Back to blog
Call quality assurance
Most contact centres still review a handful of calls per agent each month. These guides cover what changes when every call is scored: the scorecard, calibration, speech recognition in Norwegian and Swedish, and how to compare the tools.
See how Conversational Intelligence scores every callPosts on this theme
25 posts
- The Capturi alternative that doesn't make you buy a contact-centre platformCapturi is now a module in Puzzel's platform. Future Ready scores every call on the telephony you already run, shows the evidence, and coaches the gaps.Conversational Intelligence
- A Swedish court banned one sentence from a chatbot. Your AI agent's answers count as yoursA Swedish court banned one answer from Hallon's chatbot. A German court said correct data is no defence. What it means for AI agents, and 6 questions to test.Quality assurance
- Sweden's new credit law starts 20 November: five duties that land on your agents' callsFrom 20 November, Swedish lenders must spot payment trouble early and, where appropriate, offer alternatives to enforcement. The five duties and a checklist.Quality assurance
- A Danish insurer scored recorded calls with AI. The regulator let the scoring stand and struck the consentDenmark's regulator treated an insurer's AI scoring of recorded calls as quality assurance, but found its one-key consent invalid. What it means for you.Quality assurance
- An agent who fails a check on 3 calls in 10 gets a clean five-call review one month in sixThe most common QA sample is 3 to 5 calls per agent a month: enough to coach from, too few to rank anyone. How many calls each question needs, with the maths.Quality assurance
- 61% of customers prefer a human. 59% prefer AI. Both are right.Customers turn on AI customer service when it stands between them and a person. What 2026's surveys agree on, and three numbers to check on your own calls.Conversational Intelligence
- Five questions a bank or insurer's compliance team will ask about AI call scoringAI that reads agents' feelings from their voice is already banned at work. Five questions compliance will ask, and what a vendor should put in writing.Quality assurance
- Before a BPO puts AI on a client's calls, the client's DPO gets a sayUnder GDPR Article 28, your AI scoring vendor is your client's sub-processor. What to send their DPO, whose scorecard counts, and what the client sees.Quality assurance
- Sweden may ban telemarketing. France and the Netherlands already require opt-inSweden's inquiry reports by 15 January 2027. Under opt-in, every call needs proof of a yes: what that record must show, and five steps to take now.Quality assurance
- Four team leads score one call 2, 3, 3 and 5. Measure that before you judge the AIHow to calibrate AI call scoring against human QA. Two reviewers can match on 94% of calls and agree on just 2 of the 8 they failed. Measure that first.Quality assurance
- Your next renewal has an AI vendor's quote in it. How to keep the contractAI vendors now bid for your simplest queues. How an outsourced contact centre proves quality on its client's own rubric, with gains a finance team believes.Quality assurance
- Add AI call scoring to Genesys without touching routing, numbers or desktopsA quality layer scores the recordings Genesys already keeps, on your rubric. Why not your telephony's own QA module, and why your QA history survives a switch.Conversational Intelligence
- Make the QA vendor prove it on one behaviour before you buyCoach the one behaviour that's costing you on calls you already record, compare with teammates who weren't coached, and take finance a number it believes.Quality assurance
- New hires improve without coaching. How to measure what coaching addedCompare coached new hires with uncoached peers from the same intake and subtract. A worked example shows how much a before-and-after slide overstates.Quality assurance
- If nobody disputes an AI score in the first weeks, worryThe person who resists AI scoring hardest is often your QA lead. The rollout order that keeps trust: QA first, disputable scores, coaching from their own calls.Quality assurance
- Handle time rose 4% last month. The reason is in calls nobody readYour dashboard shows that handle time moved. Reading every call shows why: complaint drivers as they form, avoidable contact, and demand you can't meet.Conversational Intelligence
- Containment counts a customer who gave up as a win. Score AI like your peopleYour people are scored on quality and your AI on containment, so the riskiest queue looks safest to automate. One rubric for both, handover included.Quality assurance
- A rep's pricing score rose after coaching. How much did the coaching earn?A rep's pricing score rises 1.2 points after coaching; untrained peers rise 0.3. How to measure the part coaching caused, and four checks before you report it.Conversational Intelligence
- Your weakest agents' scores will rise even if coaching does nothingA 9-point rise, a 6-point rise in uncoached peers, and the 3 points coaching caused. The four-step method that gives you a number your CFO will accept.Quality assurance
- The AI Act's delay skipped your contact centre: AI agents must disclose nowHigh-risk duties moved to December 2027. Article 50 didn't: since 2 August 2026, chat and voice agents must say they're AI. What counts, and what doesn't.Quality assurance
- Your AI agent's wrong answers show up as successes on your dashboardHandle time, CSAT and containment can log a wrong answer as a save. How to score AI agents on your own rubric, and four questions to ask any vendor.Quality assurance
- Most agents learn to handle vulnerable customers by practising on themA new starter's first bereavement call is often a real bereaved customer. How to find the calls agents fumble, turn them into voice role-play, and check it.Quality assurance
- Your board's outcomes report rests on a few calls. How to evidence every callThe FCA found board reports full of data that didn't show outcomes. Understanding and support happen on the call: how to evidence them on every one.Quality assurance
- If two reviewers score 'empathy' 2 and 4, your AI can't score it eitherWhat a good AI QA scorecard looks like: every vague criterion rewritten as a behaviour you can quote from the transcript. With a before-and-after table.Quality assurance
- Closed-loop QA has three steps, and many loops stop after twoClosed-loop quality management scores every call, coaches each gap it finds and re-measures it on later calls. Why most contact centres stop after step two.Quality assurance
No posts match. Try another word or topic.
See it on your own calls.
Book a 30-minute call. We’ll score a sample of your own calls and show you where the gaps are, and what closing them is worth.















































