What's New in Macha — May 2026: Zendesk Routing, Studies & Credit Top-Ups

If you run Macha as an AI agent layer on top of Zendesk, all three of these change day-to-day work. Here's the full rundown — what shipped, why it matters, and the honest edges.

Quick orientation, in case you're new here: Macha doesn't replace your helpdesk. It sits on top of Zendesk (and Freshdesk, Gorgias, Front) and gives AI agents real tools — read a ticket, classify it, reply, tag it, route it, look something up in Stripe or Notion. May's releases added tools to that toolbox and a new way to analyze your ticket history in bulk.

1. Zendesk routing: assign to a group, not just a person (May 2)

Triage agents had a frustrating gap. The Assign Ticket tool could hand a ticket to a named agent — but most real support orgs don't route to people, they route to queues: Billing, Tier 2, Refunds, Onboarding. Picking a specific human at triage time is brittle (they go on PTO, they get overloaded) and it's not how Zendesk's own routing model thinks.

So we fixed it. Three changes landed on May 2:

• Assign Ticket now accepts a group, an agent, or both. An agent can route an incoming ticket straight to the Billing group or Tier 2 Support queue and let Zendesk's normal assignment logic take it from there — instead of always pinning it to one person.
• List Groups — a new read tool that returns every active Zendesk group. This is the unglamorous-but-essential piece: an agent resolves a group name ("Billing") to its Zendesk group ID before it routes, so assignment doesn't fail on a guessed ID.
• Update Ticket Type — a new write tool that sets a ticket's type to problem, incident, question, or task. Triage agents can now categorize tickets the way Zendesk natively models them, without you having to wire up a custom field for it.

Why route to a group instead of a person

This is the difference between automation that survives contact with a real team and automation that doesn't. Zendesk's own routing guidance is built around groups and assignees as trigger actions, and omnichannel routing distributes from group queues — so when Macha assigns to a group, the ticket drops into Zendesk's existing distribution rather than fighting them.

Where this beats a plain Zendesk trigger: a trigger routes on fixed conditions (this support address, this tag, this form). A Macha agent reads the ticket and decides — it can tell a billing dispute from a how-do-I question from a bug report, set the right type, and pick the matching group, even when no tag or address tells it where to go. That's the triage-and-route use case most teams want and can't get from static rules alone.

A worked example, the way you'd configure it:

Watch-out: routing to a group is only as good as your group structure. If your Zendesk has one catch-all group, there's nothing to route to — set up real groups first. And give the agent guidance on which group maps to which topic; "use your judgment" produces inconsistent routing.

2. Studies: batch AI analysis across thousands of tickets (May 30)

This is the headline. Studies lets you point an AI model at a whole batch of records, define the columns you want filled in, and get back a structured grid — one row per record, one column per question. It's the analysis counterpart to Macha's live agents: agents act on tickets in real time; Studies reads your history in bulk.

The shape of a Study is three decisions:

• Source & scope. The first input source is Zendesk Tickets. You pick them with a Zendesk search query and an optional date range, then choose which fields the AI sees — subject, description, status, priority, tags, custom fields, and the full comment thread. Narrowing the fields keeps the model focused and the cost down.
• Typed columns. Each output column is a typed question: boolean (is this billing-related?), single / multi select (category from a fixed list), number (urgency 1–5), or short / long text (a one-line summary). Add per-column guidance so the model decides the way you would. Typed columns are what make the results aggregatable into charts later.
• Run & review. Before anything runs, Studies shows you the exact record count, the credit cost per record, and the total — and nothing kicks off until you confirm. You're billed per successful record; cancelled and errored rows cost nothing.

Two guardrails worth calling out:

• Test runs sample 10–50 records so you can validate your schema and guidance before committing to the full batch. Test runs are flagged so they don't clutter your real history.
• A 20,000-record cap per run keeps cost predictable. For bigger jobs, split by date range. Every run also keeps a frozen snapshot of its config, schema, and model — editing the Study later never rewrites past runs.

The payoff step: a completed run can be pushed to a Knowledge Source. Each result row becomes a document indexed through Macha's existing embeddings pipeline — so the structured insights from your analysis become searchable context your live agents can use. Analyze the past, then let the present agents draw on it.

We wrote a full first-party walkthrough — a real 57-ticket run, 114 credits, the actual findings — in Studies: run an AI analysis across thousands of support tickets. If you've ever argued about "what percentage of our queue is really billing," that post settles it with data.

