Use cases
Call quality monitoring
With CONTACT.WORK, you know how every call went – even the ones nobody listened to. Speech analytics scores each conversation against your standards and, by morning, flags the calls where an agent broke the script or the customer hung up unhappy. Your QA team reviews only those. Complaints are checked against the recording, and when a call gets tough, a supervisor can coach the agent or take over.
CustomerEmily+7 9•• ••• 18 52Topicpayment questiondetected by AIRating86 of 100by scorecardIn CRMsummary and scoresentChallenges
Common problems with manual QA
Solution
What you get
Scoring
One standard for every call
Speech analytics transcribes each conversation and scores it against your weighted criteria. A step counts even if the agent put it in their own words, and every score comes with a quote from the call.
- Weighted QA scorecard
- Meaning-based scoring
- Quote for every score
Greetingdone+10 pointsNeeds discoveryin own wordscountedClosing summaryskipped−10 pointsSummary72 of 100quote for every itemAlerts
Problem calls flagged automatically
If an agent skips a critical step, the supervisor gets an alert right away. Each morning, a review list is ready: violations, unhappy customers, and repeat calls about the same issue.
- Critical miss
- Unhappy customer
- Repeat calls
Critical misses4alert sentUnhappy customers7for reviewRepeat calls11same questionTotal to review22 callsof 1,380 checkedLive calls
Mistakes fixed before the call ends
A supervisor joins the line silently and can listen, whisper to the agent so only the agent hears, or take the call over. Agents can also call in a supervisor with one button.
- Listen
- Whisper
- Barge-in
StaffAnna, ext. 2144:12 on callCustomer unhappyraised voicecalled the supervisorManagerconnectedcustomer can't hearModewhispercan take over the callSearch
Complaints checked in minutes
Find a recording and its transcript by the caller's number or by words spoken on the call, and it opens at the right moment. AI writes a short summary, so you don't have to listen to the whole call. You decide how long recordings are kept.
- By number
- By keyword
- Call summary
Room+7 9•• ••• 41 073 callsPhrase"discount"found at 2:35Summarydonebilling questionRetention12 monthsset by the companyCoaching
Agents improve and scores get more accurate
Agents see their scores with explanations; supervisors see how each agent's results change over time. Strong calls become examples for new hires, and scores corrected by supervisors are used to fine-tune the model.
- Own scores
- Trends
- Example calls
Average score68 → 81over 4 weeksCommon misscall wrap-uphalf as oftenReference answers3 callsfor new hiresScores corrected2for retrainingExamples
Four situations for your QA team
How it worked before and how it works now.
The customer is sure they were promised a discount
A customer says the agent promised a discount, but it isn't on the bill. Before, someone searched the call log for the recording and listened to the agent's entire shift. Now the call is found by the customer's phone number, the transcript jumps straight to the right moment, and it's clear no discount was promised. The customer gets an answer right away, and the agent isn't pulled off the line. If the agent did overpromise, it's on the same recording, and the conversation with them is based on facts.
- Search by number
- The exact line
- Fast reply
Complaintpromised a discountfrom customerConversationfound by numberOct 12Line at 3:10no discount promisedquoteCustomer replysame dayno staff interviewsFirst week on the phones
A new agent starts taking calls. Before, a mentor could listen to a couple of their calls a day and learned about mistakes too late. Now every call is scored right away, the mentor sees which steps the new hire skips most often, and in a difficult call coaches them through whisper mode. The AI assistant finds the right article for them. After a week, it's clear which mistakes are gone and which still need work.
- Instant scoring
- Most-missed steps
- Live coaching
Calls checkedall 146right after the callCommon misscall wrap-upto reviewWhisper2 callsmentor coachedAI suggestionsfrom knowledge baseon the call topicGive the recording notice, don't overpromise
You need agents to tell customers the call is being recorded – which many states require – and to avoid guarantees you can't back. Before, this was spot-checked, and violations surfaced through complaints. Now every call is checked for the required words and for prohibited ones. The supervisor sees a miss the same day. The agent replays their own recording and hears what they left out.
- Required phrases
- Prohibited promises
- Same-day alert
Recording noticeskipped in 3 callsalertExtra promise1 callfor reviewReviewedall callsin a dayFor the agentrecording senthears the miss himselfAn agent challenges a score
An agent is sure they completed a step, but lost points anyway. Before, it came down to one person's word against another's. Now each score comes with a quote from the call: the supervisor plays that clip, hears that the agent said it differently but got it right, and corrects the score. The correction is used for fine-tuning, so the same wording counts next time. The agent sees how the dispute ended in their own scores.
- Quote in the score
- Correction
- Fine-tuning
Stageneeds discoverypoints deductedQuote"what matters most to you now?"at 2:05Supervisor decisionstep countedscore correctedFine-tuningcorrection appliedwill count next timeYou decide what counts as a violation and when to send an alert; we set it up at launch.
Features
What else is included in quality monitoring
What gets checked, how scores are set, how agents are trained, and how they get help on a live call.
What gets checked
- All call typesInbound and outbound customer calls are checked, as well as internal calls between employees.
- Written conversationsEmails and chat messages are checked just as strictly as calls.
- Your recording archiveAt launch, your existing archive is checked too, so you know where to start.
- Recordings from your PBXCalls from any phone system work, as long as the voices are recorded on separate tracks (dual-channel recording).
- Talk-time ratioSee how long the agent talked versus the customer, and where the call dragged on.
