AI Agents · Early access

AI agents that know your business.

Give an agent your knowledge, your instructions, and the tools it may use. Put it in front of customers, share it with your team, or keep it for your own work.

Get Help.ai updates
Connected conversation bubbles, a knowledge document, and a completed task, illustrating an AI agent turning a question into a useful next step.
  • Answers from the knowledge you choose
  • Tools and connected accounts you approve
  • Your team one click away

Most questions are the start of a job.

A customer asks about a refund. The answer is in a policy, the order details are in another system, and the exception needs a person. An agent connects the conversation to all three, so the job moves instead of stalling.

The right answer for each audience

Customers get product help. Colleagues get internal process. You get your own reference work. Each agent uses only the knowledge and instructions you give it.

Tools that match the job

Pick from enabled tools and connect the accounts the work needs. The agent gets exactly the capabilities its role requires, and nothing more.

People in control

Approve tool steps when the agent asks. In customer conversations, an operator can take over with the full history and hand control back when done.

Set it up once. Improve it every week.

  1. Define the job

    Decide which questions it answers, which knowledge it reads, and which actions it may take.

  2. Choose where it works

    Use it in your workspace for yourself and your team. For customers, deploy it as a website widget, a hosted page, or through the API, each with its own knowledge and tools.

  3. Review and refine

    Read the questions, answers, and actions. Fill the gaps in its instructions, and keep a person on anything that needs judgment.

One builder. Many jobs.

A private helper and a customer-facing agent share the same building blocks. Give each one its own responsibility, audience, and knowledge.

Repeatable work

Save the instructions, review criteria, and references you use every week. Recurring drafts, briefs, and reviews start from the same solid base.

Team knowledge

Colleagues ask in plain language and get the answer from your shared documents, with the next step in the process.

Customer support

Customers get answers from your knowledge in the conversation, and a person picks it up when the request needs one.

Service enquiries

Understand what the customer wants and collect the details your team needs to respond well.

Customer onboarding

Explain each step, gather the required information, and keep new customers moving.

Connected actions

Take the next step in the tools you already use, through supported integrations and the accounts you authorize.

You decide what it knows and what it can do.

Each agent saves its instructions and capabilities for one role. Use it privately in your workspace or through a deployment.

Knowledge for the right audience
Organize documents, pages, and reference material in your Library, then choose what each agent may use. Check internal material before an agent faces customers.
Tools and integrations
Select enabled tools, authorized connected accounts, and required credentials. Add MCP connections when the job calls for them.
Operator conversations
Your team replies inside the customer’s existing conversation, with everything already said in view.
Conversation analytics
See recurring questions, customer needs, form activity, and where answers fall short.
Images, video, and voice
Workspace agent conversations can use enabled image, video, and text-to-speech options. What’s available depends on the models and tools enabled in your workspace.

Questions about AI Agents

What can I build an agent for?

Any job with a clear scope: answering questions about your documents, guiding colleagues through a process, or gathering client intake details. Give it clear instructions, trusted knowledge, and limits on what it may do.

When should I use an agent instead of AI Chat?

Use AI Chat to work through an open-ended question. Use an agent when the same role needs saved instructions, chosen knowledge, and configured tools across many conversations.

Can agents help both my team and my customers?

Yes. Use workspace agents for your own work and internal questions, and deployments for customers. Different audiences usually need different knowledge and instructions, so give each its own agent.

Which actions and integrations can an agent use?

The ones enabled in your workspace and the accounts you authorize. Configure the tools for the job and review any approval request. A listed integration doesn’t guarantee every action in that service.

Can a person take over?

Yes. Your team can review the conversation, take it over, reply, and hand control back to the agent when they’re done.

How do agents work with AI Forms?

The agent handles the conversation and the form captures the structured details. Together they turn a customer’s questions into information your team can act on.

Put your knowledge to work.

Get updates on agents for your customers, your team, and your own work.

Product updates by email. Privacy Policy

Explore related products and resources