How Boatwork Assist works

A Boatwork Assist agent is a configured AI assistant that works inside your business. Each agent has a name, a role, and a defined scope of work — for example handling a shared inbox, responding to leads, drafting quotes, or keeping a calendar current.

A teammate, not just a tool

Rather than a chatbot bolted onto a website, an agent is set up to do real work: read and answer email, manage calendar events, create and edit documents, and follow the playbook and persona you define. Its usefulness comes from context — the more it understands about how your business actually runs, the more accurately it can act on your behalf.

A managed platform

Boatwork Assist is managed: we run, host, patch, and maintain the runtime each agent operates on. There is no software for you to install or operate. What stays entirely under your control is access and data — you connect the accounts you want the agent to use, on permissions you approve, and you can withdraw access whenever you choose.

Each agent runs in its own isolated environment

Every agent runs in its own dedicated runtime — an isolated environment with a workspace-scoped filesystem, not shared multi-tenant compute. Depending on the channels you need, an agent is hosted either on a dedicated Mac or in the cloud (see Channels).

How an agent connects to your tools

Agents act through standard OAuth connections to the services you already use — Google Workspace or Microsoft 365 for email, calendar, and documents. You approve the permissions during a one-time setup, and each mailbox is connected by signing in as that mailbox. Access is delegated and least-privilege; the Security & data protection page covers exactly how that boundary is enforced.

How an agent generates its responses

To produce a reply, the agent’s request is routed to a third-party AI model provider, which may include Anthropic, OpenAI, and others depending on configuration. Message content is sent to the provider solely to generate the response. We select providers that do not use customer data for model training by default.

You stay in control

Every agent operates under a named human supervisor and an organization-level security policy that caps what the agent is allowed to do. See Agent safeguards & oversight for the controls that bound an agent’s behavior.