Call Centre Quality Monitoring: Why Sampling Isn't Enough
Quality assurance is one of the most compliance-critical functions in any contact centre, and one of the most under-resourced. For most operations, the gap between what QA teams can review and what regulators now expect to see evidenced has never been wider.
Most contact centres review a small fraction of their calls. A QA analyst picks a handful, scores them, flags what went wrong, and then moves on. It feels like it ticks the box for quality assurance. But for Ofcom-regulated telecoms operations and FCA-regulated financial services firms, it’s not enough, and the consequences of getting it wrong have never been higher.
This article explains why call sampling creates compliance exposure, what always-on monitoring looks like in practice, and what to look for when evaluating your current approach.
What is call quality monitoring?
Call quality monitoring is the process of reviewing agent-customer interactions to assess whether they meet your quality, compliance, and performance standards.
It typically covers:
What was said and how the agent handled the conversation
Whether compliance scripts and protocols were followed
How vulnerable customers were identified and managed
Whether the outcome was appropriate for the customer
How performance compares against your scoring framework
When call quality monitoring is done consistently, it gives you a documented evidence-base across every call type, agent, and campaign. But when it’s done poorly or too infrequently, it leaves gaps that regulators are increasingly likely to find before you do.
How do most contact centres currently monitor calls?
Sampling is the typical approach many call centres take to monitoring calls. A QA reviewer listens to a set number of calls per agent per month, scores them against a framework, and feeds the results back into coaching. It is time-consuming work, so let’s break down the numbers.
Example:
A single reviewer handles 50 calls a month at 30 minutes per call.
This amounts to 25 hours of review time.
And it is still only a fraction of the total call volume reviewed.
The problem here is not the effort; it's the coverage. On average, contact centres manually evaluate 5% of calls per week, meaning many QA operations are leaving the majority of interactions unreviewed. This means:
You don’t know whether your agents are consistently identifying vulnerable customers.
You don’t know whether compliance scripts are being followed on the calls you did not pick.
You are not building an evidence bse, only a small sample.
Manual call sampling statistics
FCA Consumer Duty: you need evidence across every interaction, not a snapshot
For debt collection, insurance, and other FCA-regulated contact centres, the stakes are different but the problem is the same. Consumer Duty requires firms to demonstrate they are delivering good outcomes for retail customers, not just on the calls they reviewed, but consistently and measurably across their entire operation.
The FCA has shifted decisively from implementation to enforcement. Regulators are no longer asking whether you have a quality monitoring process. They are asking whether you can prove, with documented evidence, that your agents are handling vulnerable customers correctly, following compliant scripts, and not causing foreseeable consumer harm. And that’s for every call, not just the ones you checked.
A sampling approach does not produce that evidence. It produces a snapshot.
For more on what the FCA now expects from contact centres in financial services and debt collection, see our Consumer Duty guide.
The problem with call sampling: A Summary
Sampling typically covers around 5% of calls per week, leaving the 95% of interactions unreviewed and unverifiable
Compliance drift happens slowly. By the time sampling catches a behaviour, it is already established and harder to coach out
Poor agent behaviour on outbound calls can go undetected long enough to trigger carrier blocking or an FCA flag
Vulnerable customers may not be identified correctly on calls you never reviewed
Good performance goes unrecognised as you cannot replicate what you cannot see
A sample tells you what happened on the calls you chose to review. It does not tell you what is happening in your operation
From sampling to monitoring: what's actually required
Moving from sampling to consistent call monitoring is not simply a matter of reviewing more calls. It requires the right infrastructure in place, and historically, that infrastructure was either too expensive, too time-consuming, or both.
At a minimum, always-on monitoring requires:
Call recording across all interactions, not just selected campaigns or call types
Transcription that converts voice to text accurately enough to be reviewed and searched at scale
A platform that connects recording, transcription, scoring, and reporting in one place rather than across multiple disconnected tools
Without all three, monitoring at scale either falls back on human reviewers (which is where the 25-hours-per-50-calls problem comes back in) or produces data too inconsistent to be useful as a compliance evidence base.
MaxContact's Conversation Analytics brings all of this together in a single platform. Call recording, real-time transcription, and reporting sit alongside each other. This gives your QA team a single place to monitor, review, and evidence what is happening across every interaction, without stitching together multiple tools or managing separate systems.
The reason most contact centres have defaulted to sampling is not because they did not want better coverage. It is because the operational cost of achieving it manually was prohibitive. A team large enough to review every call would cost more than most mid-market operations can justify. But that has changed.
How Conversation Analytics makes always-on monitoring feasible
Conversation Analytics is the platform that makes consistent, always-on call monitoring operationally viable for mid-market contact centres.
Rather than relying on a QA team to manually select, listen to, and score calls, Conversation Analytics connects call recording, transcription, scoring, and reporting in a single platform – automating quality assurance. Every interaction is captured, transcribed, and made reviewable, giving your QA team complete visibility across all call types, all agents, and all campaigns without the resourcing overhead of manual review at scale.
The cost comparison is significant. Replicating meaningful call coverage with human reviewers alone would cost an estimated £14,000 per month in analyst time for a mid-sized contact centre. Conversation Analytics delivers that coverage at a fraction of the cost, freeing your QA team to focus on coaching, calibration, and the complex calls that genuinely need a human eye.
How AI call monitoring surfaces insights faster
AI is what makes the insights from always-on call monitoring actionable rather than overwhelming.
Without AI, full call coverage creates a different problem; more data than a QA team can meaningfully review and act on. AI-powered call monitoring solves that by doing the heavy lifting on routine scoring, so your team's attention goes where it matters most.
Benefit
What it means for your operation
Structured scorecards answered automatically
Every scorecard question is answered using transcript evidence; no manual listening, no reviewer subjectivity.
Transcript-linked evidence
Every score links back to the exact exchange that informed it, giving you a defensible audit trail.
Faster review cycles
Review time drops from 30 minutes to 5 minutes per call, recovering around 4 days of analyst time every month.
Consistent scoring across your entire operation
The same criteria, applied the same way, across every agent, call type, and campaign every time.
Human oversight built in
Your QA team reviews outputs, calibrates scoring, and focuses on complex calls. AI handles the routine. Governance stays with your team.
The result is not just faster QA. It is a more reliable, more defensible evidence base built on every call, not a sample of them.
What to look for in your call quality monitoring approach
Is your evidence transcript-linked? Generic summaries are not a defensible evidence base. Scoring decisions need to be traceable back to what was actually said.
Is your scoring consistent? If different reviewers score the same call differently, your evidence has a credibility problem. Consistent scoring logic applied across all interactions removes that subjectivity.
Does your QA sit within your analytics platform? If scoring, feedback, and reporting live in separate tools, you create friction and risk. Everything should be in the same place.
Is human oversight built in? Your QA team should be able to review, challenge, and calibrate outputs. Always-on monitoring supports human-led governance, it does not replace it.
