AThe sample
Only a small sample ever gets graded.
A reviewer pulls a few conversations per agent each week. The rest go unread, including the one that went badly.
Deep on a few, blind on the rest
/ unitQ Support QA
One standard for agents and bots alike.
unitQ Support QA grades what was said and whether the bot or agent took the right action against your standard. That includes evals for your AI bots and coaching for your human agents, with each miss routed to the person who can fix it.
/ The job
Grade every connected conversation, not a weekly sample.
Most QA teams live with two blind spots. Reviewers grade the handful of conversations they have time for. And the AI bot gets graded by the team, or the vendor, that built it. Neither tells you whether your standard is holding across every conversation.
BThe bot’s own report
The bot grades itself.
The team or vendor that built the bot reports how the bot did, on its own terms. Nobody holds it to the bar your agents are held to.
Self-reported, not independent
CunitQ Support QA
One rubric on every conversation.
Your standard, applied to every conversation you connect. Human agents, BPO teams and AI bots get graded the same way, by something that didn’t build any of them.
One bar, every conversation
/ How it works
Set the rubric, then review the grades.
Write it, grade against it, hold the bot to it, route what misses. Support QA does the grading. Your team owns the standard and the coaching.
Your teamunitQ Support QA
- 01Write the rubricYou
- 02Grade every conversationSupport QA
- 03Eval your AI botsSupport QA
- 04Route each missSupport QAYour team
01Write the rubric
Write the standard once.
Put your rubric into unitQ Support QA: your policies, your tone and the right action for each situation. Did the bot or agent issue the refund, escalate the case or confirm identity when required? Start from the scorecard you already use or from your policy docs. It is your standard, not a generic one, and it lives in one place.
- You
- Write the criteria and decide what counts as a pass.
- Support QA
- Turns each criterion into a check it can grade against.
New rubricDraft
What should every conversation get right?
Refund requests, version 3
PolicyToneResolutionAction
- State the refund window correctly
- Confirm identity before any change
- Take the required next step
Save criteria
Applies toOne standard
- AYour human agents
- BYour BPO teams
- CYour AI bot
Change a criterion here and every new conversation is graded against it.
PreviewPublish rubricLive
02Grade every conversation
Every conversation, not a sample.
unitQ Support QA grades the conversations you connect against every criterion in your rubric: what was said and whether the right action followed. Not the few a reviewer had time for that week. Each grade includes the passage or evidence behind it, so anyone can see why a conversation passed or missed.
- Support QA
- Grades each connected conversation, criterion by criterion, with the evidence attached.
- Your team
- Spot-checks grades and tunes a criterion when it reads differently than you meant.
ConversationsGrading
- Refund requestChatMiss
- Late orderEmailMeets
- Billing questionVoiceMeets
- Account changeChatMeets
GradedEvery one
Refund requestRefund policy, v3
Identity confirmedMeets
Refund windowMissed
Required action takenMeets
Customer
Agent
Wrong window quotedThe evidence
03Eval your AI bots
Hold the bot to the same bar.
Point Support QA at your AI bot’s conversations and it grades them on the same rubric as your agents. That is a bot audit from outside the team that built it, not the bot grading its own homework. When an answer drifts from policy, you see the conversation, the criterion and the passage.
- Support QA
- Grades bot conversations on your rubric, next to your agents’ grades.
- Bot owner
- Sees exactly where the bot drifted, and what to change.
Same rubricSide by side
AgentsAI botBPO
Identity confirmedMeets
Refund windowMissed
Required action takenMeets
Bot auditReview
You can request a refund within 60 days.
Your policy says 30 days
Refund windowMissed
Flag for bot ownerFlagged
04Route each miss
Every miss goes to someone who can fix it.
A miss on a human conversation becomes coaching for that agent, with the conversation and the criterion attached. A miss on the bot becomes a fix for whoever owns it. Both arrive with a brief that shows the pattern, not just a score, so the next conversation goes better.
- Support QA
- Routes each miss to its owner with a brief: the criterion, the passages, the pattern.
- Your team
- Coaches the agent or fixes the bot, then watches the next grades.
MissesRouted
- Agent: refund windowTeam lead
- AI bot: refund windowBot owner
- BPO: next stepVendor lead
People getCoaching. Bots get a fix.
BriefFor the team lead
Criterion
State the refund window correctly
Suggested coaching
Send to coachingSent
/ See it
Follow one standard, start to finish.
The four moves, from writing the rubric to the coaching that follows a miss. Get a demo to learn more with your own conversations.
One standard, start to finish
/ Next step
Learn more.
Get a demo and bring your rubric, or a policy that has to show up in every conversation.
Space plays or pauses. Left and right arrows move between chapters. C turns captions on or off. F goes full screen.
Read the transcript
01 One standard
Your team publishes one rubric.
Support QA grades agents, BPO teams and the AI bot against it.
Every miss goes to the person who can fix it.
02 Write the rubric
Name the standard every conversation should meet.
Add the criteria: policy, tone, resolution and the right action.
The same rubric applies to agents, BPO teams and your AI bot.
