Start with one outcome. Grow into a platform.
Each process below runs on agents and workflows your team governs. Alongside what it does, you can see the rules it would run under: a guardrail policy, which actions ask a person, and a budget.
- IT & identity
Joiners, movers and leavers
When an employee’s record changes to leaving in your HR system, an agent disables their sign-in, removes app access, asks the manager where their files go, and proves nothing was missed.
- Where it runs
- HR system update, your directory through a private tunnel, manager approval in Slack
- Guardrail policy
- “Never remove access for anyone other than the person named in the HR record.”
- Approvals
- Deleting accounts or transferring files asks the manager first.
- Budget
- $2 per run, $50 per day
- Finance opsStarter package
Invoice to payment
An agent reads incoming invoices, matches them to purchase orders, sends exceptions to the right approver and posts the rest to your ledger.
- Where it runs
- Accounts-payable mailbox, ERP, approver in Microsoft Teams
- Guardrail policy
- “Never change bank details on a supplier from what an invoice says.”
- Approvals
- Posting a payment always asks a person; reading and matching run on their own.
- Budget
- $0.50 per invoice, $100 per day
- Starter package
- Reads invoices from Gmail into a Google Sheet.
- Customer serviceStarter package
Resolve, not just reply
An agent answers tickets in your support desk, looks up orders, issues refunds within policy and hands the rest to a person with the context attached.
- Where it runs
- Zendesk, Intercom or Freshdesk; WhatsApp and email
- Guardrail policy
- “Never promise a refund, credit or delivery date the policy doesn’t allow. Checked on every reply.”
- Approvals
- Refunds ask a support lead.
- Budget
- $0.20 per conversation, $200 per day
- Starter package
- Classifies an incoming message and files it as a Zendesk ticket.
- Sales & CRMStarter package
Pipeline that keeps itself current
New leads are enriched, scored and routed to the right rep; calls and meetings are logged; deals that have gone quiet are flagged.
- Where it runs
- Web form, CRM, a daily digest in Slack
- Guardrail policy
- “Never email a prospect who has opted out, and never quote pricing.”
- Approvals
- Sending an email to a prospect asks the rep; updates to the CRM run on their own.
- Budget
- $0.10 per lead, $30 per day
- Starter package
- Turns a Typeform submission into a HubSpot contact and tells the team in Slack.
- Logistics
Exceptions before they escalate
An agent watches shipments, chases carriers when something slips, and tells the customer before they have to ask.
- Where it runs
- Carrier updates, SMS and email to customers, the ops channel
- Guardrail policy
- “Never give a customer a delivery date the carrier hasn’t confirmed.”
- Approvals
- Rebooking a shipment asks the ops lead.
- Budget
- $0.05 per shipment checked, $40 per day
- HRStarter package
Onboarding that finishes
From the day a new hire is added in your HR system to the end of their first week: paperwork, equipment, accounts and first-week tasks, assigned, chased and tracked until every step is done.
- Where it runs
- New employee in your HR system, IT and facilities tasks, the new hire’s manager in Slack
- Guardrail policy
- “Never share one employee’s personal details with another.”
- Approvals
- Ordering equipment asks the manager; creating accounts runs on its own.
- Budget
- $1 per new hire, $20 per day
- Starter package
- Creates Asana tasks for a new BambooHR hire and messages their manager in Slack.
Policies, approvals and budgets are examples of what a team might set. A starter package installs a first step of the process, which you then build on.
See how it works.
- AgentsAlways-on agents in chat, email, support desks, SMS and phone calls, each with its own identity.
- WorkflowsDurable workflows that wait days for a person, retry what fails and reach 800+ apps.
- LLM gatewayOne endpoint for 30+ model providers on your own keys, with routing, fallback and cost control.
- Memory & contextMemory, business definitions, skills, documents and your own data model, shared by every agent.
- APIs & MCPPublish any workflow or agent as a versioned API or an MCP server, with keys, quotas and limits.
- GovernanceGuardrails, approvals, budgets, permissions and audit around every model call and action.