Tools / QA & Conversation Analytics
Zendesk QA Review (2026)
The former Klaus, acquired by Zendesk in 2024, now sold as an add-on and no longer carrying a published seat rate.
Some links on this page can be affiliate links. We may earn a commission at no cost to you. This never affects our rankings. How we work.
Visit Zendesk QA ↗Opens www.zendesk.com in a new tab.
What the score measures. Five things we can verify from published material: whether pricing is transparent, whether the product can act on your systems or only answer from documents, whether the vendor still owns its own roadmap, how much independent review evidence exists, and whether the bill stays predictable as volume grows. Weighted, then scored against an ideal platform that scores 5.
BEST FOR: ZENDESK TEAMS WANTING QA WITHOUT ANOTHER VENDOR · LAST UPDATED 2026-08-11
Zendesk QA is owned by Zendesk. Acquired as Klaus and rebranded Zendesk QA. Still sold separately, but development priority follows the Zendesk roadmap. Deal closed 2024.
Zendesk QA in depth
Zendesk QA was Klaus, an Estonian conversation-quality product Zendesk acquired in 2024. It has been folded into Zendesk's portfolio and is now sold as an add-on rather than a standalone product with its own published seat rate.
It does what QA tooling should: scores conversations rather than the small sample a manager can review, against a rubric you define, with calibration so that reviewers agree with each other. Read the exclusions before you take 100% literally, because they are real and they are only in the documentation.
What it does
- Automatic scoring across qualifying conversations
- Custom scorecards and rubrics
- Calibration between human reviewers
- Coaching workflows and agent dashboards
- AI-agent scoring alongside human agents
- PII and PCI redaction before LLM processing
More detail on how it works
The increasingly relevant part is that it scores AI agents as well as humans. As more of your volume gets handled autonomously, the question of whether the AI is doing a good job becomes the same question you already ask about agents — and most teams have no way to answer it.
100% coverage, with the small print
Every qualifying conversation scored instead of the 2% a manager samples, which is the largest single change available in support quality management and the reason QA tooling exists. Qualifying is doing some work in that sentence: the documentation requires a message from both sides and at least 10 words, skips spam and historical conversations, and applies a newly created category only to conversations that arrive after it. Short exchanges are exactly where quality problems hide, so ask what share of your queue clears the bar.
Scorecard design
The step that decides whether this is useful. The tool grades whatever you tell it to, so a long, vague rubric produces confident and meaningless numbers at scale. Few, specific, observable criteria work better.
Scoring your AI agent
As autonomous handling grows, this becomes the more important half. If Fin or Zendesk's AI agent resolves 40% of your tickets, nobody is checking those conversations unless a QA tool does it.
Setting Zendesk QA up
Realistic time to a working deployment: Days. These are the standard steps for this category. We have not published a walkthrough specific to this tool yet.
- 01
Connect your conversation source
Help desk, contact centre or CRM. Read-only in most cases.
- 02
Backfill historical conversations
Gives you a baseline before any coaching starts. Hours for large volumes.
- 03
Build the scorecard
The step that decides whether this is useful. Few, specific, observable criteria beat a long rubric.
- 04
Calibrate against human scores
Have a manager grade fifty conversations by hand and compare. If the AI disagrees systematically, fix the rubric before trusting the dashboard.
- 05
Set the sampling scope
Most teams move from the 2% a manager samples to 100% coverage. Decide what you will act on before you generate it.
- 06
Wire scores into coaching
A dashboard nobody reviews weekly changes nothing. This is the step that produces the return.
The step that takes longer than they imply. Scorecard design. The tool grades whatever you tell it to, so a badly designed rubric produces confident, useless numbers at scale.
What the AI actually does
Every platform here says “AI agent”. It covers a bot that reads your help centre and a system that can refund a customer. This is which one you are buying.
Zendesk QA grades; it does not act.
That is the correct design for a QA tool, and it is worth saying plainly because the marketing page talks about routing high-risk tickets and triggering workflows while the documentation describes scoring, flagging and dashboards.
The one automatic action we found documented is that it switches off an AI prompt of its own that keeps returning nothing useful.
- AutoQA scores conversations against system categories including spelling and grammar, greeting and closing, plus custom categories you write as natural-language prompts
- The 100% coverage claim carries documented exclusions: a conversation needs at least one message from both the customer and the agent and a minimum of 10 words, and spam and historical conversations are skipped
- AI insights grade agents and flag tickets for a human to act on. The only automatic action in the documentation is that a prompt returning over 95% not-applicable for seven consecutive days is marked inactive
- PII and PCI data are redacted before anything is sent to the language model, which is a stronger published commitment than most tools here make
- Capped at 10 active AI prompt-based categories and 20 exact text-match categories per account, and the custom prompt categories run as an Early Access Program you have to be approved for
Who Should Use It
The second list is the more useful one.
Zendesk teams wanting QA
Native integration, published pricing, one vendor. The obvious default if you are already on Zendesk.
Teams with no QA programme
Moving from ad-hoc sampling to full coverage is the biggest available improvement in quality management.
Teams running AI agents
If a meaningful share of your volume is handled autonomously, somebody should be scoring those conversations.
Who should look elsewhere. Teams not on Zendesk, where the integration advantage disappears and specialists compete on merit. Also insufficient for very complex QA programmes, where dedicated tools handle intricate rubrics and workflows better.
