Salesforce Knowledge Base Setup for Customer Service (2026)

Salesforce Knowledge is the built-in knowledge base and article management system within Service Cloud – allowing support teams to create, publish, and search a library of troubleshooting guides, FAQs, product documentation, and how-to articles that agents use to resolve cases faster and customers use to self-serve. A well-configured Knowledge Base reduces case volume (deflection), improves first-contact resolution rates, and provides consistent answers across all support channels. This guide covers enabling Salesforce Knowledge, creating article types and record types, knowledge article workflow (draft to review to publish), surfacing articles in the Service Console, and configuring the customer-facing Knowledge portal.

The best guide is the one that makes self-service feel easier to maintain.

A useful explanation should help the reader see where the knowledge base fits in support operations.

That means the guide should focus on practical service flow instead of generic content management advice.

For many teams, the value is in giving people a reliable place to find answers.

It should also show how organised content helps both self-service and agent support.

A good guide should explain what the knowledge base is meant to solve and why the structure matters.

That makes it a practical support feature rather than just a library of articles.

Salesforce knowledge base setup is useful because support teams need a place to store answers that both agents and customers can use. A well-built knowledge base can reduce repetitive work and make service faster over time.

Enabling Salesforce Knowledge

Salesforce Knowledge requires a Knowledge licence (included with Service Cloud and available as an add-on for Sales Cloud users who need self-service capabilities). Enable Knowledge at Setup, then Knowledge, then Knowledge Settings, then Enable Salesforce Knowledge.

Once enabled, the Knowledge tab appears in Salesforce. The first configuration steps:

• Enable Allow Users to Create and Edit Articles to allow non-administrator users to write articles
• Configure Knowledge language settings: if your support team handles multiple languages, enable multi-language Knowledge and designate translation workflows for article localisation
• Enable Case Deflection: allow agents to search Knowledge from within Case records and attach relevant articles to case resolutions

Knowledge Article Types and Record Types

Knowledge articles are categorised by record type – different article types have different field sets, page layouts, and publishing workflows. Common article record types for B2B support:

• FAQ: question-and-answer format with a direct answer. Simple structure, high volume, written by support agents from resolved cases.
• How-To: step-by-step procedural guides with numbered steps covering a specific task from start to finish.
• Troubleshooting Guide: diagnostic flow for resolving a specific error or problem – symptoms, root cause analysis, resolution steps, escalation path if unresolved.
• Product Documentation: reference documentation for product features – comprehensive coverage of a specific feature, appropriate for detailed technical reference.
• Release Notes: documenting product updates, new features, and known issues – time-sensitive articles linked to specific product versions.

Create Knowledge record types at Setup, then Object Manager, then Knowledge, then Record Types. Each record type has its own Page Layout – the FAQ layout might show only Question and Answer fields, while the Troubleshooting Guide layout includes Symptoms, Root Cause, Steps to Resolve, and Related Articles fields.

Article Lifecycle: Draft, Review, and Publish

Knowledge articles follow a structured publication workflow – they are not immediately visible to agents or customers when created. The article states are:

• Draft: article created but not yet published – visible only to the author and reviewers. Agents can create Draft articles directly from Case records (capturing the resolution while working the case).
• In Review: submitted for approval before publication – a knowledge manager reviews for accuracy, completeness, and style consistency before publishing.
• Published: active and visible to agents in the Knowledge sidebar and to customers in the self-service portal (if enabled). Published articles have a URL and are indexed for Knowledge search.
• Archived: removed from active search results but preserved for reference – appropriate for outdated articles that may still be referenced from external links or older case resolutions.

Knowledge Approval Process

Configure an Approval Process for Knowledge articles to enforce the review workflow:

1. Setup, then Process Automation, then Approval Processes, then New Approval Process, then select Knowledge Article Version
2. Configure entry criteria (all articles, or specific record types)
3. Assign approvers: typically a Knowledge Manager or senior support agent
4. Configure actions on approval (set article to Published) and rejection (return to Draft with comments)

When an agent submits an article for review, it routes to the Knowledge Manager for approval. The manager reviews, may edit, and approves – triggering the article to publish automatically.

Data Categories: Organising Your Knowledge Base

Data Categories are the taxonomy for organising Knowledge articles – allowing agents and customers to browse articles by product, topic, region, or audience. Enable Data Category Groups at Setup, then Data Category Groups.

Articles are tagged with categories from each group – a troubleshooting guide for Product A’s API integration would be tagged: Product A / API and Integrations / Administrators. This enables precise filtering: an administrator searching for API integration help sees only relevant articles.

Data Categories also control article visibility – articles tagged with specific categories can be visible only to users with access to those categories. This enables internal-only articles (visible to agents, not to the customer portal) and audience-specific content (developer docs visible only to developer community members).

