Salesforce Knowledge
Salesforce Knowledge is the knowledge-base feature of Service Cloud.
Definition
Salesforce Knowledge is the knowledge-base feature of Service Cloud. It provides the data model, publishing workflow, multilingual support, channel visibility, and search infrastructure that turn raw content into a customer-service deflection engine. The current generation is called Lightning Knowledge and runs on the KnowledgeArticleVersion object. The older Salesforce Classic Knowledge model still exists in legacy orgs and uses one custom object per article type, but Salesforce has not invested in it since 2019.
Salesforce Knowledge sits at the intersection of three product surfaces: the Service Console where agents attach articles to cases, Experience Cloud where customers self-serve through a Help Center, and Einstein or Agentforce where AI retrieves articles to ground its answers. Enabling Knowledge is not just turning on a feature toggle. It requires license decisions, a record type strategy, data category design, channel governance, and an editorial process. Orgs that skip the design and just enable the feature end up with thousands of zero-view articles inside six months.
How Salesforce Knowledge works as a knowledge-base platform
Lightning Knowledge vs Classic Knowledge
Lightning Knowledge uses a single object, KnowledgeArticleVersion, with record types per article variant. Classic Knowledge used a separate __kav object per article type (FAQ__kav, HowTo__kav, and so on), each with its own page layout and field set. Lightning is the supported path. Classic is in maintenance mode, cannot be enabled in new orgs, and migrating from Classic to Lightning is a one-way conversion that breaks any code referencing the legacy objects. Treat Classic as a constraint on inherited orgs, not a choice.
Licensing: Knowledge user vs Knowledge license
Two license types matter here. The Knowledge User feature license is checked on the user record and gates the ability to read articles. The Salesforce Knowledge license is part of the org-level entitlement. Public-facing readers (Experience Cloud, Help Center, anonymous web visitors) do not consume Knowledge User licenses. Internal agents who read or author articles do. The most common license surprise is discovering that every author needs the Knowledge User flag, not just the editors.
Data Category Groups: visibility and navigation
Data categories are hierarchical groupings used both for navigation and for permission control. You define a Category Group like Products or Regions, build a tree of values inside it, and tag each article with one or more values. Visibility is controlled by Data Category Visibility settings on profiles and roles, with a default and explicit selections. The system supports up to five active category groups and 100 categories per article. Orgs routinely break the model by creating one category group per team. Two or three category groups, designed up front, is the right scale.
Channels and audience segmentation
Each article has channel flags: Internal App, Customer, Partner, Public Knowledge Base. Channels are an AND with sharing settings and data category visibility. An article published to Customer still requires the user to have a Community login and a profile that grants Read on Knowledge. Articles published to Public Knowledge Base are anonymous-readable on the Help Center URL and SEO-indexable by Google. Channel governance is the single biggest source of production tickets. A weekly audit of channel mismatches saves more time than any other Knowledge ops practice.
Translation workflow and the master language
Knowledge supports multilingual content by linking translated article records to a master version. The master is the source of truth. Translations move through their own Draft, Pending Translation, and Published lifecycle. Salesforce can integrate with translation providers like Lionbridge through the Knowledge Translation API, or translations can be authored manually inside the org. When the master republishes, translations get marked outdated. Orgs that ignore the outdated flag end up serving stale Japanese articles to a customer base that thought the company was localized.
Search, synonyms, and promoted terms
Knowledge search uses the same Apache Lucene-based engine as global search. Admins can configure synonyms groups (so "lost password" and "forgot password" return the same articles) and promoted terms (so a specific search query pins a chosen article to position one). Search term history tracks what customers searched, what they clicked, and what they failed to find. The search term report is the cheapest content-gap signal in the product. Ignoring it is the most common Knowledge ops mistake.
AI grounding: Einstein, Agentforce, and retrieval quality
Modern Service Cloud uses Knowledge as the grounding layer for AI. Einstein Reply Recommendations, Einstein Article Recommendations, and Agentforce Service Agent all retrieve from Knowledge to answer customer questions. Retrieval quality is bounded by article quality. Clear titles, two-sentence summaries, structured headings, and short paragraphs win. PDF dumps, screenshot-heavy bodies, and Word paste residue lose. The AI pipeline cannot make a bad article better. It can only retrieve and quote what is there.
