Sit next to a good agent on a payments process and you will notice something odd. The customer is reading out sixteen digits, and the agent is not writing them down. Sometimes the agent is not even hearing them. The customer types the digits on the phone keypad, the agent’s headset goes quiet for a few seconds, and a green tick appears on the screen. That is not a gimmick. It is the visible part of a standard called PCI DSS, and once you see why it exists, a lot of contact-centre rules stop looking like red tape.
What the standard is, and who it binds
PCI DSS is the Payment Card Industry Data Security Standard. It is written and maintained by the PCI Security Standards Council, which was set up by the big card brands. It is not a law passed by Parliament. It is a contractual requirement: if your business stores, processes or transmits cardholder data, the card brands and your acquiring bank expect you to follow it, and your client will usually write it into the contract with you. A BPO taking card payments on behalf of a retailer or a utility is squarely inside this.
The data it protects falls in two buckets. The first is cardholder data: the card number (called the PAN), the name, the expiry date. The second is sensitive authentication data: the three or four digit security code on the card (CVV or CVC), the full magnetic-stripe data, and PIN data. The rule you should remember is blunt. You may store the first bucket if you protect it properly, for example by rendering the PAN unreadable. You may not store the second bucket after the payment has been authorised, even if it is encrypted. A security code sitting in a database is a breach waiting for a date.
Why a call recording is the classic problem
Think about what a contact centre already does. It records calls for quality and dispute handling. It keeps those recordings for months. Now a customer reads out card number, expiry and security code on one of those calls. The recording has just become a store of exactly the data the standard says must not exist. Anyone with access to the QA tool, the storage and the backups can now listen to it. Multiply by thousands of calls a month and you have a very attractive target, plus a long list of people (auditors, team leaders, trainers, IT staff) who can reach it.
Agents themselves are the other exposure. Most agents are honest. But a floor of several hundred people, with high attrition and a pen and paper within reach, is a large population to trust with a stream of card details. The standard does not say agents are thieves. It says you should not build a process whose safety depends on nobody ever being tempted.
The usual ways of shrinking the problem
The smartest idea in PCI work is not to protect data better but to hold less of it. The compliance people call this reducing scope. Every system, person and room that touches card data is in scope and must be assessed. Remove the data from a system and that system drops out. Common techniques you will meet:
- DTMF masking. The customer enters digits on the keypad. The tones are suppressed or flattened so the agent and the recording hear only a monotone, while the digits go straight to the payment provider. The agent stays on the line and can still help if the customer is confused.
- Pause and resume. The recording stops while card details are spoken and restarts afterwards. Cheaper to set up, but it relies on the agent pressing the button at the right moment, and the agent still hears everything. Auditors tend to treat it as weaker than masking.
- Transfer to a payment IVR. The agent hands the customer to an automated system that collects the card, takes the payment and returns the call with only a success or failure result.
- Tokenisation. After the first payment, the provider returns a token, a meaningless substitute that is useless outside that merchant. Your systems store the token, not the card.
Notice the pattern. None of these makes the agent more careful. They make the agent irrelevant to the card data, which is a much stronger control than training.
The floor rules that go with it
Even with masking in place, PCI-scoped processes usually come with physical and procedural rules, and they all trace back to the same logic. No pens, paper or phones on the production floor, because those are ways of copying data out. Restricted print and copy-paste. Screen locking. Individual logins, never shared ones, so every access can be traced to a person. Clean-desk checks. Network separation, so that the machines handling card data sit apart from the general office network. Access to the room, to the systems and to recordings is granted on a need basis and reviewed regularly. Logs are kept, and someone actually looks at them.
If you have ever wondered why a team leader cannot just carry a notebook onto the floor for coaching notes, this is why. It is annoying on a Tuesday and sensible the day a notebook goes missing.
How compliance is actually checked
The route depends on the size and type of the organisation. Large service providers are usually assessed by a Qualified Security Assessor, an independent firm approved by the Council, who produces a Report on Compliance and an Attestation of Compliance that you hand to clients. Smaller ones may complete a self-assessment questionnaire. Which applies to you is decided by card-brand rules and your acquirer or client, so check, do not assume. Assessments are repeated every year, and the standard itself is revised: the long-running version 3.2.1 was replaced by version 4.0, and some of the new requirements only became mandatory from 2025. If you are preparing for an audit, use whichever version your assessor names.
One more practical point. Compliance is a snapshot, a promise about how things are run on an ordinary Tuesday. The questions an assessor asks are mostly about evidence: show me the access review from last quarter, show me who can reach the recordings, show me what happens when someone leaves. A process that is secure but undocumented fails in the same way as one that is insecure.
When the controls start to leak
Where do real processes slip? Usually at the edges. A customer starts reading the card number aloud before the agent has activated masking, so the first digits land in the recording. Chat and email arrive with card numbers typed in by customers, and nobody planned for that channel. Screenshots taken for a ticket capture a payment page. Work-from-home seats appear, and the clean-desk rule now depends on a family living room. Each of these needs its own answer: automatic redaction in chat, instructions to customers not to send card details in writing, restrictions on screenshots, and stricter controls or a decision not to take payments from home at all.
So when someone says “PCI is an IT thing”, politely disagree. The IT part is real, but the standard bites hardest on how people sit, talk, record and log in. It is about operations, and operations teams carry most of the day-to-day weight.
Leave a Reply