How Do You Keep a Zendesk Knowledge Base Updated From Solved Tickets?

You keep a Zendesk knowledge base current from solved tickets by having software read each solved ticket, pull out the problem and the fix, and queue that fix for an admin to publish. Past Tickets AI Knowledge Base, Macha's Zendesk Marketplace app, did exactly that. The app has been retired; building knowledge from solved tickets now runs on the Macha platform, priced by ticket volume from $299 a month for 750 tickets.

The problem it targets is familiar. Your product shipped a new feature last Tuesday. By Wednesday, tickets about it were rolling in. By Friday, your agents had worked out the common issues and the fixes. Three weeks later, none of that has reached your documentation, because the admin has fifty other priorities and the agents are busy answering tickets.

That gap is widest in software, SaaS and other teams whose product changes faster than anyone can write it down.

"In fast-moving companies, the gap between what agents know and what's documented grows wider every day."

Why does a Zendesk knowledge base fall out of date?

Here's what the manual loop looks like in most support teams:

The manual process nobody has time for:

1. Agents solve complex issues daily but don't document them
2. Admins periodically run reports in Zendesk Explore to find knowledge gaps
3. Someone (eventually) assigns documentation tasks
4. Agents write up solutions from memory, if they remember the details
5. Content gets reviewed, edited, and maybe published
6. By then, the product has changed again

Zendesk does give you tools for the upkeep side. Article verification rules send reminders when an article is due for review, but they need a Suite Enterprise or Enterprise Plus plan (or Knowledge Enterprise). They remind an owner to check an article; they don't write the new one. For a product that releases weekly or monthly, the reminders pile up faster than the fixes.

The time problem

A knowledge base needs constant attention: tracking views, spotting gaps, updating articles. When the queue is burning, that work is the first thing pushed aside. It's never quite as urgent as the tickets in front of you.

The discovery problem

Zendesk Explore's knowledge base reporting shows article views, comments and search trends. It won't tell you which fix is missing. You can see customers searching for something, but connecting those failed searches to the ticket where an agent solved it takes detective work.

The expertise problem

Your best agents, the ones solving the trickiest problems, are too valuable on the front line to spend afternoons writing articles. Newer agents don't know enough to document complex fixes accurately. So the knowledge stays with a few senior people, and it bottlenecks when they're out.

The version history problem

Zendesk's article revisions (view and restore earlier versions) need Suite Growth or higher, or Knowledge Professional or Enterprise. On lower plans, tracking what changed and when is a manual job, and keeping articles in step across product versions, regions and customer tiers gets heavy fast.

How does Past Tickets AI turn solved tickets into knowledge?

Past Tickets AI Knowledge Base reads the solved Zendesk tickets you choose and turns similar ones into "AI Solutions": structured problem-and-fix blocks grounded in what your agents actually did. It then compares those solutions against your Help Center and flags missing articles, partial coverage and outdated information.

The mechanism: solved tickets feed the knowledge, and an admin decides what gets published.

Nothing is auto-published. AI Solutions stay internal until an admin reviews one and publishes it as a Help Center article. That's a deliberate choice: the drafting is automated, the judgment isn't.

How the traditional process runs (and why it fails)

How the Past Tickets AI loop runs

What does this look like for a software team shipping every two weeks?

Say your development team ships every two weeks. Each release brings:

• New features customers don't understand yet
• API changes that break integrations
• UI updates that confuse existing users
• Bug fixes that create new edge cases

By the time someone documents a feature by hand, it has often changed again.

Knowledge capture from the week's tickets: When Monday's release causes integration issues, agents work out the fixes through the week. Once those tickets are solved and ingested, the fixes are available to the rest of the team as AI Solutions instead of sitting in one agent's closed tickets.

Pattern recognition: Several tickets about the same API error collapse into one AI Solution, so the team sees one answer instead of five slightly different ones.

Gap detection: Because solutions are checked against the Help Center, the admin sees which of them have no article yet, which articles only partly cover the fix, and which look outdated.

Does it help teams outside software?