Are you scoring the right calls? Configurable triggers and criteria by call type, queue, campaign, or outcome, mean your monitoring effort goes where the compliance risk is highest.
The question is not whether you can afford to monitor every call
It is whether you can afford not to.
Ofcom and the FCA have both made clear that evidence of compliance needs to be consistent, documented, and demonstrable. A sampling process may satisfy an internal audit. It is unlikely to satisfy a regulator asking for proof of good outcomes across your entire customer base.
Always-on call quality monitoring closes that gap. It gives your QA team better data, gives your compliance function defensible evidence, and gives your operation a consistent view of what is actually happening on the phones across all calls, rather than just the ones you happened to pick.
Download the UK Contact Centre Regulatory Guide 2025–2027 to see how the FCA and Ofcom compliance obligations facing your sector map to your call monitoring approach and what good evidenced practice looks like in both.
(() => {
const run = () => {
const rich = document.querySelector('#rich-text');
const toc = document.querySelector('#toc');
if (!rich || !toc) return;
const headings = rich.querySelectorAll('h2');
if (!headings.length) {
toc.style.display = 'none';
return;
}
const slugCounts = Object.create(null);
const slugify = (str) => {
const base = (str || '')
.trim()
.toLowerCase()
.normalize('NFD').replace(/[\u0300-\u036f]/g, '')
.replace(/[^a-z0-9\s-]/g, '')
.replace(/\s+/g, '-')
.replace(/-+/g, '-');
const n = (slugCounts[base] = (slugCounts[base] || 0) + 1);
return base ? (n > 1 ? `${base}-${n}` : base) : `section-${n}`;
};
// Build TOC off-DOM
const frag = document.createDocumentFragment();
for (let i = 0; i < headings.length; i++) {
const h = headings[i];
const text = (h.textContent || '').trim() || `Section ${i + 1}`;
if (!h.id) h.id = slugify(text);
const a = document.createElement('a');
a.href = `#${h.id}`;
a.className = 'content_link is-secondary';
a.dataset.target = h.id;
a.setAttribute('aria-label', text);
const p = document.createElement('p');
p.className = 'text-size-small';
p.textContent = text;
a.appendChild(p);
frag.appendChild(a);
}
// Single DOM update
toc.innerHTML = '';
toc.appendChild(frag);
toc.addEventListener('click', (e) => {
const link = e.target.closest('a.content_link[href^="#"]');
if (!link) return;
e.preventDefault();
const id = link.getAttribute('href').slice(1);
const target = document.getElementById(id);
if (!target) return;
// Only compute layout once
const targetTop = target.getBoundingClientRect().top + window.scrollY;
const finalY = targetTop - 150;
window.scrollTo({ top: finalY, behavior: 'smooth' });
history.replaceState(null, '', `#${id}`);
}, { passive: false });
};
// Webflow-safe “run after everything is ready”
if (window.Webflow && Webflow.push) {
Webflow.push(() => requestAnimationFrame(run));
} else {
document.addEventListener('DOMContentLoaded', () => requestAnimationFrame(run));
}
})();
related articles
you might also like
Our articles and industry insights give you expert perspectives, practical strategies, and the latest trends to help your business connect smarter and perform better.
(() => {
const WORDS_PER_MINUTE = 200;
const MULTIPLIER = 1; // your choice
const estimateMinutes = (el) => {
if (!el) return null;
const text = (el.innerText || el.textContent || "").trim();
if (!text) return 1;
const words = (text.match(/\S+/g) || []).length;
const baseMinutes = Math.max(1, Math.ceil(words / WORDS_PER_MINUTE));
return Math.max(1, Math.ceil(baseMinutes * MULTIPLIER));
};
const findNearestTargetInItem = (itemRoot, rt) => {
if (!itemRoot) return null;
return itemRoot.querySelector('.is-text');
};
const applyWithin = (root) => {
// More forgiving selector: attribute present or equals "true"
root.querySelectorAll('[data-rich-text], [data-rich-text="true"]').forEach((rt) => {
const itemRoot =
rt.closest('[role="listitem"]') ||
rt.closest('.w-dyn-item') ||
rt.parentElement ||
root;
const target = findNearestTargetInItem(itemRoot, rt);
if (!target) return;
const mins = estimateMinutes(rt);
if (mins != null) target.textContent = `${mins} MIN READ`;
});
};
const init = () => {
applyWithin(document);
// Re-apply on dynamic changes (pagination/filters)
const mo = new MutationObserver((mutations) => {
for (const m of mutations) {
for (const n of m.addedNodes) {
if (!(n instanceof Element)) continue;
if (
n.matches('[data-rich-text], [data-rich-text="true"], [role="list"], .w-dyn-items, .w-dyn-item') ||
n.querySelector?.('[data-rich-text], [data-rich-text="true"]')
) {
applyWithin(n);
}
}
}
});
mo.observe(document.body, { childList: true, subtree: true });
};
// Robust bootstrapping
if (window.Webflow && Array.isArray(window.Webflow)) {
window.Webflow.push(init);
} else if (document.readyState === 'loading') {
document.addEventListener('DOMContentLoaded', init, { once: true });
} else {
// DOM is already ready; run now
init();
}
})();
When you get it right, you build trust, loyalty, and efficiency. If you get it wrong, you risk frustrating customers, increasing churn, and even regulatory trouble.
Long waiting times and poor call handling don’t just test patience; they directly impact outcomes. Customers who are satisfied with wait times are 2.6x more likely to repurchase, and 3x more likely to recommend a brand (Qualtrics 2025).
So, if you’re a contact centre leader aiming to reduce customer churn and strengthen reputation, then optimising call handling operations is essential. Drawing on insight from our ,, here are the five most common mistakes that hurt customer experience and compliance and how to avoid them.
1. Long average handling times without a call resolution
55% of customers abandon calls because of long waits. It’s an all too familiar frustration, but it’s the way contact centres tackle it that makes all the difference.
The knee-jerk reaction is to push agents to “be faster,” but it often backfires. When speed becomes the only metric, compliance takes a back seat, important details are missed, and customers end up calling back because their issue wasn’t fully resolved the first time around. That adds more volume, along with more cost and frustration.
The more effective approach is to focus on resolving calls the first time. That involves giving agents the tools and context to handle conversations confidently and consistently:
Skills-based call routing means customers reach the right person the first time. While routine queries go to newer agents, complex or sensitive issues are routed straight to specialists or agents with more experience.
Call recording and reporting tools can uncover repeated bottlenecks and delays, helping managers coach more effectively.
Agent assist features, such as call scripts and prompts help to keep conversations on track without rushing, reducing compliance risks.
By fixing the causes of long average handling time, such as poor routing, lack of context, or inefficient workflows, contact centres can cut waiting times and keep customers from hanging up.
