> For the complete documentation index, see [llms.txt](https://doc.aissist.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.aissist.io/success-metrics.md).

# Success Metrics

**Last updated:** May 5, 2026

Measure AI performance with two metric groups:

* **Issue rate** — how often AI makes mistakes
* **Accomplish rate** — how often AI contains or resolves the work

These metrics help you balance reliability and automation.

### Issue rate

Issue rate measures how often Aissist produces the wrong outcome.

That can include:

* incorrect replies
* incorrect tags
* poor escalation decisions
* unsafe or off-brand behavior

Classify issues by business impact.

* **P0** — severe and unrecoverable business impact
* **P1** — clear business impact, but recoverable
* **P2** — minor impact with low urgency

{% hint style="info" %}
A core goal is **trust**.

AI should avoid catastrophic failures, such as abusive language, unsafe actions, or behavior outside its intended role.
{% endhint %}

### Accomplish rate

Accomplish rate measures how much work AI handles successfully.

Track it with two metrics:

* **Contained Rate**
* **Resolution Rate**

#### Contained rate

Contained rate is:

`1 - percentage of sessions with sys_human_help`

Sessions usually get `sys_human_help` when:

* the user asks for a human
* AI lacks the information to answer
* AI finds a scenario it does not know how to handle

#### Resolution rate

Resolution rate is:

`1 - percentage of sessions with sys_human_help or sys_human_follow_up`

The added tag, `sys_human_follow_up`, usually means:

* a human should pay attention
* a human needs to complete an action
* there are unresolved questions or next steps

Targets depend on workflow complexity and the quality of your instructions, assets, and actions.

Below are practical ranges and the usual improvement path.

<table><thead><tr><th width="166">Rate</th><th width="317">Ideal Target</th><th>Actions to improve</th></tr></thead><tbody><tr><td>Contained Rate</td><td><ul><li>Sales lead qualification -> 80% - 90%</li><li>Service -> 70% - 90% (depends on complexity, eCommerce will be higher than technology service)</li></ul></td><td>This normally means that there is a gap of either instruction or assets against the incoming traffic, enhancing which will lead to higher contained rate.</td></tr><tr><td>Resolution Rate</td><td><ul><li>Sales lead qualification -> 70 - 80%</li><li>Service -> 60% - 85% (depends on complexity, eCommerce will be higher than technology service)</li></ul></td><td>High contained rate and low resolution rate could be normal because in some scenarios that you do want human team to pay attention but not necessarily take actions. Fine-tuning AI to let AI output less response like "someone from the team will reach out", etc, will improve the resolution rate.</td></tr></tbody></table>

{% hint style="info" %}
There is no universal target.

With strong documentation and clear workflows, many teams can reach `70%–80%` resolution rate and `80%–90%` contained rate.
{% endhint %}

### How to improve results

If issue rate is high:

* tighten instructions
* remove conflicting assets
* narrow the workflow scope
* keep sensitive cases out of **Auto-Pilot** until the workflow is stable

If contained rate is low:

* add missing knowledge
* refine sub agents
* improve action coverage

If resolution rate is low:

* reduce unnecessary human follow-up triggers
* improve action execution
* refine replies that hand work to humans too early

### Recommended review cadence

Review success metrics every week during rollout.

Track:

* P0, P1, and P2 issues
* contained rate
* resolution rate
* top reasons for human handoff

Then update instructions, assets, actions, or routing based on what you learn.