Built to be trusted
An assistant that can read your CRM has to be careful with it. The Layer Assistant is built so that it can only ever do what you could do yourself, and so that your company stays in control of where its information is processed.
It acts as you
Every time the assistant looks something up, it does so with your account's permissions. It cannot see a record you could not open yourself, and it cannot widen its own access. If your role does not let you see a customer, team or report, neither can the assistant while you are using it.
It is read-only
In its current release the assistant only reads your data. It finds records and explains them; it does not create, edit or delete anything. That means there is no risk of it changing your data while it answers a question.
Where your information is processed
To answer a question, the assistant sends the relevant details to an AI model. Your company chooses which AI service that is and uses its own account, so you are never pooled into a shared one. There are three levels of control:
A cloud provider with your own key - OpenAI or Claude, using the key your administrator sets.
Your own Microsoft tenancy with Azure AI Foundry - prompts and answers stay within the Azure environment you govern, under your own region, data-retention and compliance settings.
Your own hardware - a self-hosted, OpenAI-compatible model, so nothing leaves your network at all.
See Choose your AI provider for how to set each of these up.
The assistant only draws on your Layer data and our Help Centre. It is not a general web search and does not go looking outside Layer for answers.
What it does not do
It does not act on your behalf or make changes to records.
It does not see data beyond your own permissions.
It does not send your data to any AI account other than the one your company configures.