For deeper strategies on reducing average handling time (AHT) while preserving that all-important customer experience, see our guide on How to Reduce Average Handle Time.
2. Poor call routing
Yes, being left on hold is frustrating, but being bounced between agents and departments is worse. In fact, 34% of customers abandon calls after being transferred multiple times.
Every unnecessary transfer forces customers to repeat themselves, delays resolution, and erodes trust in your service. It’s inefficient and leaves people questioning whether your business is competent at all.
Intelligent call routing solves this by matching enquiries to the right place first time:
Skills-based routing directs complex or sensitive cases to the most experienced agents.
IVR menus streamline routine queries and cut unnecessary transfers.
CRM integration gives agents context immediately, so customers aren’t asked to repeat themselves.
With fewer transfers and faster resolutions, customers stay on the line and crucially, on your side.
3. Lack of agent training on compliance protocols
When customers call with complaints or sensitive issues, they want to feel heard and reassured they’re in safe hands.
But if an agent sounds unsure or handles a situation incorrectly, trust is lost. Our Voice of the Consumer Report shows this has real consequences: 35% of customers have abandoned a call because the agent couldn’t understand their situation.
So, what’s the solution? Invest in ongoing coaching and visibility:
Drag-and-drop scripting tools guide agents through tricky or regulated conversations step by step.
Speech analytics and sentiment analysis flag risks early and speed up QA reviews.
Targeted coaching and live monitoring allow managers to step in before issues escalate.
Well-trained agents aren’t just more compliant. They’re more confident, faster, and deliver better CX.
4. Not capturing or acting on customer feedback
When service quality falls short of customer expectations, they rarely hold back: 66% leave reviews after a negative experience. One of the biggest mistakes many contact centres make is failing to capture or act on their feedback, thus missing opportunities to fix the most common issues that lead to customer complaints in the first place.
With the right technology in place, contact centres can take customer feedback (one of the richest sources of improvement) and turn it into action.
Here’s how:
Speech analytics records 100% of calls, and with speech-to-text transcripts, managers can review real conversations. By spot-checking and searching for recurring issues, like complaints about wait times or poor transfers, managers can act on weaknesses in call-handling processes.
CRM integrations can also be utilised to link feedback with customer history, highlighting any customer retention risks or gaps in agent onboarding.
Post-call surveys and IVR feedback can capture immediate customer sentiment and reactions, feeding CX metrics such as CSAT and NPS.
When feedback loops are closed, customers feel heard, agents improve faster, and compliance risks are caught before they escalate. This is only possible with a well-defined, data-driven contact strategy guiding your approach.
5. Over-reliance on AI or automation
70% of customers prefer to speak to a human for complex situations, and 20% say AI doesn’t understand them. The takeaway from those statistics is that automation is powerful. But only if it supports rather than replaces live agents.
Used well, automation makes operations more efficient:
IVR menus and AI agents handle simple, repetitive tasks like account checks or payment reminders instantly.
Seamless escalation to live agents ensures sensitive or complex queries are handled with empathy and context.
This hybrid approach keeps costs and waiting times down, but also reassures customers that when they need to speak to a human, someone is available.
These five mistakes; long handling times, poor routing, inadequate training, ignoring feedback, and over-reliance on automation do more than just hurt customer experience. They also increase compliance risks due to inconsistent handling and breaching customer trust.
Used effectively, software and automation can support your people, not replace them. By combining the right tools and processes, contact centres can deliver a service that’s faster, more compliant, and more human, even under cost pressure.
See how MaxContact helps you meet customer expectations and improve compliance. Book a demo today.
Blog
5 min read
WHAT OFCOM AND PECR REGULATIONS REALLY MEAN: A QUICK GUIDE
The rules governing how UK contact centres dial, who they can call, and what it costs to get it wrong have changed significantly. This guide cuts through the complexity so you know exactly where you stand.
(() => {
const WORDS_PER_MINUTE = 200;
const MULTIPLIER = 1; // your choice
const estimateMinutes = (el) => {
if (!el) return null;
const text = (el.innerText || el.textContent || "").trim();
if (!text) return 1;
const words = (text.match(/\S+/g) || []).length;
const baseMinutes = Math.max(1, Math.ceil(words / WORDS_PER_MINUTE));
return Math.max(1, Math.ceil(baseMinutes * MULTIPLIER));
};
const findNearestTargetInItem = (itemRoot, rt) => {
if (!itemRoot) return null;
return itemRoot.querySelector('.is-text');
};
const applyWithin = (root) => {
// More forgiving selector: attribute present or equals "true"
root.querySelectorAll('[data-rich-text], [data-rich-text="true"]').forEach((rt) => {
const itemRoot =
rt.closest('[role="listitem"]') ||
rt.closest('.w-dyn-item') ||
rt.parentElement ||
root;
const target = findNearestTargetInItem(itemRoot, rt);
if (!target) return;
const mins = estimateMinutes(rt);
if (mins != null) target.textContent = `${mins} MIN READ`;
});
};
const init = () => {
applyWithin(document);
// Re-apply on dynamic changes (pagination/filters)
const mo = new MutationObserver((mutations) => {
for (const m of mutations) {
for (const n of m.addedNodes) {
if (!(n instanceof Element)) continue;
if (
n.matches('[data-rich-text], [data-rich-text="true"], [role="list"], .w-dyn-items, .w-dyn-item') ||
n.querySelector?.('[data-rich-text], [data-rich-text="true"]')
) {
applyWithin(n);
}
}
}
});
mo.observe(document.body, { childList: true, subtree: true });
};
// Robust bootstrapping
if (window.Webflow && Array.isArray(window.Webflow)) {
window.Webflow.push(init);
} else if (document.readyState === 'loading') {
document.addEventListener('DOMContentLoaded', init, { once: true });
} else {
// DOM is already ready; run now
init();
}
})();
Industry regulator Ofcom produces regulations around what contact centres can and can't do when it comes to automatic dialling and CLI. On top of that sits the Privacy and Electronic Communications Regulations (PECR), enforced by the ICO. This covers consent, suppression against the TPS and CTPS registers, and fines for getting marketing calls wrong.
The regulations can be confusing, having undergone numerous iterations over many years. There is no single source of truth - there isn't even a complete version of all the regulations in one place. And as of 2026, the landscape has shifted again.
With that in mind, we've produced a no-nonsense guide to the current regulations surrounding the use of automatic calling equipment (ACS), updated to reflect the Data (Use and Access) Act 2025 (DUAA) changes now in force, the updated Ofcom CLI guidance applied from 29 January 2025, and the complaints provisions due to take effect from 19 June 2026.
It's important that you know these regulations, because neither Ofcom nor the ICO accept ignorance as a defence. And the penalties have changed significantly. The maximum PECR fine (which used to be capped at £500,000) is now £17.5 million or 4% of global annual turnover, whichever is higher. That's a 35-fold increase. Directors can also be held personally liable for serious breaches.