Surfacing Knowledge in the Service Console

The Knowledge Base delivers its case resolution value through the Knowledge Sidebar in the Lightning Service Console. When an agent has a Case open, the Knowledge Sidebar automatically suggests articles based on the Case’s Subject and Description – Einstein AI reads the case content and ranks relevant articles by semantic similarity.

Agent Knowledge workflows:

• Article-assisted resolution: agent finds the relevant article in the Knowledge sidebar, follows the troubleshooting steps, and resolves the case. After resolution, the agent clicks Attach to Case – linking the article to the Case record for reporting on article usage and effectiveness.
• Email article to customer: from the Knowledge sidebar, the agent can insert a formatted Knowledge article excerpt directly into the case email response – or send the customer a link to the article in the self-service portal.
• Create article from case: when an agent resolves a case with no existing Knowledge article, they can click Create Article from the Case record – the Case data pre-populates a Draft article for the agent to format and submit for review.

Customer-Facing Knowledge Portal

Knowledge articles published to a Public channel (Setup, then Knowledge, then Channel Configuration) are accessible via Salesforce Experience Cloud or the Lightning Customer Portal – enabling customer self-service before they submit a case.

Configure the customer portal Knowledge component:

1. Create an Experience Cloud site (Setup, then Digital Experiences, then Sites)
2. Add the Knowledge Search and Article List components to the portal home page
3. Configure Data Categories visible to guest users (unauthenticated visitors) vs. authenticated customers – restrict internal articles from the public channel
4. Configure Article Votes and Comments if you want customer feedback on article helpfulness

Case Deflection metrics: Salesforce tracks how many customers searched Knowledge before submitting a case, how many found a relevant article, and – crucially – how many never submitted a case after finding an article (the deflected cases). This metric quantifies the Knowledge Base’s operational value: each deflected case is a case the support team did not have to handle.

Knowledge Base Maintenance

A Knowledge Base is a living asset that requires active maintenance – outdated articles provide worse service than no article at all:

• Article review schedules: set a Review Date on each published article (a custom date field). A Scheduled Flow runs monthly, identifies articles with Review Date in the past, and creates a Review task for the article author.
• Article effectiveness reporting: Salesforce tracks article views, case attachment frequency, and customer ratings. Articles with high views but low ratings indicate content quality problems.
• Case-to-Article conversion rate: the percentage of resolved cases that resulted in a Knowledge article. A target of 15-25% drives continuous Knowledge Base growth.
• Search term gaps: Knowledge Search analytics (Setup, then Knowledge, then Search Analytics) show which search terms returned zero results – these are topic gaps where new articles are needed.

Integrating Salesforce Knowledge with Agentforce and AI Grounding

One of the most significant advances in Salesforce Knowledge for 2026 is its role as the primary data source for Agentforce AI agents. When you connect your Knowledge Base to Agentforce, published articles serve as grounding data – the AI model retrieves relevant articles before generating responses, ensuring answers are accurate, brand-consistent, and based on your actual product documentation rather than general training data.

To configure Knowledge as an Agentforce data source, navigate to Setup, then Agentforce, then Data Sources, and add the Knowledge object. You can filter which record types and data categories the AI can access – for example, restricting the customer-facing agent to articles in the Customer and Public channels while preventing it from surfacing internal-only troubleshooting notes. This separation is essential: without category-level filtering, an AI agent could inadvertently expose internal pricing notes or unresolved escalation procedures to customers.

Additionally, the Knowledge Feedback feature (now generally available in Spring 2026) creates a feedback loop between AI-generated responses and your article library. When a customer rates an AI-generated answer poorly, the feedback is logged against the underlying article, triggering a review task for the article author. This means your Knowledge Base continuously improves based on real interaction data – not just agent judgement.

Can Salesforce Knowledge be used without Service Cloud?

Yes, but with limitations. Salesforce Knowledge is included at no additional cost in Service Cloud and Unlimited Edition. For Professional and Enterprise Edition users without Service Cloud, Knowledge is available as a paid add-on. Without Service Cloud, you lose access to the Lightning Service Console Knowledge Sidebar – the primary interface for agent article search during case resolution – as well as Case Deflection analytics and Einstein Article Recommendations. Sales Cloud users who add Knowledge typically use it for internal reference articles accessible via the Knowledge tab rather than the agent-assisted case workflow. If your primary use case is customer self-service via an Experience Cloud portal, Knowledge works without Service Cloud as long as you have the Experience Cloud licence.

How do you measure the ROI of a Salesforce Knowledge Base?

