How Does Past Tickets AI Turn Solved Zendesk Tickets Into a Knowledge Base?

Past Tickets AI Knowledge Base was a Zendesk Marketplace app from Macha that read your solved tickets, turned them into structured Problem and Solution records called AI Solutions, and showed the top 5 relevant ones in the ticket sidebar when an agent opened a similar new ticket. The app has been retired. Building knowledge from past tickets now happens in the Macha platform, priced by ticket volume from $299 a month for 750 tickets, whose AI Knowledge Builder turns historical tickets into a knowledge base for AI agents.

Every support team has the same problem: the fixes for customer issues are buried in thousands of old tickets, and nobody has time to write them up. The knowledge base drifts out of date, new agents learn by asking senior colleagues, and the team solves the same problems again and again.

Your team has already solved most of the problems it will see this month. The answers exist. They're just locked in ticket history where nobody can find them.

What is Past Tickets AI?

Past Tickets AI Knowledge Base is a Zendesk app that ingests your solved tickets, extracts clean Problem and Solution records (AI Solutions), and surfaces those fixes inside Zendesk when agents need them. It was listed on the Zendesk Marketplace and worked with Zendesk Support plans that have access to solved ticket data and Help Center publishing.

Instead of asking "Has anyone seen this before?" in Slack or digging through old tickets by hand, agents get the fix that worked last time, with the source tickets linked.

How does Past Tickets AI turn tickets into knowledge?

It runs in four steps.

1. Ingestion

After you installed the app, it read your solved tickets, up to your plan's limit. You didn't have to feed it everything. Admins can target tickets with Zendesk Search API queries, filtering by tags, date ranges, subject keywords, brands, groups, channels or custom fields, combined with AND/OR operators.

2. Solution extraction

The AI turns each ticket into a structured AI Solution with:

• Problem statement (what the customer asked)
• Solution (the fix that worked)
• Source tickets (links to the original conversations)

If a customer had trouble with API authentication and an agent solved it by explaining the key regeneration delay, that problem and fix become one AI Solution.

3. Automatic grouping

Similar tickets are consolidated under one AI Solution. If ten customers asked about the same authentication problem, you get one Problem summary instead of ten variations, with all ten source tickets still visible. When new solved tickets arrive, the app updates the existing Solution.

4. In-Zendesk suggestions

When an agent opens a ticket, the sidebar shows the top 5 relevant AI Solutions, ranked by semantic match, each with the Problem summary, the Solution, a relevance score and the source tickets. Agents can also search all Solutions from the sidebar by description, symptoms or phrases, and it matches on meaning rather than exact keywords.

What changes for a support team that uses it?

Fewer escalations for questions someone already answered

When the fix is on screen, an agent can answer on first contact instead of researching or passing the ticket up. The gain depends on how repetitive your queue is: a team that sees the same 50 problems every month gets far more from it than one handling mostly novel cases.

Less reliance on tribal knowledge

Every team has the senior agent who knows everything. When that agent solves a tricky webhook configuration issue, the fix becomes an AI Solution the whole team can reuse, and it stays when they go on holiday or leave.

Faster onboarding for new agents

New hires see how experienced agents handled real tickets, with context and the exact steps, instead of learning by trial and error.

A knowledge base that follows the queue

A traditional knowledge base needs someone to update it. Here, new solved tickets refine existing AI Solutions, so new features, edge cases and regional quirks show up as your team handles them.

What does it look like on a real ticket?

A customer writes in:

"Our API requests are failing with an 'Invalid Key' error even though the key looks correct. What could be wrong?"

Without Past Tickets AI:

• Agent searches old tickets by keyword
• Maybe finds something relevant, maybe doesn't
• Asks a senior colleague on Slack
• Tries possible fixes
• Total time in our example: 20-30 minutes

With Past Tickets AI:

• Agent opens the ticket
• Sidebar shows an AI Solution: "Invalid Key API error when using regenerated credentials"
• Solution: "Regenerated keys can take up to 10 minutes to propagate. Wait or re-authenticate the integration. Ensure the new key is used in both staging and production."
• Source tickets linked for context
• Agent adapts the solution and sends it
• Total time in our example: 2-3 minutes

These are illustrative timings for one worked example, not a measured average. On a ticket like this one the saving is 25+ minutes, and the answer matches what the team told the last customer.