For reference:
ACS = Dialler software
Abandoned call = dropped call
Connect = Live call passed to an agent
The guide is split into two parts: the Ofcom rules that govern how you dial, and the ICO/PECR rules that govern who you can call and what it costs if you get it wrong.
Part 1: Ofcom rules for automated dialling
Silent calls
There are three main causes of silent calls:
False positives
Abandoned calls (with no message)
Agents receiving calls and not speaking or being on mute, which is a reasonably significant cause.
In 2010 Ofcom said companies “should ensure their abandoned call rate is less than 3% of all live calls.”
That seemed clear enough. If dropped (or abandoned) calls were less than 3% of your total, you were safe. Or so many of us thought.
But as it turned out, this was a misinterpretation. Ofcom has since clarified that 3% was meant as a measuring stick and the actual target is 0%. In other words, you can attract regulator scrutiny with any percentage of dropped calls.
Despite Ofcom’s clarification, many people in the industry still think the 3% rule applies – and some even run their diallers with 3% in mind. We have even seen suppliers of dialler equipment advertise the 3% rule as compliant.
It isn’t. This is why we developed our un-droppable algorithm, which helps contact centres stick to 0% dropped calls even in blended environments.
Knowing how to calculate the drop call rate accurately is essential if you want to remain compliant, and there are still quite a few systems that get it wrong. They either use the wrong formula or, knowingly or otherwise, add inbound and manual dial connects into the equation to artificially reduce the drop call rate.
The correct formula is:
Drop rate = Drops/(Drops + Connects)
It’s worth checking Ofcom’s guidance (see Section A3.8) on calculating the drop call rate. It has examples with AMD both on and off.
15 second minimum ring time
This rule stipulates that you have to call a number for a minimum of 15s before disconnecting as a no answer.
The rule was introduced to prevent the practice of “pinging”, whereby diallers would ring a number for 1 second to try and generate a missed call. The next day agents would dial all the numbers found to be “live”. You can imagine how annoying this was for customers.
The only contentious point around the 15s minimum ring revolves around preview calls, which are automated but in every other way are the same as a manual call. Manual calls are not restricted to the 15s minimum. Because of this, whether or not you need to adhere to the minimum ring time rule on preview calls is open to interpretation. The Ofcom wording is in Section A2.15.
Rotating CLI
Not long ago, there was a trend for calling line identification (CLI) calls to display local numbers. The system would present a Manchester number when calling a Manchester number, a Newcastle number when calling a Newcastle number, and so on.
This was primarily done to increase connect rates, and it worked pretty well. It also rotated the numbers, so people didn’t recognise you.
But because Ofcom and the ICO base complaints on the CLI, it was seen as a way of avoiding investigation because contact centres were presenting 150+ different numbers. So complaints to a single number were low, even when complaints about the call centre as a whole might have been very high.
Ofcom acted, and carriers are no longer allowed to automatically rotate numbers, so the practice is vastly reduced. But you can still present different numbers if you have a valid reason. For example, a well-established car dealership with branches around the UK, but a central call centre, can present a local dealer’s number.
There’s a lot of controversy around AMD use. Ofcom now makes firms take false positive rates into consideration when calculating drop rates (though it has suggested it may drop this).
This is because inadequate systems with poor AMD were causing silent calls on false positives. In other words, when the system thought a live person was an answering machine, no message played and the call was not passed to an agent (see below on leaving a message).
But AMD use is not banned, and it can play an essential role in the productivity of contact centres. In fact, it can increase efficiency by 100% in some cases, but it must be used correctly.
CPA stands for Call Progress Analysis and is part of the AMD function. Two seconds is the amount of time your CPA is given to determine whether the called party is an answering machine or a live person.
Many providers say two seconds is not enough and, as a result, AMD can’t be used compliantly. They’re wrong. In our case, if we can’t determine whether the answered call is an answering machine or not, we pass the call through to an agent and play it safe.
If your current system “holds” onto the call longer than two seconds you are non-compliant, which is why you often get recommended to turn AMD off rather than the vendor resolving the problem.
In a silent call, someone answers the phone only to be met with silence from the other end. Needless to say, Ofcom is not in favour of silent calls.
It’s recommended that an answer phone message is left, so false positives would no longer result in silent calls. The problem here is the huge proliferation of answer phone messages, but Ofcom’s number one priority is to tackle silent calls. It all comes down to the contact centre: you have to make the choice.
When dialling predictively you are calling more people than you have agents, which can occasionally result in dropped calls. Instead of silent calls, Ofcom insists a drop call message is played to the consumer. It also offers guidelines on what the message should and shouldn’t say.
In the context of AMD, a false positive occurs when the system detects an answer machine even though a person has answered the phone. This results in silent calls and complaints from consumers.
The recommendation from Ofcom here is to play an answer machine message. Consumers will hang up, but at least they won’t experience a silent call that can be at best annoying and at worst frightening.
Carrier-level call blocking has moved from an emerging measure to an established part of Ofcom's approach to nuisance and silent calls, and it has intensified. What it means is that termination endpoint carriers such as EE, Vodafone, BT and O2 (Telefónica), can block you if they think you are making nuisance calls to their customers. This isn't the carrier you are using to make calls, it's the network the called party is on.
Carriers share a database of nuisance callers. If one blocks you, the others may follow.
This is an automated response to perceived nuisance calling, based on two main factors:
Connect rate or ASR (Answer Seizure Ratio). If your connect rate to Vodafone (for example) is low (i.e. people aren't answering the phone) then Vodafone will interpret this as people not wanting to talk to you.
ACD (Average call duration). Taking Vodafone again as an example, if call duration length is low Vodafone might view this as customers not wanting to speak to you. "Low" typically means less than 30 seconds. Leaving an answer machine message can help, as it increases the average call duration.
These factors will often result in an automatic calling block to Vodafone numbers from your CLI.
The only way to get around this is to change your number. You should also make a complaint to the termination carrier if you think they have blocked you by mistake, but be aware there is no formal appeals process, so prevention through disciplined call pacing, accurate targeting and message quality is a far better strategy than remediation.
What changed in January 2025
Ofcom's updated CLI guidance came into effect on 29 January 2025, closing a long-standing loophole that allowed scammers abroad to spoof UK numbers. Under the updated guidance, calls entering the UK from overseas that present a UK number as a Presentation Number must be blocked, with a limited number of legitimate exceptions (for example, UK mobile users genuinely roaming overseas).
For legitimate UK contact centres, the direct impact is small, but the indirect impact matters. Consumer trust in unknown numbers is at a historic low, carriers are more willing than ever to block, and the bar for what looks like "legitimate" dialling behaviour to a carrier algorithm keeps rising.
There is also a live Ofcom consultation (published July 2025) on extending blocking to calls from abroad that spoof UK mobile (+447) numbers. This is likely to land in 2026 and will tighten things further.
