Nobody is listening to 98% of your customer calls

By Arsalan Kamran, CEO and co-founder of AssistRing

Ask a support director how their team is performing and you will usually get a confident answer. Ask how they know, and the answer gets thinner.

Most quality assurance in customer support works like this. A supervisor pulls a handful of recorded calls at the end of the week, listens to them, fills out a scorecard, and files it. In a team handling 5,000 calls a month, that supervisor might review 100. Across the outsourcing industry, standard practice sits somewhere between one and three percent.

The other 97 percent are never heard by anyone except the agent and the customer.

I want to be careful about what that does and does not mean. Most of those unheard calls are fine. Agents are not sabotaging your business when nobody is watching. But the calls that matter most are almost never in the sample. The compliance slip. The refund that should have been escalated. The customer who was told something the company cannot actually deliver. The caller who hung up polite and never came back.

You do not find those by listening to 2 percent. You find them by accident, usually months later, through a complaint or a chargeback or a churn report that arrives with no explanation attached.

Where this started

I started AssistRing in 2024 with my co-founder Oscar Lausen. We are a managed BPO with registered entities in the United States and Germany, and our delivery operation runs out of Karachi, Pakistan.

I did not come into this from a call center background, and the thing that bothered me in the first few months was how little anyone actually knew about their own support function. Clients would send us monthly reports full of averages. Average handle time, average CSAT, average resolution rate. Averages are comfortable. They are also where problems go to hide.

A team with 97 percent satisfaction and a handful of badly mishandled compliance calls looks identical on a dashboard to a team with 97 percent satisfaction and none. The average absorbs it.

Every BPO will tell you they care about quality, then hand you a report built on two percent of the calls. If you are only listening to two percent, you are not measuring quality. You are guessing at it and hoping the sample was representative.

So the question I kept coming back to was simple. What changes if you review every call instead of a sample?

What we built

That question turned into ARIS, the intelligence layer that now runs across every AssistRing account. It is not software we license separately. It comes with the team.

Sage is the part that answers the original question. It scores 100 percent of conversations against the client's own scorecard, every day. When a supervisor opens the dashboard on Monday, they are not looking at a hand-picked selection. They are looking at everything, with the outliers already surfaced. The practical effect is not that supervisors work harder. They stop guessing which calls to listen to and spend their time coaching instead of sampling.

Around Sage sits the rest of the suite. Edith handles live voice translation across more than 100 languages, so an agent and a caller who do not share a language can still have a conversation in real time. Agent Copilot pulls the right SOP, policy or answer in front of the agent while they are still on the call, which shortens ramp time for new hires and stops knowledge gaps turning into escalations. VOX covers the automated voice and chat tiers, the repetitive contacts that do not need a person, and hands off cleanly the moment one does.

A human stays in the loop across all of it. My honest view is that most of the industry has the AI question backwards. Everyone is racing to automate the conversation while leaving the measurement manual. We did the opposite. Automate the measurement, give agents better tools in the moment, and protect the human on the calls that decide whether a customer stays.

A number I keep coming back to

Royal Nawaab is one of the largest Pakistani hospitality groups in the United Kingdom, and one of our earliest accounts. High volume, heavy peaks, reservations and order enquiries stacked on top of each other through dinner service.

Seven agents handle over 900 calls a day on that account. Average handle time runs between one and two minutes, and most of the team sits at 97 to 98 percent customer satisfaction. Calls come in across Hindi, Urdu, Punjabi and English, sometimes switching mid-conversation, which is normal for that customer base and close to impossible to staff for conventionally. Their verified Clutch review documents a call drop rate that fell from 40 percent to under 3 percent after we took the account.

Seven people, 900 calls. Roughly 128 calls per agent per day, well above what a conventional team of that size manages. It works because the agents are not doing the parts a machine does better. They are not writing call notes by hand, not hunting through a knowledge base while the customer waits, and not blocked when a caller switches language.

That number does not transfer to every account. A technical support queue with 12 minute calls will never look like a restaurant booking line. But the principle holds everywhere we operate. A small, dense, well-instrumented team beats a larger conventional one, and it costs less.

We work the same way in healthcare staffing. Locum Smart places pharmacists and dispensers into permanent roles across the UK, and their bottleneck was never headcount. It was candidate response speed, because in recruitment the firm that calls back first usually wins the placement. Full QA coverage in that context is not politeness scoring. It is knowing that every candidate conversation followed the process, every time, without a manager listening in. AmazeWholesale runs order and returns management with us across ecommerce and wellness.

What working with us actually looks like

We are the legal employer of every agent on your account. Not a staffing agency, not an employer of record, not a marketplace passing contractors through. We hire them, train them, manage them and answer for them. It sounds like a technicality until something goes wrong, at which point it is the difference between one accountable provider and a chain of intermediaries arguing about whose problem it is.

Teams are dedicated and sized to the account rather than pooled across clients, and we scale headcount as volume grows. We start at two agents, full time. I mention that specifically because most BPOs will not take you seriously below 20 seats, and that is a mistake in a market where a 40 person company can have a support problem just as expensive as a 400 person one.

Every engagement includes ARIS. There is no premium tier where quality coverage suddenly becomes available. If we run your support, all your calls get scored.

For UK and European clients we operate GDPR-compliant data handling. For healthcare and healthcare-adjacent work we maintain HIPAA-ready workflows, including access controls, agent training and audit trails. Our US and German entities exist so clients can contract locally and, for DACH accounts, deal with German-speaking leadership rather than a time zone away.

The part most people skip

Outsourcing customer support fails for boring reasons. Not because the offshore team is bad, but because nobody defined what good looked like, nobody checked whether it was happening, and by the time the numbers showed a problem the relationship had already soured.

Full QA coverage does not fix a bad process. What it removes is the excuse. When every call is scored there is no argument about whether a problem is real, no debate about whether the sample was representative, and no six month lag between something going wrong and someone noticing.

That was the gap I wanted to close when we started AssistRing, and it is what ARIS is for.

If you run a support operation and you are working off a 2 percent sample, go and check what the other 98 percent contains. You can do that with us or without us.

Arsalan Kamran is the CEO and co-founder of AssistRing, a managed BPO providing customer support operations across the United States, Germany and Pakistan. More at assistring.com, the ARIS suite at assistring.com/aris, and client work at assistring.com/portfolio.