Can it write Help Center articles from tickets?

Yes. From any AI Solution, the app can draft a Help Center article with a problem description, the solution steps, optional troubleshooting notes and examples from similar tickets.

Nothing is published automatically. AI Solutions stay internal until an admin reviews and publishes them. That matters, because an AI-drafted article that goes live unreviewed is how wrong refund policies end up in front of customers.

The app also compares your AI Solutions against existing Help Center articles and flags gaps: topics with no article, articles with partial coverage, and articles that look out of date next to how agents now answer. If an authentication fix has come up 50 times and has no article, that's the next article to write.

How does Past Tickets AI compare to other tools?

A few tools build knowledge from tickets. They differ in who they serve and how they charge.

Zendesk's native AI

Zendesk sells agent-facing AI as Copilot at $50 per agent per month on Suite Professional and above, and bills its AI agents per automated resolution. Zendesk's pricing page lists that rate at $1.50 committed and $2.00 pay-as-you-go. For a 20-agent team, Copilot alone is $1,000 a month before any resolution fees, because the price scales with headcount.

Past Tickets AI has been retired. Its successor, the Macha platform, is priced by ticket volume (from $299 a month for 750 tickets), not by agent count, and has no per-resolution fee.

Ariglad

Ariglad is a Zendesk Marketplace app that creates and updates knowledge base articles from support tickets. Its website doesn't publish prices; it offers a refund of the first month if you're not satisfied. It's aimed at keeping the article library current rather than at putting fixes in front of agents mid-ticket.

Past Tickets AI, now retired, showed relevant AI Solutions in the sidebar while the agent was working the ticket.

eesel AI and Fini

These tools can use past ticket conversations as a knowledge source, but they're built mainly around AI agents that answer customers, not around structured knowledge for human agents.

If you want an AI agent that answers customers from that same ticket history, Macha's main platform does that: its AI Knowledge Builder builds a knowledge base out of historical tickets, and agents then read, reply and write back on Zendesk, Freshdesk, Gorgias, Front, HubSpot or Intercom. Past Tickets AI was the smaller tool for teams that wanted their human agents to reuse what they'd already solved; it has since been retired.

How much does Past Tickets AI cost?

Nothing now: Past Tickets AI has been retired, and its plans are no longer sold. To build knowledge from past tickets today, use the Macha platform, which has one plan priced by ticket volume, starting at $299 a month for 750 tickets, with setup and monitoring by the Macha team included and no per-agent fee. See the pricing page. The free trial gives $50 of usage with no credit card.

How do you get started?

The app is retired, so start on the Macha platform instead:

1. Start the free trial ($50 of usage, no credit card).
2. Connect Zendesk to your Macha org.
3. Run the AI Knowledge Builder on the solved tickets you choose.
4. Review what it builds before any agent uses it.

Frequently asked questions

Q: Will the AI make mistakes?

AI Solutions are generated by AI, so they can be wrong. Admins and knowledge managers review Solutions and decide what gets published, and every Solution links to its source tickets so agents can check the original conversation.

Q: Can we control which tickets are analyzed?

Yes. Admins choose which solved tickets are ingested, using filters or Zendesk Search API queries on tags, dates, brands, groups, channels or custom fields.

Q: Does this replace our Help Center?

No, it feeds it. The app flags gaps in your Help Center and drafts articles from AI Solutions, and an admin publishes them.

Q: What about security and privacy?

The app works inside your Zendesk account and only ingests the tickets an admin selects. Ask Macha for its data-processing terms before you connect it if your security review needs them.

Q: Does it work on every Zendesk plan?

It works with Zendesk Support plans that have access to solved ticket data and Help Center publishing.

Who should use Past Tickets AI?

Teams with a repetitive or technical queue get the most from it: the same product issues, the same configuration questions, the same error messages, answered slightly differently by different agents. If that's your queue, the app itself is retired, so start a Macha trial, build knowledge from a filtered set of recent solved tickets from your busiest group, and check whether the answers match what your best agent would have said.