Publish it once.
03 Grade every conversation
Support QA grades the conversations you connect.
Every one, not a weekly sample.
Grade what was said and whether the right action followed.
The passage or evidence behind the grade comes with it.
04 Eval your AI bots
The bot is graded on the same rubric as your agents.
By Support QA, not by the team that built the bot.
When an answer drifts from policy, you see exactly where.
The miss is flagged for the bot owner.
05 Route each miss
Each miss goes to its owner.
People get coaching. Bots get a fix.
A brief shows the criterion, the passages and the pattern.
Your team sends it to coaching, then watches the next grades.
/ Where it connects
Review the conversations behind a Monitor issue.
Grades sit on the same record as the issues unitQ Monitor already tracks. So when refund complaints rise, you can see whether your agents and your bot are handling those conversations the way your standard says, and route the fix.
unitQ Monitor
IssueRising
Customers say they were told the wrong refund window
Seen inTickets and reviews
Shows the issue
unitQ Support QA
GradesSame issue
Refund window, stated correctly?
AgentsMostly landing
Returns botMissing
Shows which conversations miss
Your team
FixYou own it
Update the bot’s answer, coach the stragglers
Send to the bot owner
Fixes it at the source
Tie grades to what unitQ already knows.
Monitor issues
Line grades up with the issues customers raise in unitQ Monitor, so QA work follows what customers are actually running into.
Proof against a control
Compare a coached group with one that wasn’t coached, so you can tell whether the change worked or the week was just quieter.
Revenue signals in the grade
See which misses show up in conversations that matter most to the business. For the full cost picture, pair it with unitQ Impact.
agentQ is how AI bots show up in Support QA, and it answers questions about grades in plain language, inside unitQ.
MCP is how all six products show up in the agents your teams already use. Explore the use cases
agentQ and MCP are platform features, not separate products.
/ When to use it
When the standard has to hold.
Three moments teams lean on Support QA, and what the grades show at each one.
01 / 03
RubricNew policy
Refunds now need an identity check first
Graded onEvery new conversation
WhoAgents, BPO, AI bot
Is it landing?Grading
Support teamLanding
BPO teamNot yet
AI botLanding
Bot auditLaunch
Is the new returns bot ready for more traffic?
Graded byunitQ Support QA
RubricSame as your agents
Bot conversationReview
Sure, I’ve started your return.
Identity confirmedSkipped
Coaching briefPattern
Next step not stated at the end of billing calls
EvidenceConversations attached
ForTeam lead
One-on-oneYour team
Here’s the conversation. Here’s the criterion. Here’s what good sounds like.
Start coaching
QA leads
Stop picking which conversations to read.
Spend reviewer time on calibration and coaching, not on pulling a sample.
Support ops
Roll out a policy and see it land.
Know which teams, BPO partners and bots picked up the change, and which need help.
AI and bot owners
Get a grade you didn’t write.
See your bot on the same bar as your agents, with the passage behind every miss.
CX leaders
One standard across every kind of agent.
Hold in-house agents, BPO teams and AI bots to the same rubric, and compare them fairly.
SOC 2 Type II report. GDPR and CCPA compliance.
ISO certifications apply to our cloud infrastructure provider.
/ Questions
Questions about the product.
What Support QA grades, who decides what a miss means, and how bots fit in.
Do we have to use your rubric?
No. unitQ Support QA grades against your standard: your policies, your tone, your resolution steps and whether the bot or agent took the right action. That can mean issuing a refund, escalating a case or confirming identity when required. Bring the scorecard you already use, or write the criteria from your policy docs.
Does it really grade every conversation?
It grades the conversations you connect, instead of a sample a reviewer picks. Which channels and systems you connect depends on your setup, and we will walk through that with you in a demo.
How are AI bots graded?
On the same rubric as your human agents, by Support QA rather than by the team or vendor that built the bot. You see the bot’s grades next to your agents’, with the passage behind every miss.
Who decides what a miss means?
Your team does. Every grade shows the passage behind it. If Support QA reads a criterion differently than you meant, you tune the criterion. Coaching stays with the people who know your customers.
Does it replace our QA reviewers?
No. It takes over the grading so reviewers can spend their time on calibration and coaching, instead of choosing which handful of conversations to read.
What about proof and revenue?
Support QA can compare a coached group against a control, and bring revenue signals into the grade. What that looks like depends on the data you connect, so we show it with your setup in a demo rather than promise a number here.
How does it fit with the rest of unitQ?
Grades tie back to the issues in unitQ Monitor. unitQ Impact puts a number on what those issues cost. unitQ Research asks customers why. unitQ Social and unitQ Compete bring in public conversations about you and your competitors. agentQ answers questions about all of it in plain language, and MCP brings the products into the agents your teams already use.
For connectors, see integrations.
Where can I learn more?
Take the tour to learn more, or get a demo. Bring your rubric.
/ Bring your rubric
See how Support QA checks your policy.
Get a demo and bring the rubric your team already uses, or a policy that has to show up in every conversation, agents and bots alike.