AI Automation
In Forest Workflows, AI makes decisions, drafts text and calls your tools. You choose which steps run on their own and which wait for your team.

TL;DR
In Forest Workflows, AI decides, drafts, suggests values and calls your tools, one step at a time.
Each suggestion surfaces the right data and a recommended next step, which your operator confirms, edits, or overrides, or which runs automatically once you trust it.
Every AI step runs under the operator's permissions, follows your approvals, and lands in the same audit trail as a manual action.
Every AI step starts as help for an operator. Once your team trusts it, switch it to automatic. The steps you can write as rules stay deterministic.
AI decisions
A decision step can be made by the operator, assisted by AI, or left to AI, which picks the branch from the context gathered in earlier steps.
Drafts and suggested values
Guidance tasks draft the reply or the note, and Update Data steps suggest field values. The operator confirms, edits, or overrides.
Tool calls through MCP connectors
An MCP task lets AI act in Stripe, HubSpot, Slack or any MCP server, limited to the tools you allow on that step.
Manual or automatic, step by step
Each step completes when the operator confirms, or automatically. Trigger Action steps can execute on their own once you trust the AI's pick.
Your model or ours
AI steps run on Forest's AI server by default, with nothing to configure, or on your own LLM provider through Forest Runtime.
Same audit as a manual action
Whether the operator accepts a suggestion or AI acts on its own, the audit trail captures both the suggestion and the decision.


Built for regulated ops
Rules first, AI second
Deterministic steps handle every check you can write as a condition. AI handles reading, drafting and the cases rules cannot settle.
Role-aware AI
AI steps surface only the data and actions the operator can see. AI agents inherit the same scope.
Same workflows, same audit
AI steps follow your Forest rules, run through your approvals, and land in the same audit trail as a manual action.
Vendor agnostic
Frequently asked questions
You still have question ?
Book a conversation with a forest expert
What can AI do in a Forest workflow?
Make a decision and pick a branch, draft guidance text, suggest field values, pick the action to trigger, and call tools through MCP connectors. Each is a step type in the workflow editor.
Which actions can AI suggest or run?
Any Forest Action the operator has permission to trigger. Suggestions are scoped to the operator's role and to the workflow the case belongs to.
Can AI run a step without anyone confirming?
Yes, when you set it to. Each step's completion can be manual or automatic, and Trigger Action steps can execute automatically. Actions with a form still need an operator to fill it in.
How do we stop AI from calling a sensitive tool?
Restrict the Tools field on the MCP task. AI can only call the tools listed there, so leave out anything that sends an email or deletes a record.
What model does Forest use?
Forest's AI server by default, with nothing to configure. The model is configurable: point Forest Runtime to your own provider, such as Claude or GPT, and the call and response are logged.
What data goes to the model?
AI steps and MCP tasks send prompt content, which can include record data, to the LLM provider and to the tools they call. Everything else stays in your Forest back-end, in your infrastructure.
Is the suggestion audited?
Yes. The suggestion, the operator's response (accept, edit, override, ignore), and the final action are all recorded against the case.
How do you ensure customer data never leaves our environment?
The Forest back-end runs inside your infrastructure, alongside your databases. Customer data travels to a provider only at the point where its service is called, and the response is logged back to the case. Data stays in your systems of record.