The primary ROI metric is case deflection rate – the percentage of customers who searched the Knowledge Base and resolved their issue without submitting a case. Salesforce tracks this natively in the Service Cloud Knowledge Dashboard (Reports, then Dashboards, then Service Cloud Knowledge). Secondary metrics include average handle time per case (cases where agents used a Knowledge article close faster than cases resolved from scratch), first-contact resolution rate (agents with Knowledge access resolve more cases on the first contact), and agent onboarding time (new agents with a comprehensive Knowledge Base reach productivity faster). The standard benchmark is that a 10% improvement in case deflection reduces support headcount growth by approximately one full-time equivalent per 1,000 monthly case volume, based on Gartner self-service research.

What is the difference between Lightning Knowledge and Classic Knowledge?

Classic Knowledge used separate custom objects for each article type – a FAQ article and a Troubleshooting article were stored in completely different Salesforce objects, making cross-type reporting and search configuration complex. Lightning Knowledge consolidates all article types onto a single Knowledge object (Knowledge__kav), with article types differentiated by Record Type – the same architecture as any other Salesforce object. This simplification means admins can use standard Salesforce tools (reports, list views, flows, validation rules) on the Knowledge object without article-type-specific workarounds. Lightning Knowledge also introduced version control, where each edit creates a new version rather than overwriting the published article – preserving history and enabling rollback. Salesforce retired Classic Knowledge in 2020; all orgs should be on Lightning Knowledge by now, but some migrated orgs may still have orphaned Classic article type objects that can be cleaned up via the Migration Tool in Setup.

How should Knowledge articles be structured for Agentforce AI to use them effectively?

For Agentforce to retrieve and use Knowledge articles accurately, article structure matters as much as content accuracy. Each article should have a clear, descriptive Title that matches how customers phrase their questions – conversational phrasing performs better than internal jargon because it matches search patterns. The Summary field (a 250-character plain-text field displayed in search results) should contain the single most important takeaway – this is what Agentforce uses to evaluate article relevance before reading the full body. The article body should use structured HTML with clear headings, numbered steps for procedures, and a Symptoms or When to use this article section at the top – helping the AI identify which customer situations the article addresses. Avoid embedding resolution steps in long unstructured paragraphs; AI models extract structured content more reliably than dense prose. Finally, keep articles focused on a single topic – an article that covers five different issues will be retrieved for all five but will be less useful for any individual situation than five focused articles would be.

Problem: Knowledge Search Returns Irrelevant Articles in the Service Console

Agents frequently complain that the Knowledge Sidebar surfaces off-topic articles – a case about a billing error returns installation guides instead of billing FAQs. The root cause is almost always missing or inconsistent Data Category tagging on articles. Fix this by auditing your article library: export all published articles using the Knowledge Report (Reports, then Knowledge Reports, then All Knowledge Articles), then filter for articles with no data category assignments. Assign appropriate categories to every article. Next, review your Search Configuration (Setup, then Knowledge, then Knowledge Settings, then Search) and ensure the relevant fields – Title, Summary, Article Body – are all enabled for search indexing. Finally, if you are using Einstein Article Recommendations, navigate to Setup, then Einstein, then Einstein Article Recommendations and retrain the model after completing the category audit. Agents should see noticeably improved suggestion relevance within 24-48 hours of retraining.

Problem: Articles Published but Not Visible in the Customer Portal

A common post-launch issue is that agents can see published articles in the Service Console but customers cannot find them in the Experience Cloud portal. This is almost always a channel visibility configuration problem. Check the article’s Channel settings: open the article in edit mode and confirm that the Customer or Public channel checkbox is ticked – articles default to the Internal channel only. Then verify that the Experience Cloud site’s guest user profile has Read access to the Knowledge object (Setup, then Sites, then Guest User Profile, then Object Settings, then Knowledge). If the portal uses Data Category visibility rules, also confirm that guest users have access to the relevant data category group. After correcting any of these, re-publish the article – archive it first, then publish the draft again – to refresh the channel assignment.

Problem: Article Creation from Cases Is Not Maintaining Draft Quality

When agents create articles directly from cases, the resulting drafts are often incomplete – populated with raw case notes, informal language, and missing structured fields. This degrades Knowledge Base quality over time. The fix is to implement a Knowledge Article Template with field-level defaults and a guided layout for each record type. In the page layout for each record type (Setup, then Object Manager, then Knowledge, then Page Layouts), add required fields for Symptoms, Resolution Summary, and Related Product – making these fields required prevents agents from submitting a draft without completing them. Additionally, configure an Approval Process with a checklist-style approval step: the knowledge manager’s approval page layout should display a custom checklist field (a multi-select picklist covering checked grammar, verified steps, added data categories, and added related articles) that must be completed before approval is possible. Pair this with a monthly Knowledge Quality report showing average article ratings per author – visible to team leads – to create ongoing accountability.

The best knowledge-base setup is the one that keeps answers easy to find and update. If the structure is messy, self-service loses much of its value.

Frequently Asked Questions