AI Contexts
AI Contexts are reusable pieces of information that you can attach to any AI feature in Zero. Think of them as a shared knowledge base for your AI agents — once you write down something the AI should always know (your company description, your ICP, your tone of voice), you can reuse it everywhere instead of pasting the same paragraphs into every prompt.
Manage your contexts at Settings → Workspace → Agent Context.
What are AI Contexts?
An AI Context is a named document containing background information that gets injected into the prompt before an AI feature runs. Each context has:
A name — how you'll recognize and reference it (e.g. "Company info", "ICP", "Email tone").
A content body — a rich-text document where you write the actual information. You can use headings, bold text, lists, and paragraphs.
Contexts are stored at the workspace level. Anyone in the workspace can create, edit, and use them, and they're available across every AI surface in Zero.
Contexts are never used automatically. Creating a context doesn't change how any existing AI feature behaves — you have to explicitly attach a context to an AI Property, an automation AI step, or any other AI surface for it to take effect there. This is intentional: it lets you keep contexts focused and only pay the token cost where the information is actually relevant.
When an AI feature runs, the contexts you've attached to it are loaded, converted into plain text, and prepended to the prompt sent to the model. The model is told that this information was "provided by the user to give background information about the query" and uses it to inform its response.
Why would you need them?
Most AI features in Zero — AI properties, automation AI steps, email generation — work by sending a prompt to an LLM. Without context, the LLM only knows what's in that single prompt. That means:
It doesn't know what your company does, so it can't tell whether a lead is a good fit.
It doesn't know your ICP, so it can't score deals against your real qualification criteria.
It doesn't know your tone of voice, so any email it drafts sounds generic.
It doesn't know your product positioning, your competitors, or who your investors are.
You could solve this by pasting the same paragraphs into every AI prompt you write. AI Contexts let you write that information once and reuse it everywhere. When your ICP changes or you launch a new product, you update the context in one place and every AI feature using it picks up the change automatically.
A few common reasons to use them:
Consistency — every AI output reflects the same understanding of your business.
Reusability — write your style guide once, use it in every email automation.
Maintainability — when something changes, you update one document instead of dozens of prompts.
Quality — AI outputs get noticeably better when the model has real context about your business instead of having to guess.
Where AI Contexts are used
AI Contexts can be attached to any AI surface in Zero:
AI Properties
When configuring an AI Property on a company, contact, or deal, you can attach one or more AI Contexts to it. The contexts are sent to the model alongside the property's prompt every time it runs, so the AI has consistent background information when extracting or generating data.
For example, an AI property that scores deals can use a "Scoring criteria" context, while an AI property that drafts outreach can use an "Email writing style" context.
Automation AI Steps
The "Run AI task" action in Automations supports AI Contexts the same way AI Properties do. Attach a context to the action and the AI step will use it when the automation runs.
This is especially useful for automations that:
Score or qualify inbound leads (attach your ICP and scoring criteria).
Draft outreach emails (attach your writing style and company info).
Enrich records with structured data (attach your company info so the AI knows what to look up).
Anywhere else AI runs
Any feature in Zero that lets you write a prompt can also use AI Contexts. The selector uses the same component everywhere, so once you understand it in one place you'll recognize it in the others.
How to write a good AI Context
A good context is short, factual, and specific. Long context documents waste tokens and dilute the signal. A few guidelines:
Be specific. "We sell to mid-market B2B SaaS companies between 50 and 500 employees" is more useful than "We sell to businesses".
Stay factual. Lists, bullet points, and short paragraphs work better than prose narratives.
Don't repeat yourself. If you have a "Company info" context and an "ICP" context, the ICP context shouldn't restate everything that's already in Company info.
Use one context per topic. It's better to have several focused contexts than one giant document. You can attach only the ones relevant to each AI feature.
Keep it current. If your positioning, pricing, or ICP changes, update the context. Outdated context is worse than no context.
Write for an LLM, not for a human. You don't need preambles, transitions, or polish. Just structured information.
A typical context might be 5–20 short paragraphs or bullet points. If it gets much longer than that, consider splitting it into multiple smaller contexts.
How to configure AI Contexts
Managing your contexts
All AI Contexts for a workspace are managed from one place: Settings → Workspace → Agent Context.
From this page you can:
Create a new context — click "Add Context" to open the editor. Give it a name, write the content, and save.
Edit an existing context — click any context card to open the editor with the existing content loaded.
Delete a context — hover over a context card and click the trash icon. If the context is currently in use by any AI properties or automations, you'll see a warning listing exactly what will be affected. Confirming the delete will remove the context and clean it up from any places that referenced it.
The editor is a rich-text field that supports headings, bold text, paragraphs, and bullet lists. Save is disabled until the context has both a name and some content.
Starting from a template
When you create a new context, the editor shows a few starter templates you can use as a starting point. They give you a working example you can edit to fit your business:
Company information — a short factual description of what your company does, who it serves, and any other relevant background. Use this as your default attached context for almost every AI feature.
Investor information — a description of your funding history, investors, and partnerships. Useful when AI features need to be aware of your business relationships.
Email writing style — tone-of-voice guidelines for AI-generated emails. Includes examples of how you want your emails to sound (casual vs. formal, length, formatting rules, words to avoid).
Clicking a template button fills the editor with that template's content. From there you can edit it freely — the templates are starting points, not fixed. You can also create contexts entirely from scratch without using a template at all.
Attaching contexts to an AI feature
Once a context exists in your workspace, you can attach it to any AI feature:
Open the configuration for an AI Property or AI automation step.
In the prompt editor, look for the AI Contexts selector at the top.
Pick one or more contexts from the list. You can attach as many as you want, but more contexts means a longer prompt — only attach the ones that are actually relevant to that feature.
Save the AI Property or automation. The next time it runs, the selected contexts will be included in the prompt sent to the model.
You can also create a new context directly from the AI feature's configuration without leaving the page — the same editor opens inline, and the new context is immediately available to attach.
Last updated