Watch-outs, honestly:

• It costs real credits. A big batch is a real bill — that's exactly why the estimate-and-confirm gate exists. Test on 10–50 first, narrow the date range, and only feed the fields the question needs.
• Plan-gated. Studies is on Professional and Enterprise — not Starter or Trial.
• One source today. Zendesk Tickets is the first and only input source so far; more record types are on the roadmap. If your data lives elsewhere, Studies can't reach it yet.

3. Credit top-up packs: buy credits without changing plans (May 30)

Credits in Macha are per AI action — every model response or tool run draws down your monthly allowance (default GPT-5.4 Mini is 1 credit; heavier models cost more). Until May, if a busy month or a large Study drained your balance, your only lever was upgrading the whole plan. Now there's a lighter one: one-time top-up packs.

Three packs, bought from the billing page via an in-page Chargebee checkout:

The bigger the pack, the lower the per-credit rate — but each pack is still priced as a premium over the equivalent monthly plan allowance (top-ups buy flexibility, not the cheapest credits). The mechanics that matter:

• Top-up credits never expire. Unlike monthly credits, they roll over month to month and are spent automatically after your monthly allowance is used up — you never have to think about which bucket you're drawing from.
• For active paid subscribers. The top-up section shows for live Starter and Professional subscriptions. Trial users upgrade first; Enterprise plans get bypassed credit checks instead, so they don't need packs.
• Safe to click twice. Purchases are recorded idempotently against the Chargebee invoice ID, so a duplicated webhook can't double-credit (or double-charge) an account.

There's also a "Need more? Buy a top-up pack" deep link right on the Studies Review screen — so if a Study's estimate is bigger than your remaining balance, you can top up without losing your place.

Watch-out: top-ups are a buffer for spikes, not a pricing strategy. If you're buying the large pack every month, the per-credit math on a bigger plan is better — top-ups are deliberately priced as a premium for flexibility. Check the pricing page for the per-credit comparison.

4. Billing UI facelift (May 30)

Small but overdue. The billing Usage card now splits monthly plan usage from your top-up balance — a bar that resets each cycle for the monthly allowance, and a separate, persistent figure for top-up credits, with a note that top-ups are spent after the monthly bucket. Pack cards got cleaner too: price, per-credit rate, and savings badge sit on one row, and the "Most popular" ribbon is now Macha purple to line up with the plan cards above it.

Acting on these Macha product updates: should you do anything?

A quick triage of May's Macha product updates by who they're for:

• Running triage agents on Zendesk? Revisit your routing instructions — you can now send tickets to groups and set ticket type, which usually means simpler, more reliable triage than tag-based hacks. Worth a config pass. See the Zendesk product page for how agents act on tickets.
• Curious what's actually in your queue? Run a Study (Professional/Enterprise) on the last 30 days before you build or change an agent — it tells you what's worth automating and what's rare.
• Bumping into your credit ceiling? Top-up packs are the no-replan release valve. If you hit them every month, look at a plan with a bigger allowance instead.

FAQ

Can a Macha agent route a Zendesk ticket to a group, not a person? Yes — as of May 2, 2026, the Assign Ticket tool accepts a group, an agent, or both. Use the List Groups tool so the agent resolves a group name to its ID before routing.

What is Macha Studies? A batch AI-analysis tool: define typed columns, run them across a batch of Zendesk tickets in parallel, and get a structured grid plus a report with charts. You can export it or push it back as agent knowledge. It's on Professional and Enterprise.

How much does a Study cost? It's credit-based, priced per record by the model you pick, and you see the exact total before you confirm. You're billed only for records that process successfully — cancelled and errored rows cost nothing.

What are credit top-up packs? Three one-time credit packs (1,000 credits for $99, 5,000 credits for $449, 10,000 credits for $799) you buy on top of your monthly allowance. They never expire and are spent automatically after your monthly credits run out. Available to active Starter and Professional subscriptions.

Is Macha replacing my Zendesk? No. Macha is an AI agent layer that runs on top of Zendesk — it reads, replies to, tags, types, and routes tickets using Zendesk's own model. Your helpdesk stays your helpdesk.

Try it

Most of May is live for you right now. Start a 7-day free trial, no credit card required, connect Zendesk, and either point a triage agent at the new routing tools or run your first Study. The full feature reference lives in the docs, and every shipped change is logged on the changelog.

Written by Abbas (Customer Support & AI, Macha) · Reviewed by Ankeet Guha (Co-founder & CTO) · Published 2026-06-24 · Last updated 2026-06-24.