Unfortunately, as Ofcom makes clear, this blocking policy is actively encouraged:
Page 6 – Technical Measures (Ofcom forms Strategic Working Group)
Page 5 – Strategic Working Group (Ofcom confirms the SWG has blocked a large number of nuisance calls since its formation)
You are not allowed to call previous dropped calls within 72 hours, or answer machines the same day (note: not 24 hours), unless you can guarantee an operator is present. Preview or manual dialling is fine. This is one of the reasons MaxContact has mixed-mode dialling, which means we can automatically move records from “predictive” to “preview” and back again.
What exactly “harassment” means is not clearly defined and is open to interpretation. The term “a reasonable amount” is often used, but it’s unclear what that means in practice.
This needs Ofcom clarification. It can boil down to the nature of the calls. Debt collectors have a valid reason for calling but how much is too much? We have seen agreements in place between the FCA and some contact centres that allow X calls per day but no more than Y times per week, for example.
Maximum attempts to a user
Again, this is not defined in the regulations and needs clarification. Common sense should be applied, and the strategy modified in line with the nature of the call.
DNC list
Do you need an internal one? It's not in the Ofcom regulations, but it's good practice and operationally sensible. There does need to be a way for people to remove themselves from your calling lists, either by speaking to an agent or more commonly through self-service IVRs.
It's also worth being clear that the Telephone Preference Service (TPS) is the UK's official Do Not Call register for landlines and mobile numbers. If a number is registered with TPS, you are legally required under PECR to refrain from calling it for unsolicited sales and marketing purposes and this is not discretionary. Your internal DNC list and TPS suppression are separate obligations and both need to be in place. See Part 2 for the full PECR and TPS compliance picture.
The mention of DNC lists in the Ofcom regulations is in Section A1.25 here.
Part 2: ICO and PECR consent, suppression and fines
This is where things have changed the most. The rules on who you can call, and what evidence of consent you need, have been in place for years. But the financial consequences of getting them wrong are no longer what they were.
PECR fines explained
The Data (Use and Access) Act 2025 (DUAA) received Royal Assent on 19 June 2025, with key enforcement provisions coming into force on 5 February 2026. The headline change is the fine cap. Previously, the maximum PECR fine was £500,000. Under the DUAA, PECR fines are now aligned with UK GDPR: a maximum of £17.5 million or 4% of annual global turnover, whichever is higher. That's a 35-fold increase for calling TPS-registered numbers without consent, making automated marketing calls without consent, or failing to produce evidence of consent when the ICO asks.
Director personal liability also applies. Since 2018, the ICO has been able to issue personal fines directly to company directors for PECR breaches specifically to stop directors escaping penalties by dissolving firms and restarting under a new name. This is a board-level risk, not an operations one.
In September 2025, the ICO fined Home Improvement Marketing Ltd for instigating more than 2.4 million automated marketing calls without prior consent. The pattern is consistent: the ICO is fining fewer organisations but at higher individual values, and contact centres are over-represented. Making thousands of calls a day across multiple campaigns multiplies the opportunity for a TPS-registered number or a broken consent chain to slip through. Getting this right is now a commercial risk question, not a technical configuration one.
TPS compliance
If you are making unsolicited sales or marketing calls to UK consumers, you are required by law under PECR to screen against the Telephone Preference Service (TPS) the UK's central opt-out register for individuals, sole traders and partnerships who don't want to receive marketing calls. There is no threshold, no exemption for small volumes, and no grace period.
A few points that catch contact centres out in practice:
TPS applies at the point of the call, not the point of data purchase. "The data was TPS-screened when we bought it" is not compliant.
Consent overrides TPS. If someone has given valid, specific, evidenced consent to receive calls from you, you can call them even if they're on TPS. The burden of proof sits with you.
Screening means before the call, not after. You can't call to check whether someone is happy to be called - that call is the breach.
The ICO standard is that marketing lists must be screened at least every 28 days if the list is still being called. That's 28 days (four weeks) not "monthly". If you're still calling the same list, that's 13 screenings a year, not 12. Real-time screening via an API into the dialler is the strongest position to be in.
If your third-party data supplier says it screens against TPS, you're still responsible for confirming that it does. Spot-check the data, ask for screening logs, and ask what consent statement was presented at the point of data collection. If the answer is vague, you share responsibility for any breach.
CTPS is the business-to-business equivalent of TPS. The Corporate Telephone Preference Service covers corporate subscribers. For example, limited companies, PLCs, LLPs, Scottish partnerships, government bodies, schools, colleges and charities, who have registered their objection to receiving unsolicited marketing calls.
If you're running B2B outbound, CTPS compliance is a legal requirement. The same PECR enforcement regime applies, which means the same £17.5m or 4% of turnover cap.
Two failure modes trip B2B contact centres up repeatedly. First, assuming B2B is exempt. It isn't. Second, screening against TPS but not CTPS; a B2B list screened only against TPS will still contain CTPS-registered numbers. If you're calling businesses, you need both.
Automated decision-making and AI in your dialler
This is a genuinely new area, and one that will matter more through 2026.
The DUAA makes it easier to deploy AI-driven processes in contact centres (automated lead scoring, AI-driven call routing, AI-assisted collections prioritisation, AI quality scoring) but introduces mandatory safeguards. If you're using AI to decide who gets called, in what order, or with what script, you now have specific obligations. Customers must be informed of significant automated decisions that affect them, must be able to challenge those decisions, and must be able to get a human to look at them. From 19 June 2026, they'll also have a new right to complain directly to you as the data controller.
In practice:
Map where AI is making decisions in your contact strategy
Document your lawful basis for each
Make sure there's a human-override route your agents and team leaders know how to action.
ICO guidance on ADM under the DUAA is expected through 2026.
MaxContact is your friend here. Our sophisticated customer engagement solution will keep dropped calls to an absolute minimum (in fact, close to zero), while helping you remain compliant.
The regulatory landscape for UK contact centres has shifted significantly, and it will keep shifting.
Download the UK Contact Centre Regulatory Guide 2025–2027 for a full breakdown of what Ofcom, the ICO, and the FCA now expect from your operation, and what the consequences look like if you get it wrong.
Blog
5 min read
How to Improve Quality Assurance in a Call Centre
Do your QA processes generate a lot of data but don’t drive change on your contact centre floor?