Setting up Salesforce Knowledge in a Service Cloud org
Enabling Salesforce Knowledge is a deliberate setup with license, record type, data category, and channel decisions. Done well, the org gets a deflection engine. Done in a hurry, the org gets thousands of unfindable articles.
- Enable Knowledge in Setup
Setup, Knowledge Settings, Edit, check Enable Lightning Knowledge. This is a one-way switch in new orgs and cannot be reversed. Confirm with leadership before flipping it. The setting also enables related objects: KnowledgeArticleVersion, KnowledgeArticleVote, KnowledgeArticleViewStat.
- Assign Knowledge User feature licenses
Open the user record for every internal author and editor. Check Knowledge User. Without this flag, the user cannot create or edit articles even if their profile grants object permissions. The flag is a feature license, separate from the Service Cloud user license.
- Design record types and page layouts
Create record types on KnowledgeArticleVersion for each article variant: FAQ, How-To, Known Issue. Each gets a page layout with the right field set. Keep the list to four or five record types. One record type per team is an antipattern.
- Configure data category groups
Setup, Data Categories, New. Build the hierarchy carefully. Categories are referenced in visibility settings and cannot be deleted if articles use them. Two category groups (one for product, one for audience or region) covers most needs.
- Set Data Category Visibility on profiles
For each profile that needs Knowledge access, open Data Category Visibility and explicitly choose the categories the profile can see. Default visibility (All) is rarely the right choice and silently leaks internal articles to external users.
- Add the Knowledge component to the Service Console
Open the Service Console app, edit a case Lightning record page, drag the Knowledge component into the right rail. Configure auto-suggestion to search article suggestions against the case Subject and Description.
- Publish the first articles and review channel selection
Create three pilot articles, publish each one to the right channels (Internal, Customer, Partner, Public), and confirm visibility in each surface (console, Help Center, agent search). Channel selection is the single biggest source of post-launch tickets.
One-way toggle that activates the KnowledgeArticleVersion object and the Lightning Knowledge UI. Cannot be reversed.
Per-user flag that gates the ability to read or author articles. Counted separately from the Service Cloud user license.
The master language for articles. Translations link back to this language. Cannot be changed after the first article publishes.
When enabled, each article supports translated versions tracked through PendingTranslation and Published states.
Shows excerpts with matched terms highlighted in search results, including in the Service Console article picker.
- Enabling Lightning Knowledge is irreversible in new orgs. Sandbox the decision and rehearse the article model before flipping the production switch.
- Data Category Visibility is profile-based, not user-based. A user inherits visibility from their profile only. Permission sets do not grant data category visibility.
- Channel selection at the article level overrides nothing on the sharing model. An article that is Public but the Knowledge object Org-Wide Default is Private still requires explicit sharing on the article to internal users.
- Search term history is purged on a rolling 90-day window. Build a Saved Report Type and a recurring snapshot if you want long-term content-gap data.
- Translations do not republish automatically. Set up a review queue or content goes stale silently in non-English markets.
Trust & references
Straight from the source - Salesforce's reference material on Salesforce Knowledge.
- Salesforce Knowledge OverviewSalesforce Help
- Set Up Lightning KnowledgeSalesforce Help
- Data Categories in Salesforce KnowledgeSalesforce Help
About the Author
Dipojjal Chakrabarti is a B2C Solution Architect with 29 Salesforce certifications and over 13 years in the Salesforce ecosystem. He runs salesforcedictionary.com to help admins, developers, architects, and cert/interview candidates sharpen their fundamentals. More about Dipojjal.
Test your knowledge
Q1. What customer experience metric would Salesforce Knowledge help improve?
Q2. Which Salesforce Cloud includes Salesforce Knowledge as a key feature?
Q3. What business function does Salesforce Knowledge primarily support?
Discussion
Loading discussion…