The AI builders are useful on their own, but they get considerably better when they know how your organization works. This article covers the ways you can shape what they produce, from the structure they start with to the organization-level controls that govern them.

Templates are the most direct way to make the AI builders follow your standards today. A template is one of your own projects that has been marked as a template, so it can be used as the starting point for new work.

The reason this works is simple: the assistant reads whatever is already in the project you have open. Create the project from your template first, then open the AI Builder. It sees your step ordering, your standard demographic questions, your contact capture approach, and your page layout, and it edits within that structure instead of inventing its own.

A practical way to use this:

Templates can be scoped by department, so a large organization can give each department its own set without cluttering everyone else's list. Your account also has access to a set of general example templates alongside your own.

Both builders accept a document as context: a PDF, Word file, text file, or CSV. The assistant reads it and works from your actual language and commitments rather than generic phrasing.

This is a per-use attachment rather than a permanent organizational setting, but it is the most direct way to carry your own material into a build. An adopted plan, a scoping memo, or an engagement plan all work well. Attach the relevant section rather than an entire long document, since the assistant reads a substantial but finite amount of any single file.

The email Assistant additionally reads the project a campaign is linked to, which is usually a faster way to get real content into an email than attaching anything.

Separately from anything you configure, the AI Project Builder can search a curated library of questions drawn from real engagement projects, and check what answer options are commonly used. You do not need to set this up. It is worth knowing about because it explains why its suggestions do not read like generic survey output.

All of the AI assistants across PublicInput are governed by a single organization-level setting. Go to Settings, then the General tab, and find the Agentic AI section. It contains one checkbox, Allow agentic AI tools, marked Beta.

Turning it off disables the AI assistants for everyone in your organization. Turning it back on restores them. This is your control, not something that has to be requested.

The section also states the data handling commitments your IT and procurement teams will want to see. The models are securely hosted, Microsoft Azure based, strictly hosted in the United States, and never trained on your data. There is a "Learn how we keep your data safe" link beside it, and a fuller explanation in the data safety article below, which is the right thing to send to a security reviewer.

The ability to upload your brand guidelines, public involvement plan, style guide, and standard operating procedures once and have every AI builder follow them automatically is in development. When it ships, the idea is that you maintain a library of guidelines, switch individual ones on and off as your team's needs change, and have them applied to everything anyone builds.

Until then, starting from a template and attaching your own documents are the ways to keep output aligned with how your organization works.