(() => {
const WORDS_PER_MINUTE = 200;
const MULTIPLIER = 1; // your choice
const estimateMinutes = (el) => {
if (!el) return null;
const text = (el.innerText || el.textContent || "").trim();
if (!text) return 1;
const words = (text.match(/\S+/g) || []).length;
const baseMinutes = Math.max(1, Math.ceil(words / WORDS_PER_MINUTE));
return Math.max(1, Math.ceil(baseMinutes * MULTIPLIER));
};
const findNearestTargetInItem = (itemRoot, rt) => {
if (!itemRoot) return null;
return itemRoot.querySelector('.is-text');
};
const applyWithin = (root) => {
// More forgiving selector: attribute present or equals "true"
root.querySelectorAll('[data-rich-text], [data-rich-text="true"]').forEach((rt) => {
const itemRoot =
rt.closest('[role="listitem"]') ||
rt.closest('.w-dyn-item') ||
rt.parentElement ||
root;
const target = findNearestTargetInItem(itemRoot, rt);
if (!target) return;
const mins = estimateMinutes(rt);
if (mins != null) target.textContent = `${mins} MIN READ`;
});
};
const init = () => {
applyWithin(document);
// Re-apply on dynamic changes (pagination/filters)
const mo = new MutationObserver((mutations) => {
for (const m of mutations) {
for (const n of m.addedNodes) {
if (!(n instanceof Element)) continue;
if (
n.matches('[data-rich-text], [data-rich-text="true"], [role="list"], .w-dyn-items, .w-dyn-item') ||
n.querySelector?.('[data-rich-text], [data-rich-text="true"]')
) {
applyWithin(n);
}
}
}
});
mo.observe(document.body, { childList: true, subtree: true });
};
// Robust bootstrapping
if (window.Webflow && Array.isArray(window.Webflow)) {
window.Webflow.push(init);
} else if (document.readyState === 'loading') {
document.addEventListener('DOMContentLoaded', init, { once: true });
} else {
// DOM is already ready; run now
init();
}
})();
Quality assurance in a call centre is the process of monitoring, evaluating, and improving agent interactions to ensure consistent customer experience and performance standards. All contact centres have a QA process. But most struggle to drive change from the data it provides.
For years, manual QA was the only option, and for many contact centres it still is. Supervisors sample a handful of calls, score them against a checklist and then file the results. Roughly 5% of interactions get reviewed on average. And then any feedback is given to agents days later (if at all).
It was never a great system. But as interaction volumes rise, agent workloads increase, and 42% of customers say they'll switch providers after a single poor experience, the cost of that insight-to-action gap is getting harder to absorb.
This guide covers how to make the shift from using QA as a monitoring exercise to using it as a driver of performance with AI-powered QA software.
You're monitoring quality. But are you actually improving it?
Knowing how to improve quality assurance in a call centre starts with an honest question: is your QA process actually producing change, or just producing data?
Traditional QA has a lag problem built in. A call happens and a supervisor reviews it days later. Feedback reaches the agent at a point where they've had dozens of conversations since the one that’s been reviewed. The connection between the behaviour and the coaching is weak, and the window for meaningful learning has already closed.
There's also the sampling issue. Manual QA typically covers around 5% of interactions.
“Leaders want answers, but those answers sit behind small QA samples, anecdotal feedback, and performance dashboards that only tell part of the story.” Connor Bowler, Principal Product Manager at MaxContact
The result is stark: manual QA gives you a story about some of your calls while an AI-powered platform gives you the truth about all of them.
Stop treating QA as an audit. Start treating it as a coaching tool.
Improving quality assurance starts with how you think about the QA function. It’s not an audit, but rather a coaching engine.
Approach QA with an audit mindset, and you’ll get reports. Approach QA with a coaching mindset, and you’ll get improvement. Contact centres that use QA to drive real behaviour change tend to do three things differently:
What They Do
Why It Works
Close the feedback loop fast
Feedback delivered within 24 hours lands harder. Agents have context, they remember the call, and the learning is concrete rather than abstract.
Make QA data visible to agents, not managers only
When agents can see their own scores and track their own trends, QA becomes something they're engaged with rather than something that's done to them. That ownership is where improvement starts.
Coach patterns, not just incidents
A single low-scoring call is an incident. Five with the same failure point is a pattern. Coaching patterns is where QA data creates lasting change.
As AI handles monitoring and scoring at scale, QA teams move away from manual call reviews and closer to coaching, analysis, and performance design; a more valuable role, and a more sustainable one.
It's a shift some organisations are already making.
The ICX Use Case
ICX, a customer engagement provider for brands including Nissan, Suzuki, and Stellantis, replaced manual call reviews with MaxContact's Conversation Analytics platform. Quality assessors moved from repeated audio replays to transcript-based reviews, with AI-powered search surfacing compliance issues, objection patterns, and coaching opportunities across every interaction. Training is now built directly from sentiment and objection data, feeding into one-to-ones and agent development.
Call centre quality assurance metrics: What they're actually telling you
Once you've made the shift from audit to coaching mindset, the next question is: what is your QA data actually telling you?
The most common scorecard measures script adherence, handling time, first call resolution, CSAT, and compliance markers. All of these are valid, but they can mislead if you're drawing conclusions from a 5% sample. The same metrics applied across 100% of interactions tell a very different story.
A few principles that make QA data more actionable:
1. Work out whether you've got a data problem or a coaching problem.
An agent who consistently mis-dispositions calls might not need coaching, they might need better data or a clearer process. An agent whose sentiment scores drop in the last hour of every shift has a different problem entirely. QA data is most valuable when it helps you tell the difference.
2. Don't look at scores in isolation. Connect them to outcomes instead.
A call that scores well on process but ends in a complaint tells you the agent followed the script and still got it wrong. Map your QA scores against CSAT, NPS, or complaint rates to find out which quality indicators actually predict good outcomes and which ones are just measuring process-following.
3. Track how quickly agents improve after coaching.
The rate of improvement following a coaching session is more useful than the score itself. If coaching isn't producing measurable change within a defined window, perhaps it’s the approach that needs to change, not just the agent's behaviour.
4. Use sentiment data to find what scores can't show you.
Scores tell you what happened procedurally. Sentiment analysis tells you how the customer felt at the start of the call, at the point of objection, and at sign-off. The gap between a high compliance score and a negative end sentiment is often where the most valuable coaching insight sits.
MaxContact's Auto QA applies customisable scorecards consistently across every selected interaction, including auto-fail criteria for non-negotiable standards, and surfaces sentiment alongside compliance scoring in a single view. QA managers spend less time manually reviewing calls and more time acting on what the data reveals.
Auto QA Score Card
The difference between feedback that lands and feedback that doesn't
Call centre quality monitoring best practices all point to the same conclusion: data doesn't change behaviour. Coaching does.
Specific beats general, every time. "You need to listen more actively" isn't actionable. "On this call at 2:34, the customer mentioned they'd been waiting three weeks and you moved on without acknowledging it. Here's what it sounds like when it's handled well" is constructive, actionable feedback. .
Frequency matters more than depth. Regular short coaching sessions (ten minutes a week focused on one call or one skill) tend to produce better outcomes than monthly deep-dives. Behaviour change is cumulative.