Strengths and Weaknesses
What Works
- Scores every qualifying conversation instead of the 2% a manager samples
- Scores AI agents on the same scorecards as humans, which almost nothing else does
- PII and PCI redaction before the LLM sees anything, stated publicly
- Native to Zendesk, and still connects to Salesforce, Intercom, Freshdesk and Front
- The Klaus product underneath is mature rather than newly built
What Does Not
- No standalone price on Zendesk's pricing page; QA is listed as requiring an add-on
- The 100% coverage claim has exclusions that only appear in the documentation
- Custom AI prompt categories are Early Access and capped at 10 per account
- Value is decided by your scorecard design, and that work is yours
- Weakest case if you are not on Zendesk, where specialists compete on merit
Zendesk QA pricing
We previously published this at around $35 per seat per month. Read again on Zendesk's own pricing page in August 2026, no standalone Quality Assurance price appears: QA shows up in the plan comparison as a feature requiring the QA add-on, and the priced product nearest to it is the Workforce Engagement Bundle at US$50 per agent per month paid yearly. Take the $35 as historical until Zendesk publishes a current figure.
For scale, the base plans on the same page are Support Team at $19, Suite Team at $55 and Suite Professional at $115 per agent per month paid yearly, with Copilot a further $50. Quality assurance is therefore a meaningful percentage on top of a seat you are already paying for, not a rounding error.
The full arithmetic
Against MaestroQA and other specialists, the argument is integration: if you already run Zendesk, this connects without work. Against them on capability the specialists are arguably deeper, particularly on complex rubric design. You are buying per seat for the people reviewing rather than per conversation scored, which at least makes the cost easy to forecast.
Custom and private pricing
Zendesk QA does not publish a rate. You have to ask, which means a sales conversation before you can compare it against anything else on this site. Sold as an add-on to Support or Suite. No standalone price on Zendesk's pricing page; the nearest priced product is the Workforce Engagement bundle.
Contact sales ↗Ask two things on that call: what triggers a billable event, and what the rate is at twice your current volume.
INTEGRATIONS
Our Recommendation
Who we would tell to buy this, and who we would not.
Zendesk QA is the straightforward choice for a Zendesk team that wants conversation quality management: native integration and the mature Klaus product underneath it. What it no longer offers is a price you can read, because the standalone rate has come off Zendesk's pricing page. The part that will matter more each year is that it scores AI agents alongside humans — if autonomous handling is taking a growing share of your volume, nobody is checking that work unless something like this does. Spend your effort on the scorecard rather than the vendor choice, because the tool grades exactly what you tell it to.
Where we write about Zendesk QA
9 pieces on this site name it.
- As the volume moves to AI, nobody is reviewing what the AI said unless a QA tool does it.Support QA was built to review what human agents did. Now tier-one is answered by software, and the question is whether the grader can grade it. Here is what the QA platforms in this directory actually document, including the one publishing evidence against its own automation.
- 78 is average. Whether that is good depends entirely on what you sell.The cross-industry average, benchmarks for eight sectors, and the three ways a CSAT score gets inflated before it reaches a dashboard.
- Most metrics dashboards measure the team, not the serviceWhich numbers change a decision, which are vanity, and the four that get gamed the moment you put them on a wall.
- Two teams reporting the same FCR are usually not measuring the same metricThe formula is trivial. What it counts is not — internal measurement overstates FCR by 10 to 20%, and the callback window has no industry standard at all.
- Support structures fail at predictable headcounts, not graduallyTiered against swarming, the four thresholds where a structure stops working, and the roles worth hiring before you think you need them.
- The help desk is the smallest part of a support stack that worksStatus pages, session replay, workforce planning and eleven other categories that change support outcomes more than switching help desk vendor does.
- The job is not disappearing. The entry-level rung is.What actually happened at the companies that automated hardest, and which parts of the role are genuinely at risk.
- The biggest agent in the category was just bought by SalesforceSalesforce is buying Fin for $3.6B. Zendesk alone has bought three, and one of those no longer exists as a product you can buy. If you are signing three years, read this first.
- Can AI grade support conversations?Guide — Quality assurance and scoring
Similar Tools
Others in qa & conversation analytics.
MaestroQA
Teams with an existing QA programme they do not want to change
2.9 / 5 · Not publishedKaizo
Managers who want coaching output, not dashboards
Not rated · Not publishedLevel AI
Contact centres needing intent analytics alongside scoring
Not rated · Not publishedLoris
Teams already using Contentsquare for digital analytics
1.8 / 5 · Not publishedFrequently Asked
The questions people actually search for about Zendesk QA.
Is Zendesk QA the same as Klaus?
Yes. Zendesk acquired Klaus in 2024 and now sells it as Zendesk QA.
How much does Zendesk QA cost?
Zendesk no longer publishes a standalone figure. Its pricing page lists QA as requiring an add-on, and prices the Workforce Engagement Bundle at US$50 per agent per month paid yearly. Our earlier $35 is not on the page today.
Does Zendesk QA work with other help desks?
It integrates with Salesforce, Intercom, Freshdesk and others, though the strongest case is if you already run Zendesk.
Can Zendesk QA score AI agents?
Yes, alongside human agents. This matters increasingly — if AI handles 40% of your tickets, nobody reviews those conversations unless a QA tool does.
Does Zendesk QA really score 100% of conversations?
Of qualifying ones. Its documentation requires a message from both customer and agent and at least 10 words, skips spam and historical conversations, and applies a new category only to conversations arriving after you create it.
What makes a good QA scorecard?
Few, specific, observable criteria. The tool grades whatever you define, so a long vague rubric produces confident meaningless numbers at scale.
What are the main Zendesk QA alternatives?
MaestroQA, Kaizo, EvaluAgent for published pricing without a sales call, Observe.AI if you want QA and live assist together.