By 31 October 2026 Vee had handled 48,626,959 cases, and every one of them was executed by Vee's AI. Vee employs no call agents, so this is not the automated slice of a larger human operation; it is the whole operation. This note explains the four cumulative numbers we publish, how to read them, and what they do not tell you.

We publish these figures for a reason that has little to do with size. Large numbers are easy to quote and hard to interpret, and our industry has a habit of quoting them without definitions. If a number carries no unit and no scope, it is decoration. So here are ours, with both.

The four numbers#

Vee's cumulative operating aggregates to 31 October 2026
AggregateValueWhat it counts
Cases48,626,959Records assigned to Vee for handling. A case is not a person and not a call.
Conversations38,325,799Connected conversations conducted by Vee's AI.
AI conversation minutes28,569,847Time spent in conversation by Vee's AI; human agents are not included.
Call attempts298,684,764Single dial attempts, whether or not anyone answered.

Cumulative to 31 Oct 2026, Vee operating data. The four figures use different units and should not be combined into ratios.

Each of the four measures a different thing. Cases measure the work assigned to us. Attempts measure the effort spent trying to reach people. Conversations measure how often that effort turned into a dialogue. Minutes measure how long those dialogues lasted. None of them, on its own, measures whether the work succeeded.

What a case is, and what it is not#

The case is the unit we chose to lead with, because it is closest to how our clients think about work. A case is a record assigned for handling: a loan account in early delinquency handed to us this week, a patient whose appointment needs confirming, a lead who filled in a form an hour ago, a customer to be surveyed after a hotline call.

A case is not a person. One account can produce several cases in a month, for example when a bank reassigns it after a broken promise or when it moves to the next delinquency stage. A case is also not a call. Some cases are closed in one conversation; others need several attempts across days, an SMS before or after, and a callback at an agreed time.

We keep three units apart in every project: the account (the customer relationship, such as a loan contract), the case (each assignment of that account to a process) and the attempt (a single dial). We also keep four status axes apart for every case: technical, dialogue, processing and business. The business status is the client's version of the truth, in its own language, and it is not comparable across projects. That discipline is what makes a figure like 48.6 million meaningful at all.

What sits behind a single case#

To make the unit concrete, follow one case through an early-collections process at a bank. The bank's analysts decide which accounts to assign, from which portfolio and bucket, and when. The account arrives in our system as a case with the data the dialogue needs and nothing more.

  1. The dialler schedules the first attempt inside the call window allowed in that market, from an outbound number chosen for that base.
  2. If the customer answers, Vee's AI introduces itself as AI and verifies identity before any debt detail is mentioned. That rule is enforced by the platform, not left to a script line.
  3. The conversation is shaped by the stage of delinquency and adjusted by history: a broken promise from last month is named, not ignored.
  4. If the customer commits to pay, the date and amount are recorded; if they describe hardship, ask for a person or dispute the debt, the agreed route applies: a transfer, a stop or a referral.
  5. If nobody answers, the next attempt is scheduled according to what happened: follow-up attempts and any text message follow the contact rules agreed for the process.
  6. Every day the bank receives case-level statuses it can act on, and later the payment data tells both sides whether the promise held.

One case, in other words, may mean several attempts, one or two conversations, a text message and a status update, spread over days. Multiply that by tens of millions and the reason for keeping units apart becomes obvious.

Why “every one executed by Vee's AI” matters#

When a contact-centre operator publishes a large volume, the first question is what share was handled by automation and what share by people. A hybrid operation is a legitimate business, but it is a different one.

Our number is not a fraction of anything. The conversations behind those 48.6 million cases were executed by Vee's AI, because Vee does not employ people to make or take calls. The people at Vee design dialogues, engineer telephony and integrations, listen to calls every day, analyse results and improve the process.

We want to be precise about what this does not mean. It does not mean no human is ever involved. Our processes include agreed transfers to the client's own staff: a customer in financial hardship, a dispute, a request for a person, a sales conversation ready to close. Those routes are designed deliberately. What it means is that the work we were assigned was carried out by our AI, at a scale where every weakness in a dialogue, a retry rule or a voice shows up quickly.

