# Implementing Kustomer AI

> This guide outlines a phased approach to implementing AIpowered customer support in Kustomer It explains the crawl walk run roadmap and the data and organiz

Source: https://help.kustomer.com/en_us/implementing-kustomer-ai-rkzh7vzRg

Last updated: 2026-07-14T22:12:26.904Z

This guide outlines a phased approach to implementing AI-powered customer support in Kustomer. It explains the crawl, walk, run roadmap and the data and organizational requirements needed to scale AI safely.

Use this guide to plan, launch, and mature AI across customer automation and agent assistance.

## Understanding the AI implementation phases

AI adoption progresses through three phases. Each phase builds on the previous one and increases capability, impact, and risk.

### Crawl: enabling conversational knowledge access

#### Goal

Provide fast, accurate answers to common questions and reduce manual lookup for agents.

#### Capabilities

*   [Add public URLs as knowledge sources](https://help.kustomer.com/en_us/add-public-urls-r1IqUxWKJg) for your AI Automations.
*   Enable conversational [responses based on approved guidance](https://help.kustomer.com/en_us/managing-ai-automations-SkaCf6nFbg#writing-guidance).
*   [Generate summaries](https://help.kustomer.com/en_us/set-up-summaries-BkESHdVmee) and [draft replies for agents](https://help.kustomer.com/en_us/using-typeahead-HkmnbKe9be).

#### Success criteria

*   Customers receive instant answers to common questions.
*   Agents spend less time searching for information.
*   AI responses stay within approved knowledge sources.

#### Key metrics

*   Knowledge answer accuracy.
*   Reduction in average handle time.
*   Agent usage of AI-assisted responses.

### Walk: assisting and personalizing agent workflows

#### Goal

Improve resolution speed and consistency by embedding AI into agent workflows.

#### Capabilities

*   [Configure AI tone and voice](https://help.kustomer.com/en_us/managing-ai-automations-SkaCf6nFbg#guidance) to match your brand.
*   Use [customer and conversation data for personalization](https://help.kustomer.com/en_us/use-evaluations-to-measure-ai-performance-BJy49yooxe).
*   Provide [suggested responses](https://help.kustomer.com/en_us/using-typeahead-HkmnbKe9be) in the agent interface.
*   Generate [handoff summaries and closing notes](https://help.kustomer.com/hand-off-options-BkAfYVRqkl).

#### Success criteria

*   Agents rely on AI suggestions during conversations.
*   Fewer errors caused by missing or outdated context.
*   Consistent tone across interactions.

#### Key metrics

*   Agent adoption rate.
*   First-response time improvement.
*   Customer satisfaction (CSAT or U-SAT).

### Run: automating actions with integrations

#### Goal

Enable AI to complete service actions, not just respond to questions.

#### Capabilities

*   [Integrate with third-party systems](https://help.kustomer.com/en_us/categories/kustomer-integrations-SJmPRmjzU).
*   [Trigger workflows using APIs](https://help.kustomer.com/en_us/categories/workflows-BJjk_bBeI).
*   [Enable AI agents to complete end-to-end tasks](https://help.kustomer.com/en_us/understand-ai-agent-tools-ByB09BdAA).

#### Example use cases

*   Update order or payment status.
*   Trigger refunds or account changes.
*   Resolve common requests without agent involvement.

#### Success criteria

*   High containment for selected use cases.
*   Reliable execution of automated actions.
*   Clear escalation paths for edge cases.

#### Key metrics

*   Containment rate.
*   Escalation accuracy.
*   Error and rollback frequency.

## Preparing to scale AI

These practices support all phases and should begin early.

### Defining the plan

#### Defining the business problem

**What to do**

Define the specific customer support problem AI will solve.

**Why it matters**

A clear problem prevents over-automation and keeps the rollout focused.

**Examples**

*   Reduce backlog.
*   Automate order status requests.
*   Improve policy accuracy.
*   Reduce handle time.

#### Defining required roles

**What to do**

Assign ownership for decisions, data, and adoption.

**Why it matters**

Clear ownership improves alignment and speeds decision-making.

**Roles to include**

*   Product or operations for business context.
*   Data or IT for systems and integrations.
*   Support representatives as change champions.

### Preparing data and content

#### Mapping customer data sources

**What to do**

Identify where customer, order, and payment data reside. Document how systems connect.

**Why it matters**

AI accuracy depends on consistent and reliable data.

**Examples**

*   Standardize status fields.
*   Map order and payment systems.
*   Verify compliance requirements.

#### Structuring the knowledge base

**What to do**

*   Create single-topic articles.
*   Write step-by-step instructions where needed.

**Why it matters**

Structured content improves AI accuracy and reduces errors.

**Best practices**

*   Split multi-topic articles.
*   Rewrite vague or ambiguous policies.
*   Use consistent terminology and formatting.

### Preparing the organization

#### Building trust in AI

**What to do**

Communicate how AI supports agents and involve them early.

**Why it matters**

Early involvement improves adoption and surfaces issues sooner.

**Examples**

*   Demonstrate AI features.
*   Share rollout plans.
*   Collect feedback across teams.

#### Aligning on success criteria

**What to do**

Define success metrics for each phase.

**Why it matters**

Teams need clear thresholds before scaling.

**Examples**

*   Accuracy thresholds.
*   Handle time reduction.
*   Containment goals.
*   Agent adoption targets.

## Testing and rollout

### Running testing

**What to do**

Test AI using real conversations and edge cases.

**Why it matters**

AI behavior is probabilistic and improves through iteration. [Use evaluations](https://help.kustomer.com/en_us/use-evaluations-to-measure-ai-performance-BJy49yooxe) to repeatedly test and optimize behavior for the intended outcome. 

**Test scenarios**

*   Measure responses to specific customer concerns.
*   Surface behavioral issues before a wide rollout.
*   Optimize tool access and agent behavior, then see your results fast.

### Rolling out gradually

**What to do**

Follow a phased rollout aligned to crawl, walk, run.

**Example progression**

*   Sandbox testing.
*   Agent assistance only.
*   Limited automation.
*   Full automation for approved use cases.

## Summary

A successful AI rollout depends on:

*   Clear problem definition.
*   High-quality data and knowledge.
*   Agent trust and adoption.
*   Defined success metrics.
*   Gradual, test-driven scaling.

Following this roadmap helps teams deliver AI-powered support that is accurate, scalable, and trusted.