- Call topicsEach call is tagged with a topic, and you see which topics are growing.
- Customer sentimentSee how the customer's mood shifted and the exact line where the conflict started.
- Competitor mentionsCalls that mention a competitor or a key service go straight to review.
- Sales stagescustomized for youA report shows exactly where in the conversation agents lose customers.
How scoring works
- Manual scoring alongside AIReviewers can score calls themselves on the same form. Reports show which scores came from AI and which from a person.
- Reviewer calibrationSeveral reviewers score the same call, and the system compares their scores and highlights the gaps.
- Testing a new scorecardA revised scorecard is tested on past recordings first, and only then do its scores count.
- Auto-fail on critical violationsMiss a critical item, and the whole call fails, no matter how well the rest went.
- Customer rating alongsideAfter the call, the customer leaves a rating, and calls rated low by both the customer and the reviewer are reviewed first.
- Self-assessmentAgents score their own calls and then compare with the supervisor's score.
- Scores in the summaryAn agent's QA score sits next to their other metrics in one summary.
- Branches and departmentsCompare call quality across branches and departments.
Training and oversight
- Training assignmentscustomized for youAfter a review, the supervisor assigns a task and sees whether it was completed and whether scores changed.
- Scores and resultscustomized for youSee whether an agent's sales and customer ratings went up after training.
- Leaderboards and contestsEveryone sees where they rank among peers, and top results are visible to all.
- Results in your CRMThrough the API, the score, topic, and call summary are saved to the right customer record in your CRM.
- Call translationcustomized for youCalls in another language are translated so a supervisor can review them.
- Audit logThe system logs who listened to calls and who changed scores.
- Personal data redactionCard numbers, ID details, and other personal customer information are hidden in transcripts.
During the call
- AI assistant suggestionsBased on the call topic, the AI assistant instantly shows the agent the right article and the next script step.
- "Call a supervisor" buttonWhen a conversation gets difficult, the agent asks a supervisor for help with one click.
- Broadcast announcementsSupervisors send announcements to agents or the team, for example about a new rule.
- Chat with a supervisorAgents ask a supervisor a question in chat without interrupting the customer call.
Modules
Contact center modules
Launch
Implementation step by step
- 01AssessmentWe look at how calls are reviewed today: who listens, by what rules, and how many calls get scored.
- 02ChecklistWith your reviewers, we build the QA scorecard: call steps, their weights, required phrases, and prohibited language.
- 03RecordingsWe confirm the agent and customer are recorded separately and connect your PBX if calls go through it.
- 04Pilot scoringWe score a sample of past recordings, compare the results with your reviewers' scores, and refine the criteria where they differ.
- 05AccessWe set up roles: who can see recordings and scores, and who can join calls and coach.
- 06Go-liveWe turn on scoring for every call and the morning review list. Disputed scores help refine the scorecard.
Preparation
What to decide in advance
Q&A
Frequently asked questions
Can you score calls from our PBX?
As long as your PBX records each side of the call separately. The platform's built-in telephony works that way out of the box. You can score recordings from past months, too.
Are chats and emails scored?
They're scored against the same criteria as calls. The separate-track requirement doesn't apply to written conversations.
How much does call scoring cost?
In our cloud, call scoring costs from $0.015 per minute. On your own server, there are no per-minute fees. Launch and scorecard setup are done by our support team and billed hourly under your support plan.
Will a step count if the agent says it in their own words?
The system looks at meaning, so exact wording isn't required. A supervisor can correct a wrong score manually, and the correction is used for fine-tuning.
How do we get reviewers to score consistently?
Use calibration: several reviewers score the same call, and the system shows where their scores differ. You review the gaps and refine the scorecard.
Will the customer hear the supervisor?
When a supervisor listens or whispers, the customer can't hear them. If the supervisor takes over, the customer continues the conversation with them.
Can agents see their own scores?
Agents see their scores along with comments and quotes, so they can work through mistakes on their own or with a supervisor.
Is it worth it for a small team?
Call scoring, call search, and live supervisor coaching are useful even with just a few agents.
Glossary
Terms on this page
- Supervisor
- Leads a team of agents: watches the queue and quality, helps and coaches. Learn more →
- Whisper coaching
- Supervisor coaching that only the agent can hear. Learn more →
- IP PBX
- A company phone system that runs over the internet and your local network. Learn more →
- Model fine-tuning
- An existing model is trained further on company materials so it speaks the company's language. Learn more →
- Speech analytics
- The system analyzes recorded calls automatically: topics, script adherence, customer emotions. Learn more →
- Cloud deployment (cloud)
- The platform runs on the vendor's servers, and agents connect over the internet. Learn more →
- Dual-channel recording
- Each speaker is recorded on a separate track. Learn more →
- CRM – customer relationship management
- Where a company keeps its customers, deals, and full communication history. Learn more →
- Contact center quality management
- Calls and chats are scored against a scorecard to find mistakes and coach agents. Learn more →
- Sentiment analysis
- The system automatically detects whether a customer is happy or frustrated. Learn more →
- API (application programming interface)
- The way programs exchange data with each other without a person in the loop. Learn more →
- Customer record
- The screen of customer data an agent sees during a conversation. Learn more →
- Personal data
- Anything that points to a specific person: a name, a number, a voice. Learn more →
- Agent assist
- Suggests answers from the knowledge base to the agent during a conversation. Learn more →
- Speech-to-text
- Turns spoken conversation into text. Learn more →
Updated: