All Docs
FeaturesSaaS FactoryUpdated February 19, 2026

Blog: Your Support System Should Get Smarter Every Time It Helps Someone

Your Support System Should Get Smarter Every Time It Helps Someone

v1.0.39 — Intelligence Release

Every time your AI support agent closes a ticket, something valuable is created: a record of a real problem and a proven solution. Until now, that record lived in a closed ticket and went nowhere. Starting with v1.0.39, it becomes the seed of a knowledge base article.

The Problem with Manual Knowledge Bases

Knowledge bases are only useful if they're kept up to date. And keeping them up to date has always required someone to sit down, recognize that a support interaction was worth documenting, write the article, tag it correctly, and publish it — after already handling the ticket that generated the insight.

In practice, this doesn't happen consistently. The busier your support operation, the wider the gap between what your team knows and what's actually written down.

The result: the AI support agent answers the same question from institutional memory every time, instead of pulling from a growing, explicit knowledge base. Each interaction is equally hard. Nothing compounds.

How v1.0.39 Changes This

The new auto-generation pipeline connects two signals that were already being collected — AI resolution confidence and customer CSAT scores — and turns them into a third thing: draft knowledge base articles.

Here's the flow:

  1. The AI agent resolves a ticket with high confidence.
  2. The post-resolution CSAT follow-up goes out automatically.
  3. If the customer responds with a score of 4 or 5, the platform knows two things: the AI got it right, and the customer agrees.
  4. Claude takes the full ticket thread and produces a structured article draft — problem, solution, tags — stripped of customer-specific details.
  5. The draft lands in a new Suggested Articles section in your Knowledge Base tab, inactive and invisible to customers.
  6. A reviewer approves it with one click.
  7. The article is live. The next customer with the same issue gets a faster answer. The AI cites the article directly.

That's it. No manual writing step. No deciding whether something is worth documenting. The system decides based on outcomes: did the AI solve it, and did the customer confirm it worked?

What Compounds Over Time

The first approved article doesn't change much. But the tenth, the fiftieth, the two hundredth — each one makes the next resolution faster and more accurate. The AI has more to work with. Customers wait less. Escalations drop.

This is what a self-learning support system looks like: not a fixed set of FAQs written at launch, but a knowledge base that grows at the same rate your support volume does, with no additional labor cost.

What Stays the Same

Nothing about the existing knowledge base workflow changes. The Add Article form is still there. Manually authored articles work exactly as before. The Suggested Articles queue is additive — it surfaces drafts when they exist, and stays out of the way when they don't.

Reviewers remain in the loop. No article goes live without a human approving it. The AI drafts; you decide.

Available Now

The self-learning knowledge base pipeline is included in v1.0.39. No configuration is required — the pipeline activates automatically when the CSAT and resolution confidence thresholds are met.

For full technical details, see the Self-Learning Knowledge Base feature documentation.