Self-review builds ownership. Agents who listen back to their own calls and score themselves before a coaching session arrive with more self-awareness and more investment in the gaps. The manager is coaching, not judging.
Use your best calls as teaching tools. Sharing anonymised examples of top-performing interactions gives the whole team a concrete standard to aim for. Not a number. A behaviour.
Data from MaxContact's Conversation Analytics platform drawn from over 700,000 objections across a six-month period, shows agents successfully overcome just 39% of objections, while 61% remain unresolved. The most challenging category is need objections ("not interested", "no immediate need"), which represent 46% of all objections but carry the lowest conversion rate. That's a pattern, and it needs a pattern-level response.
For BPOs like ICX, managing quality assurance across multiple client accounts at scale, running separate manual processes for each campaign simply isn't viable. "Anything that helps us connect to more of the conversations, especially given the volume we handle, is incredibly valuable. The team does a fantastic job, but no one can review everything manually. With Conversation Analytics, we can proactively support our agents and maintain complete oversight, so we never miss a critical moment or insight," said Sarah Franks, Call Centre Manager at ICX.
The hidden reason your QA findings never make it to the coaching conversation
There's a practical barrier between QA insight and coaching action that doesn't get talked about enough: post-call admin.
After every interaction, agents log outcomes, complete call notes, update CRM records, and prepare for the next contact. In high-volume environments, where over 52% of contact centre leaders report agent workloads have increased year-on-year, that wrap-up time absorbs the space that could go into engaging with coaching materials or reviewing their own performance data.
When agents are constantly catching up on admin, QA becomes something that happens to them in scheduled sessions, not something they engage with actively. The feedback loop gets longer. The shift from "QA as audit" to "QA as improvement" stalls.
Agent Wrap-Up Summary changes this directly. By automatically capturing call summaries, key outcomes, sentiment, topics, and follow-up actions at the point of wrap, it creates a consistent record for every interaction without adding work for the agent. That frees up the time and headspace to engage with performance data in real time.
What better QA actually means for your bottom line
For contact centre leaders asking how to improve quality scores in their call centre, the answer comes down to this: quality improvement has to show up in numbers the business cares about.
For BPOs, that means client retention. Clients expect their customers to be handled to a defined standard; when quality slips, renewals are at risk. With agent attrition running at 31% across the industry, AI-powered QA becomes even more valuable. New agents can be held to the same standard from day one, without relying on institutional knowledge that walks out the door.
For financial services and insurance contact centres, the case is just as direct. Agents who handle complaints well, identify vulnerability accurately, and resolve queries first time produce better CSAT, fewer escalations, and stronger retention. With first call resolution rates dropping from 43% to 37% year-on-year, the contact centres that reverse that trend through better coaching protect both customer relationships and commercial performance.
Either way, QA only delivers value if it produces measurable change, closing the loop between monitoring, coaching, and results, consistently and at pace.
How to make all of this work when you're dealing with real call volumes
The barrier has always been operational: the volume of calls, the limits of manual review, and the admin overhead that eats into coaching time.
The contact centres that improve quality consistently aren't doing something fundamentally different. They've just stopped treating QA as something that happens after the call and started building it into how the operation runs; faster feedback, visible data, coaching that's based on patterns rather than incidents.
At the volumes most contact centres are dealing with, that only works if the infrastructure supports it. AI-powered QA removes the manual overhead that makes it impractical, covering every interaction, surfacing what matters, and giving agents and managers the time to actually act on what the data shows.
If your QA process is still generating reports instead of results, it's time to change how it works. See how MaxContact's Auto QA and Agent Wrap-Up Summary turn insight into action.
Blog
5 min read
MaxContact named one of the UK's most thriving companies to work for
We're proud to announce that MaxContact has been named a winner of the Culture 100 Awards 2026, recognising us as one of the top 100 growing companies in the UK with a genuinely people-first working environment.
(() => {
const WORDS_PER_MINUTE = 200;
const MULTIPLIER = 1; // your choice
const estimateMinutes = (el) => {
if (!el) return null;
const text = (el.innerText || el.textContent || "").trim();
if (!text) return 1;
const words = (text.match(/\S+/g) || []).length;
const baseMinutes = Math.max(1, Math.ceil(words / WORDS_PER_MINUTE));
return Math.max(1, Math.ceil(baseMinutes * MULTIPLIER));
};
const findNearestTargetInItem = (itemRoot, rt) => {
if (!itemRoot) return null;
return itemRoot.querySelector('.is-text');
};
const applyWithin = (root) => {
// More forgiving selector: attribute present or equals "true"
root.querySelectorAll('[data-rich-text], [data-rich-text="true"]').forEach((rt) => {
const itemRoot =
rt.closest('[role="listitem"]') ||
rt.closest('.w-dyn-item') ||
rt.parentElement ||
root;
const target = findNearestTargetInItem(itemRoot, rt);
if (!target) return;
const mins = estimateMinutes(rt);
if (mins != null) target.textContent = `${mins} MIN READ`;
});
};
const init = () => {
applyWithin(document);
// Re-apply on dynamic changes (pagination/filters)
const mo = new MutationObserver((mutations) => {
for (const m of mutations) {
for (const n of m.addedNodes) {
if (!(n instanceof Element)) continue;
if (
n.matches('[data-rich-text], [data-rich-text="true"], [role="list"], .w-dyn-items, .w-dyn-item') ||
n.querySelector?.('[data-rich-text], [data-rich-text="true"]')
) {
applyWithin(n);
}
}
}
});
mo.observe(document.body, { childList: true, subtree: true });
};
// Robust bootstrapping
if (window.Webflow && Array.isArray(window.Webflow)) {
window.Webflow.push(init);
} else if (document.readyState === 'loading') {
document.addEventListener('DOMContentLoaded', init, { once: true });
} else {
// DOM is already ready; run now
init();
}
})();
The Culture 100 Awards, run by Maya, evaluate thousands of companies across more than 22 industry sectors. What makes this recognition different is how it's determined: not by self-reported data, but by anonymous sentiment surveys and open-ended responses from employees across participating organisations. Companies are assessed on verified employee benchmarks - the kind designed to uncover how people actually feel about where they work, not just how a business wants to present itself. For us, that's exactly what makes it meaningful.
As a team of around 70 people, we've grown steadily as demand for cloud-based contact centre and engagement technology has increased, and we're thrilled to have been selected for our commitment to building an environment that holds our people as a genuine competitve advantage.
Hannah Holmes, our Head of People, put it well: "We've been working hard to build an environment where expectations are high, accountability is clear, and people feel genuinely supported. This recognition tells us that work is landing in the right way."
CEO Ben Booth sees it as central to how we run the business: "Building a high-performing culture isn't a side project for us. We believe that getting our people strategy right is what enables us to serve our customers well and grow sustainably."
Being listed among the UK's most thriving places to work is something the whole team has earned, and it reflects the kind of company we're committed to being as we continue to grow.