We don't manage people, so we control the service far better: one change and every live conversation already reflects it. – Dawid Wójcicki, Founder & CEO of Vee

In regulated sectors there is a second consequence. When conversations are executed by AI from approved, versioned content, every one of them can be reconstructed afterwards: what was said, which intent was recognised, which data field was filled, which decision was taken and which version of the script was live. A bank or insurer auditing its supplier is not sampling the memory of individual agents; it is reading a record. Across 48.6 million cases, that consistency is worth more than any single headline figure.

Reading attempts and conversations#

The largest number, 298,684,764 attempts, is also the least glamorous and in some ways the most instructive. In our outbound processes, most dialling attempts never become a conversation. Phones are busy, switched off or unanswered; voicemail picks up; the customer is driving, at work or asleep; the number is out of date.

That is why reach is the first rung of every process we run. Without reaching the customer, there is nothing else: no continuation, no information, no business result. A large part of what we have learned over the years sits in this gap between attempts and conversations. It concerns outbound numbers, calling windows, retry rules and the role of text messages, and the answers differ by market and by base.

We deliberately do not publish a ratio between attempts and conversations. The ratio depends almost entirely on the base: a fresh lead, a known customer in a loyalty programme and an account worked over by several previous campaigns behave in completely different ways. A ratio across all of them would describe our process mix, not our skill, and it would invite exactly the kind of comparison that misleads buyers.

The 38,325,799 conversations, for their part, are the moments where reach succeeded and the dialogue began. Each one is a recorded, reconstructable interaction: recording, logs, statuses and the version of the script that ran. That is the raw material for the daily listening and analysis that improve the next conversation.

Reading the minutes#

The 28,569,847 minutes count time in conversation by Vee's AI. They do not include human agents, either ours (we have none) or the client's staff who receive transfers.

For years minutes were the industry's default measure of activity, and for a long time they were part of how we were paid. We have moved away from both. A minute rewards long calls, while in many processes the result comes from short ones. In early collections in banking, for example, cases ending with a promise to pay have historically averaged 116.5 seconds of handling time; short is not rude when the conversation is specific.

Since the start of 2026 our minutes have been falling faster than our cases. Calls have become shorter, the campaign mix has changed, and foreign processes have grown. On 15 September 2026, in current report ESPI 11/2026, we ended regular publication of case volumes and minutes as operating indicators. We still publish cumulative aggregates like the ones in this note, as a record of experience. We no longer present them as a measure of performance.

How the numbers grew#

The company was founded in 2012, and production at scale began in 2017 in regulated sectors. The cumulative totals reflect years of steady growth, with a visible acceleration in the last two.

Selected milestones in annual and monthly volume
PeriodCasesMinutes
2024 (year)8.2 million5.0 million
February 2025First month above 1 million–
October 20251.164 million (monthly record)–
Q4 20253.02 million (best quarter)–
2025 (year)11.61 million (+45.7% y/y)7.18 million
Cumulative to 31 Oct 202648,626,95928,569,847

Vee operating data and periodic reports. Regular publication of case and minute volumes ended with ESPI 11/2026 (15 Sep 2026).

Behind the growth is the Land & Expand pattern that shapes most of our relationships. A client starts with one process or portfolio. When the results are proven, the cooperation extends to further buckets, portfolios, business units and, increasingly, countries. Volume is the by-product of that trust, not a goal in itself.

What the aggregates do not tell you#

It would be easy to let a number like 48.6 million do the selling. It should not. Here is what it cannot show.

  • Effectiveness. Volume says nothing about whether a customer paid, attended, renewed or bought. Effectiveness is measured per process, against a comparator, in the client's systems.
  • People reached. Cases are not people. The number of distinct customers behind them is lower and is not something we aggregate across clients.
  • Revenue. Our financial results are published in our periodic reports. The relationship between volume and revenue varies widely by process and billing model, and it has become weaker as billing moves away from minutes.
  • Comparability with other suppliers. Others may count calls, sessions, interactions or contacts. Without the same definitions, totals cannot be compared.