Yes, anywhere the answer changes faster than the documentation:

E-Commerce

Product catalogs change daily. Shipping policies vary by region. Promotions have exceptions. Solved tickets record how agents handled each variation, which is the raw material for a playbook of edge cases.

Financial Services

Regulations change and new products launch. Here the admin review step matters most: an AI Solution drafted from tickets should be checked against current policy before anyone publishes it.

Healthcare & Services

Procedure updates and insurance changes land constantly. Capturing how experienced agents handled them makes that know-how available to the rest of the team, with a human deciding what becomes official.

Which gaps does it show that Zendesk Explore doesn't?

Zendesk Explore shows article views and search patterns. It can't tell you which fixes exist in your tickets but not in your Help Center. Past Tickets AI Knowledge Base compares the two and flags:

• Recurring problems with no Help Center article at all
• Articles that only partly cover the fix agents use
• Articles that look outdated next to recent tickets
• Which topics need formal Help Center articles next

That gives the knowledge manager a ranked to-do list built from real tickets rather than a guess about what's missing.

How long does setup take?

There's no migration. The app works with Zendesk Support plans that have access to solved ticket data and Help Center publishing:

What does Past Tickets AI cost, and how do you work out the return?

Past Tickets AI has been retired, so its plans are no longer sold. The Macha platform now handles this, priced by ticket volume (see pricing):

We haven't published measured before-and-after figures, so don't plan on a vendor's number. Do the arithmetic on your own queue: if 1,000 solved tickets a month each save an agent 5 minutes of research, that's about 83 hours a month against a $299 bill. Beyond handle time, the savings show up as:

• Fewer escalations to senior staff
• Faster onboarding, since new agents learn from proven fixes
• Higher first-contact resolution
• Fewer customers searching a Help Center that doesn't have the answer

Why doesn't Knowledge-Centered Service fix this on its own?

Knowledge-Centered Service (KCS) has been the standard advice since the 1990s: agents should document as they solve. In practice it rarely sticks, because agents are measured on tickets closed, not articles written. The incentive points the wrong way.

Past Tickets AI takes the writing step off the agent. They solve tickets; the app drafts the knowledge; an admin approves it. KCS still describes the goal, but nobody has to fight their own metrics to get there.

Who controls the data and what gets published?

The controls sit with your admins:

• Admins choose which solved tickets are ingested
• AI Solutions stay internal until an admin publishes them
• Nothing is auto-published to your Help Center
• The app is aimed at admins and knowledge managers, who review solutions and decide publication
• Each solution traces back to real support cases

If your security review needs more (retention, sub-processors, training use), ask the Macha team for the current documentation rather than relying on a blog post.

What if you want AI agents to use that knowledge, not just agents in the sidebar?

The marketplace app is Zendesk-only and feeds human agents and the Help Center. The broader Macha platform has its own AI Knowledge Builder, which builds a knowledge base from historical tickets for AI agents that draft replies and act inside the ticket. It runs on Zendesk, Freshdesk, Gorgias, Front, HubSpot or Intercom, and the plan is priced by ticket volume, starting at $299 a month for up to 750 tickets (see pricing).

Documentation stops being a separate task and becomes a byproduct of solving tickets. That shift doesn't need anything exotic, just a step that reads what your team already wrote.

How do you get started?

1. Assess your current state: How many tickets could be solved faster with better documentation? In Zendesk Explore, look at tickets with several public comments; they often mean research time.
2. Calculate your ROI: Multiply a realistic handle-time saving (even 5 minutes) by your monthly ticket volume, and compare it to the plan price.
3. Start small: Ingest your highest-volume ticket categories first and expand from there.
4. Measure the impact: Track first-contact resolution, average handle time and agent satisfaction for a few weeks before and after.

Your team already creates useful knowledge every day. The problem isn't expertise; it's that the knowledge disappears into closed tickets. The best documentation comes from real fixes to real problems, and those fixes are already in your tickets.

"For fast-moving teams, reading solved tickets is the only way documentation keeps up."