Want to be part of it?
We're hiring. If you're looking for a place where the culture is real, not just a slide in an onboarding deck, take a look at our open roles.
Blog
5 min read
After-call work isn’t an efficiency problem- it’s a trust problem.
After-call work isn't just a time drain - it's a trust problem. Discover how inconsistent CRM records erode customer loyalty, and how AI-generated call summaries close the loop.
(() => {
const WORDS_PER_MINUTE = 200;
const MULTIPLIER = 1; // your choice
const estimateMinutes = (el) => {
if (!el) return null;
const text = (el.innerText || el.textContent || "").trim();
if (!text) return 1;
const words = (text.match(/\S+/g) || []).length;
const baseMinutes = Math.max(1, Math.ceil(words / WORDS_PER_MINUTE));
return Math.max(1, Math.ceil(baseMinutes * MULTIPLIER));
};
const findNearestTargetInItem = (itemRoot, rt) => {
if (!itemRoot) return null;
return itemRoot.querySelector('.is-text');
};
const applyWithin = (root) => {
// More forgiving selector: attribute present or equals "true"
root.querySelectorAll('[data-rich-text], [data-rich-text="true"]').forEach((rt) => {
const itemRoot =
rt.closest('[role="listitem"]') ||
rt.closest('.w-dyn-item') ||
rt.parentElement ||
root;
const target = findNearestTargetInItem(itemRoot, rt);
if (!target) return;
const mins = estimateMinutes(rt);
if (mins != null) target.textContent = `${mins} MIN READ`;
});
};
const init = () => {
applyWithin(document);
// Re-apply on dynamic changes (pagination/filters)
const mo = new MutationObserver((mutations) => {
for (const m of mutations) {
for (const n of m.addedNodes) {
if (!(n instanceof Element)) continue;
if (
n.matches('[data-rich-text], [data-rich-text="true"], [role="list"], .w-dyn-items, .w-dyn-item') ||
n.querySelector?.('[data-rich-text], [data-rich-text="true"]')
) {
applyWithin(n);
}
}
}
});
mo.observe(document.body, { childList: true, subtree: true });
};
// Robust bootstrapping
if (window.Webflow && Array.isArray(window.Webflow)) {
window.Webflow.push(init);
} else if (document.readyState === 'loading') {
document.addEventListener('DOMContentLoaded', init, { once: true });
} else {
// DOM is already ready; run now
init();
}
})();
Ask a contact centre leader about after-call work and they'll usually frame it as a time problem. Wrap time is too long. Agents aren't “going available” quickly enough. AHT is inflating. The fix, in most conversations, is operational: better templates, tighter ACW targets, more monitoring.
That framing is not wrong, but it is incomplete. After-call work is not just a time problem. It’s a quality problem, one which has a direct customer-facing cost that most operations are not measuring.
What actually happens when the call ends
The call ends. The agent is under pressure to “go ready” and be available for the next call in the queue. They have notes to write, a CRM record to update, a disposition to log. Often with multiple systems to update. They have approximately two minutes to do all of that before the queue moves. So, they write what they can. A sentence, maybe two. A shorthand that makes sense to them right now but will mean nothing to the agent who picks up next week's call. Sometimes nothing at all, and a disposition code carries the entire context of a complex interaction. Now multiply that across your team. Ten agents handling the same call type will leave ten different records. Some thorough, some minimal. Some missing the most important detail entirely - what was promised, what was escalated, what the customer was told to expect next. This is the quality problem, and it compounds quietly.
The customer pays for it twice
The first cost is visible: longer calls, higher AHT, agents unavailable for longer than they should be. This is what gets measured. The second cost is less visible but more damaging. The customer calls back. A different agent picks up. They open the record - and it tells them almost nothing useful. So, they ask the customer to explain themselves again. That moment - the repetition, the sense that the company was not paying attention - is where trust erodes. It’s not dramatic. It does not show up immediately in CSAT. But it accumulates, and eventually it becomes the reason a customer switches.
Our Voice of the UK Consumer 2026 research found that 42% of UK consumers have already switched provider due to poor contact centre experience. The word ‘already’ matters. These are not consumers who are at risk of switching – they’ve already left. The post-call gap is not just an internal inefficiency. It's a retention risk dressed up as an admin problem.
Why training cannot fix this
The instinct, when notes are inconsistent, is to retrain. Set clearer standards. Remind agents what a good record looks like. Monitor more closely. This rarely works. Not because agents do not want to do it well, but because the system is not set up to support consistency at pace. An agent writing notes under queue pressure, with no template and no structure, will produce exactly what the conditions allow. Varying quality, varying detail, varying usefulness. The problem is not discipline or intent. It is that the task is being done manually in the least forgiving conditions possible.
What changes when AI writes the notes
Agent Wrap Up Summary generates a structured call record automatically the moment the call ends; drawing on the conversation to produce a consistent summary of what was discussed, what was agreed, and what happens next. Every call. Every agent. Every time.
Consistency is the point. Not just the time saving, though that is real: wrap time typically accounts for 15–20% of an agent's working day, and a 50% reduction returns meaningful capacity to productive contact time. For a 50-agent team, that translates to an illustrative annual saving of £175,000: based on 50 agents, 50 calls per day, a 50% reduction in wrap time, and an average fully loaded agent cost of £25,000 per year.
The more significant change is downstream. When every call produces a reliable, structured record, that record becomes the foundation for what the next agent sees before their call begins. Customer History in Contact Hub surfaces that context automatically - so the agent who picks up next week is starting the call informed.
This is how personalisation at scale works. Not by asking agents to memorise histories or search through fragmented notes. By generating a complete record on every call, so context accumulates and becomes genuinely useful over time.
The record is where the loop closes
Agent Wrap Up Summary is the start of a feedback loop, not the end of one. The structured data it generates - consistent, covering 100% of calls - feeds everything downstream.
Conversation Analytics can analyse that data at scale, identifying coaching opportunities, surfacing compliance drift, and enabling AI Call Scoring that cuts QA review time from 30 minutes to approximately 5 minutes per call. Real Time Agent QA (available in Beta Q4 2026), uses it to guide agents in the moment, surfacing compliance prompts, flagging sentiment shifts, and steering conversations towards the outcomes that best records show actually work.
Better calls produce better records. Better records enable better coaching. Better coaching produces better calls. The loop only works when it is closed. And it closes after the call ends.
Start with the audit
You do not need a platform overhaul to find out where you stand. Pull a sample of CRM records from last week. Read them. Ask a simple question: if the next agent had only this record to go on, what would they know? The answer will tell you more about the state of your post-call process than any metric can.
Want to see how Agent Wrap Up Summary works in practice? Download The Assisted Agent - our practical guide to AI-enabled agent assistance across the full call lifecycle. Or if you'd rather see it live: book a demo with the MaxContact team.
.png)