Where we publish effectiveness, we publish it with scope. In early-stage collections, historically, we have measured an increase of 7–13% in on-time payment frequency versus a human-agent comparator. In soft collections in insurance, historically, 61.9% of cases were closed at first contact. In a CSAT survey in the energy sector, historically, the fully completed survey rate was 43.5%. Each of these figures belongs to a specific process and period, and none of them should be read as a universal rate.

Where the cases come from#

The total is spread across sectors and process types that behave very differently. A few examples show the range.

  • Banking and consumer finance. Early collections from the first days of delinquency, receivables monitoring, lead qualification and cross-sell. Speed matters because each day without contact moves an account closer to the next bucket.
  • Insurance. Soft collections, renewals and cross-sell during the life of a policy, and booking medical benefits for insured persons, where we call partner clinics until we find a slot that matches the patient's preferences.
  • Healthcare. Appointment confirmation that is, in business terms, early cancellation: learning that a visit will not happen while there is still time to refill the slot.
  • Energy and utilities. Satisfaction surveys after a hotline or point-of-sale contact, collections, contract renewals and confirming access before a technician drives out.
  • Telecommunications. Retention, renewals and pre-sales at subscriber-base scale, with interested customers passed to salespeople.

Each of these has its own definition of success, its own rhythm and its own regulatory constraints. The aggregate hides that variety, which is one more reason not to read it as a single performance figure.

What scale has taught us#

Scale is not just a number; it is exposure. Running tens of millions of cases means that rare situations stop being rare. A phrase that confuses one customer in a thousand becomes a pattern. A time window that works in one market fails in another. A voice that converts in a warm base sounds wrong in a cold one.

  • Many campaign types, especially outbound. Early collections, receivables monitoring, lead qualification, appointment confirmation and cancellation, renewals, surveys, inbound FAQ lines.
  • Demanding sectors. Banking and consumer finance, insurance, healthcare, telecommunications and energy, where mistakes cost money and licences.
  • Daily human listening. In every project our people listen to selected calls every day, especially unusual ones, to hear what the metrics cannot.
  • Change at the speed of configuration. Because conversations are executed by our AI, an approved improvement applies to every live conversation at once.

The other lesson is that scale is an infrastructure question, not a staffing one. When demand from our large clients jumped at the start of the pandemic in March 2020, we expanded telephony capacity rather than recruiting. The same principle lets us start a new portfolio in a new country without building a local call floor.

Scale also changes how we test. With enough volume, a proposed change can be tried on a comparable part of the base and judged within days rather than months. The client decides whether the change goes live; the data makes the decision easier.

How we will report from here#

The end of regular volume reporting in September 2026 was not a retreat from transparency. It reflected a change in what our business is. As billing increasingly depends on effectiveness and results, equal minutes or equal cases can carry very different economic value. Work on our future reporting standard focuses on three dimensions: the economic value of the business, operational effectiveness expressed through measurable results, and process economics.

We will continue to publish cumulative aggregates from time to time, because they are a fair record of experience and because readers ask for them. When we do, they will carry the same units and scope lines as in this note.

There is a practical reason we still count at all. Cumulative volume is the closest thing an operator has to flight hours. It does not prove that the next process will succeed, but it does mean that most of the failure modes a new client fears, from a base full of outdated numbers to a sudden spike in hang-ups after the company name, are things we have already met, diagnosed and fixed somewhere else.

If you are comparing suppliers#

Volume claims are everywhere in our industry. These questions will tell you quickly whether one is meaningful.

  1. What exactly is being counted: cases, accounts, people, calls, attempts or minutes?
  2. What share of the volume was handled by automation, and what share by people?
  3. Over what period, and is it cumulative or annual?
  4. In which sectors and process types does the volume sit?
  5. What effectiveness figures sit alongside it, with which scope and comparator?
  6. Does the supplier run the process, or only provide the software?

Our answers are in this note. Forty-eight million cases is a record of what we have done since 2012, in demanding sectors, at the client's side throughout. What it buys a new client is not the number itself but the experience behind it: knowing, before launch, most of the ways a process can go wrong.

Follow The First Second

Occasional issues on the mechanics and economics of AI-powered customer contact. No noise, no vendor news.

Follow via RSS