AI & Automation
Tool and agent orchestration
Composable agents that take real actions across Thorbis, every step gated behind your approval and a clear audit trail.
Overview
An assistant that only talks can't move work forward. Thorbis is building agent orchestration that lets AI chain tools together to draft a job, prep an invoice, or queue a follow-up across the platform. Each action runs against your permissions and waits for human approval before anything changes, and every step is logged. You stay in control while the busywork gets staged for you.
What we're building
An assistant that only talks can't move work forward, but an assistant that acts on its own is a liability. Contractors need AI that can actually take the busywork off their plate without being trusted to change things unsupervised.
Thorbis is building agent orchestration that lets AI chain tools together to draft a job, prep an invoice, or queue a follow-up across the platform. Every action runs against your company's permissions, waits for human approval before anything changes, and is logged step by step, so nothing happens behind your back.
You stay in control while the multi-step busywork gets staged and lined up for you, review exactly what an agent intends to do, then approve, change, or stop it before any of it runs.
What you'll be able to do
- Chain multiple tools to complete a multi-step task
- Stage actions for human approval before they run
- Respect company permissions on every action
- Track each step with a full audit trail
- Take action across jobs, invoices, and follow-ups
- Review exactly what an agent will do before it runs
- Stop or change a staged action before you approve it
Rollout plan
- 01
Approval-gated actions
In progressLet agents stage single actions that run only against your permissions after you approve them.
- 02
Multi-step orchestration
In progressChain multiple tools into a full task like drafting a job or prepping an invoice.
- 03
Cross-platform reach with audit
In progressExtend staged actions across jobs, invoices, and follow-ups with a complete step-by-step audit trail.
Milestones
- Not started:
Tool action layer
Build permission-checked actions agents can call across the platform.
- Not started:
Approval gate
Stage every change for human approval before it runs.
- Not started:
Multi-step chaining
Let agents sequence tools to complete a multi-step task.
- Not started:
Audit trail
Log each step so every agent action is traceable.
More in AI & Automation
How to send useful feedback
Name the blocked role
Product feedback is more useful when it names the owner, dispatcher, technician, office user, or customer affected.
Explain the workaround
Describe what your team does today and where the process fails under real work.
Define success
Share what would change after launch: time saved, fewer manual steps, cleaner records, or better customer follow-up.