What This Update Actually Is
HubSpot's Breeze Knowledge Base Agent is a Breeze AI tool built into Service Hub. It reads your closed support tickets, compares that data against your existing knowledge base articles, and surfaces two types of output: drafted new articles for topics your content doesn't yet cover, and targeted edit suggestions for articles that have gone stale, unclear, or off-brand.
The pricing change is the news here. Starting September 16, 2026, every new article the agent drafts costs 200 credits, which is $2. Suggested edits to articles you already have are included in your Service Hub seat at no extra credit cost. This is a straight shift from a free beta to a credit-consuming feature.
The feature is available to Service Hub Professional and Enterprise portals. It works with one knowledge base per portal. Note that a Super Admin must enable the Customer conversation data toggle in AI settings before any of this can run.
Why HubSpot Shipped This
Knowledge base content decays. That's not a theory. It's a pattern we've watched play out across hundreds of portals. A team publishes a solid set of articles, then six months later a product update ships, pricing changes, or the brand refreshes its voice. Nobody updates the knowledge base because nobody has a system to catch what's broken.
The result is a double problem. Humans searching for answers find outdated content and lose trust. Support agents keep fielding tickets for questions the knowledge base should answer but doesn't. Manual audits are expensive, easy to skip, and usually reactive.
HubSpot's answer is to use the ticket data you're already generating as a feedback loop. Every closed ticket is a signal about what humans needed help with. The agent mines those signals to tell you where your content has gaps and where it's gotten fuzzy.
How to Use It Step by Step
Before you start, confirm these three things are true for your portal:
- A Super Admin has enabled the Customer conversation data switch in AI settings.
- You have at least five closed ticket records in the account.
- A knowledge base has already been created and is live.
You'll also need Service Access with Knowledge Base Articles permission to complete the setup.
Setup steps:
- In HubSpot, click More, then navigate to Service, then Knowledge Base. If More doesn't appear, go directly to Service, then Knowledge Base.
- Click the Suggestions tab on the knowledge base index page.
- Click Create Knowledge Base Agent.
- Use the Knowledge Base dropdown to select your knowledge base.
- Use the Pipelines dropdown to select which ticket pipelines the agent should pull from. It will only scan closed tickets in the pipelines you choose.
- Click Finish. The agent will begin scanning and populating the Suggestions tab.
Reviewing new article drafts:
- Go to Service, then Knowledge Base, then the Suggestions tab.
- Use the Filter dropdown and select Create to see new article suggestions. Sort by Sources (most) to prioritize articles backed by the most tickets.
- Click a suggestion to open the draft. Click Source conversations to see which tickets and gaps informed it.
- Click Save as draft to keep editing, or Dismiss to reject. Remember: accepting a new article draft consumes 200 credits.
Reviewing edit suggestions for existing articles:
- On the Suggestions tab, use the Filter dropdown and select Update.
- Click a suggestion to see the recommended change and the edit type (Clarity, Stale data, Branding, AEO, Voice and formatting, Incomplete, or Decay).
- Click Accept all and publish, Accept all and save as draft, or use individual Accept and Dismiss controls for each suggestion item. These edits don't consume credits.
What It Touches in Your HubSpot Strategy
This isn't just a content tool. It connects several layers of your service and RevOps setup in ways worth mapping out before you flip the switch.
Service Hub ticket pipelines: the agent's quality depends on the quality of your ticket data. If your pipelines are messy, poorly categorized, or have very few closed tickets, the suggestions will reflect that. Clean pipeline hygiene directly improves agent output.
Knowledge base structure: the agent works with one knowledge base. If you're running multiple knowledge bases for different brands or regions, you'll need to choose which one gets the agent. That's a strategic decision, not just a setup click.
Credit budget and governance: new article generation costs credits. That means someone on your team needs to own the decision of which drafts to accept. It also means you should set a rough monthly credit budget for this feature before the agent starts producing suggestions at scale.
Key Takeaway
Edit suggestions for existing articles are free on Service Hub seats. Only generating brand-new article drafts costs 200 credits ($2) each. Triage your Suggestions tab with a filter on Update first to capture the free wins before spending any credits.
AEO and content visibility: one of the seven edit suggestion types is Answer Engine Optimization. That's specifically about how your articles appear in AI-generated answers and search features. If you're thinking about how your content performs in AI overviews and conversational search, this is directly relevant. The agent connects to a broader shift happening across HubSpot's Breeze platform.
