the template catalogue, and design a new project or make one from a template. Before it creates anything it shows the proposed structure and waits for you to confirm.

A connection never publishes: the draft appears in the Project designer, with a link back from the chat, for you to review. It runs on your own Claude or ChatGPT plan, not on the organisation’s credits.

Publish project checks the draft and creates the cloud project for the organisation, private until you share it with the crew. From then on the crew collects into it with Mapit Pro or Pro+ in the Mapit GIS app, and a later change, a new layer, field or pick list, by hand or drafted by the AI, ships to every phone as a change to the structure rather than a new download.

Project sync is what happens after that.

Designing is for the organisation’s admin and managers, while its subscription is active. It happens in a browser and needs no app licence; collecting is what needs Mapit Pro or Pro+. Contributors collect into what is published, and viewers see what is shared with them.

• The AI never sees collected data: no records, locations, photos, WFS or web maps.
• Nothing publishes itself. A draft reaches a crew only when an admin or a manager publishes it.
• It does not set how a layer looks on the map; styles are set in the app.
• It does not check its suggestions against paid standards it cannot read. Check field names and values against your own copy before relying on them.
• From Claude or ChatGPT, what the connection reads, the structure of your projects, goes to Anthropic or OpenAI under their own terms and may be held outside the EU. Everything Mapit GIS Cloud stores stays in the EU.
• A published project is changed through its structure editor and a release, never by redesigning it from scratch.