An AI agent that works
within your team's rules.
Your team sets how far Aide goes on each topic, writes the procedures it follows and approves the actions you choose. Every change is on the record, and Aide tests what it learns before keeping it.This is the Agent Governance Engine: not a separate product, but the rules that govern everything Aide already does. Topics and AI levels, procedures, reply verification, approvals and the Action Trace.What does an AI agent governance platform do?
An AI agent governance platform lets a team decide what its AI agents may say and do, test them before they go live, and review every action afterwards. In Aide, your team sets an AI level for each topic, writes the procedures Aide follows, approves the actions you choose, and has an audit log of every change.UpdatedWhat governs the agent, today
Governance in Aide is not a policy document. It is a set of settings your team sees and changes, and checks that run on every conversation. Here is each one, as it looks in the product.
Aide can propose actions, but a person confirms them before anything happens.
- None
- Drafts
- AI Agent
When to use
The customer says an item arrived broken or damaged. Not when the parcel is only late.
Scoring the latest conversations…
Matches of the latest 50 conversations
Matched
- Kettle arrived dented96%Nora Feld · 30 Oct
- Grinder lid cracked93%Ravi Menon · 24 Oct
Declined
- Still waiting on my mugs31%Lucia Romero · 29 Oct
The parcel is only late. Nothing arrived broken.
Use for this
Burr grinder1 × $189.00$189.00Awaiting approval · Aide
- Get the workspace's inbox settings
- Update the workspace's inbox settings
| Field | Before | After |
|---|---|---|
| AI refund cap | 50 | 75 |
Replies judged acceptable, matched on 17 topics and weighted to Aide's mix of them, against 5 teammates
- Aide79–91%
- Team median72–89%
- Top quartile83–95%
Bars are 95% intervals: the range the true share likely sits in.
Aide can propose actions, but a person confirms them before anything happens.
- None
- Drafts
- AI Agent
No trade-off between capable and safe
An agent that changes orders, accounts and records has to be held to rules your team reads and sets. Those rules apply on every conversation, on every channel.Works with your existing stack
The same rules apply wherever Aide works: Aide Chat, email, or the helpdesk your team already uses. No migration, no engineering.
Watch only first, then live per brand, with an Activity log of what Aide did.
Answers in the inboxes you take live, with the Aide panel beside each ticket.
Answers chat and email once an inbox is live, with Aide's card in the sidebar.
Aide reads only the contact and account fields you choose.
Reads orders and tracking; cancels, changes addresses and refunds with approval.
Reads orders, products and customers, so answers carry the live order.
Search knowledge, read conversations, and act from a Claude chat.
Frequently asked questions
AI agent governance is the set of rules that decide what an AI agent is allowed to do, and the record that shows it stayed inside them. In Aide, your team sets an AI level on every topic, writes the procedures the agent follows, approves the actions you choose, and reads every change in the audit log.
Autonomous AI decides for itself what to say and do on every conversation. Governed AI works inside limits your team sets: in Aide, each topic has an AI level (AI Agent, Drafts or None), actions run only from procedures your team set live, and the actions you choose wait for a teammate's approval.
Aide acts on its own only on topics your team has moved to AI Agent, and only within the procedure it follows. A reply that promises something the topic's level rules out, leaves a placeholder or claims to be a person is revised or handed off before it reaches the customer. Refunds wait for your team by default.
New topics start at Drafts, where a person confirms any action. A procedure stays a draft until an owner or admin sets it live: check its trigger on your last 50 conversations and replay a past conversation with the draft. In Zendesk, Front and Gorgias, an inbox runs in Watch only, where Aide decides and drafts without writing anything, then goes live for test senders, a share of conversations, and everyone.
Yes. Each conversation shows the topic on each message, the sources under each reply, the actions offered, run or sent for approval, and the handoff notes. The audit log, for owners and admins, records every change to agents, procedures, topics, macros, settings and knowledge, and every refund approval, with who, when, from where and a before and after. It exports to CSV and is kept 2 years.
Aide's Quality page shows a pass or fail scorecard against a written quality bar, and replays eval cases against the agent every night. Aide's replies and your teammates' replies are graded on the same scorecard, with Aide beside your team's median and top quarter. Every change Aide learns is listed with an Undo.
Yes. Connect your own data sources and internal systems through an API or an MCP server, and Aide takes action in them only from a procedure your team wrote. Each action is confirmed by the customer, approved by a teammate where the topic asks, and logged like an order action.
When Aide is not sure, it hands the conversation to your team with a summary and sends the handoff message your team wrote. A customer who asks for a person gets one at once, without the AI answering first.
Yes. When a customer asks, Aide answers with a fixed, truthful line naming itself as an AI agent for your business, which no setting or instruction changes. Any reply claiming to be a person is rejected before it is sent.
Aide Superintelligence runs safe steps directly. Anything customers would notice, such as putting an agent live, changing its instructions, replying to a customer, acting on an order or changing a setting like the refund cap, shows an Approve or Decline card with the exact change first. Every change it makes is in the audit log, marked as made with Superintelligence on behalf of the teammate who asked.
Aide is SOC 2 Type II certified, GDPR compliant, and HIPAA compliant where applicable, with zero data retention on Claude model calls and open-weight models on Aide's own servers. Add AI levels per topic, approvals and the audit log, and most of a security review's questions already have answers. See aide.app/security for details.
Most teams connect their helpdesk or a channel and see first value in 24 hours. Start with your highest-volume topic at Drafts, preview it on real conversations, then move it to AI Agent when your team is ready.
Start free with 150 AI replies a month and no card. Paid plans start at $39 a month with a 7-day free trial, and you pay for AI replies, not seats. See aide.app/pricing for every plan.
Give an AI agent real work, under your rules
Start every topic at Drafts, preview it on real conversations, and move it to AI Agent when your team is ready. Every change stays on the record.