Customer Agent and self-service strategy: if you're running or planning HubSpot's Customer Agent for chat-based self-service, a well-maintained knowledge base is the foundation that agent draws from. The Knowledge Base Agent and Customer Agent are designed to work together. We covered how Customer Agent Segments let you personalize those self-service experiences by audience, and fresh, accurate knowledge base content is what makes that personalization actually useful.
Key Takeaway
The Decay suggestion type flags articles where traffic or engagement has dropped significantly. This is a useful signal for content pruning decisions. An article flagged for decay may need a rewrite, a redirect, or outright removal. Don't auto-accept every suggestion without reviewing that context first.
Language and localization: the agent generates drafts and suggestions in the primary language set on the knowledge base it's connected to. If you support multiple languages, plan for that. You can't point the agent at a secondary-language knowledge base at the same time.
Who Should Care Most
This update matters most to a specific set of roles and company profiles. Here's how to think about it:
- Customer success and support leaders who manage a knowledge base and have more than 50 closed tickets per month. The agent's signal quality scales with ticket volume.
- Content ops and marketing ops humans who own knowledge base governance but don't have bandwidth for manual audits. This automates the detection work and puts review decisions in front of the right humans.
- RevOps leads responsible for deflection rates and support costs. Every high-quality self-service article that prevents a ticket is a measurable cost reduction. This tool systematically builds that library.
- Companies on Service Hub Professional or Enterprise who are already using Customer Agent or planning to. A stale, incomplete knowledge base is a liability for any AI-powered chat or self-service setup.
- Portals that have built a knowledge base but haven't touched it in six-plus months. The Stale data, Clarity, and Decay suggestion types were built exactly for this situation.
George's Take
I've watched teams build beautiful knowledge bases, publish them with pride, and then never touch them again. The articles get stale, the humans who need answers find wrong information, and support ticket volume climbs for questions the KB should already answer. That's not a content problem. It's a system problem. What I like about the Knowledge Base Agent is that it treats your ticket data as a feedback loop instead of a dead end. Every closed ticket is already telling you something about where your content has a gap. The agent just makes that signal visible and actionable. My caution: don't let the agent run on autopilot. The Decay and Branding suggestion types in particular need a human judgment call before you publish. Use the Source conversations link on every suggestion before you accept. Know why the agent flagged it, not just that it did.
“Every closed ticket is already telling you something about where your content has a gap. The Knowledge Base Agent just makes that signal visible and actionable. But a human still has to decide what to do with it.”
If you want to see how this fits into a broader Breeze AI strategy, our breakdown of what HubSpot's Breeze Assistant can actually do across your portal is a solid next read.
If you're on Service Hub Pro or Enterprise and want to build a knowledge base system that actually stays accurate, let's map it out together. Book a strategy call with the Sidekick team and we'll show you exactly how to set this up so it flourishes over time, not just on launch day.
Frequently Asked Questions
How much does HubSpot Knowledge Base Agent cost?
Starting September 16, 2026, generating a new knowledge base article using the Knowledge Base Agent costs 200 HubSpot credits, which equals $2 per article. Suggested edits to existing knowledge base articles are included in Service Hub Professional and Enterprise seats at no extra credit cost.
What does HubSpot's Knowledge Base Agent actually do?
It scans your closed support tickets and existing knowledge base articles, then surfaces two types of output: drafts for new articles on topics your KB doesn't cover, and edit suggestions for existing articles that are unclear, outdated, incomplete, off-brand, or showing declining engagement.
What are the requirements to set up HubSpot Knowledge Base Agent?
You need Service Hub Professional or Enterprise, at least five closed ticket records in your portal, an existing knowledge base, and a Super Admin must enable the Customer conversation data toggle in AI settings. You also need Service Access with Knowledge Base Articles permission to complete the setup.
What types of edit suggestions does Knowledge Base Agent generate?
The agent flags seven edit types: Clarity (unclear phrasing), Incomplete (missing information), Stale data (outdated facts), Branding (off-brand language), Answer Engine Optimization (AEO for AI search visibility), Voice and formatting (tone consistency), and Decay (articles with significantly declining traffic or engagement).
Can HubSpot Knowledge Base Agent work across multiple knowledge bases?
No. The Knowledge Base Agent is limited to one knowledge base per portal. During setup, you select the knowledge base and ticket pipelines you want the agent to scan. If you manage multiple knowledge bases, you'll need to choose which one the agent analyzes.
Who in my organization should own the Knowledge Base Agent review process?
Typically, a customer success lead or content ops manager should own the review workflow. They should triage edit suggestions first (these are free), then evaluate new article drafts before accepting them and spending credits. Someone with both service context and content judgment gets the best results from this tool.
