One contact centre for eleven countries — one platform, not eleven islands

The situation

Customer service in eleven countries almost always means, in practice, eleven systems that grew separately: their own numbers, their own announcements, their own opening hours, their own reporting — and no way to compare two countries without manually reconciling the figures first.

On top of that sits the problem every agent has daily and no phone system solves: who is calling is in the CRM. What they bought is in the ERP. What the product does is in the product system. The agent has three systems open and a customer on the line.

The approach

A shared contact centre platform on Twilio Flex for all eleven countries, with voice and SMS as channels.

Call control lives in configurable flows rather than hard-wired logic: blocked numbers, opening hours, public holidays, direct dial and routing to the right queue all run as per-country configuration through one shared mechanism. Distribution to agents is handled by the platform’s task routing, with skills and languages as criteria.

Data enrichment is the part agents actually notice. On an incoming call the number is resolved and the workspace shows, merged into one view, what the source systems hold — customer master data and orders from SAP, cases and ownership from Salesforce, article and document data from the product system. No alt-tab, no second window, no phone number retyped somewhere.

Reporting is custom-built rather than out-of-the-box. A platform’s bundled metrics answer the platform’s questions. This organisation’s questions — comparability across eleven countries, behaviour per brand, reachability per service line — needed their own reports over the raw data.

Where it could have gone the other way

One instance for eleven countries, not eleven instances.

The obvious route, with eleven countries and eleven sets of habits, is to give each country its own environment. It is tempting because it is politically easy: every country changes what it likes without asking anyone.

The cost would have been eleven times the operations, eleven times the upgrades, eleven times the security review — and above all eleven pools of data from which no defensible country comparison can be built. That comparison was one of the goals.

So the decision went the other way: one platform, and everything that distinguishes a country is configuration. Numbers, announcements, opening hours, queues, languages and destination systems exist as data, not as code. The AI voice assistant built later on this platform makes the pattern explicit — it draws its permitted destinations at runtime from exactly that country configuration.

The downside is real and has to be lived with: a change to the shared mechanism affects all eleven countries. That forces coordination where previously nobody had to ask, and a mistake lands wider. The trade is comparability and operability against local autonomy — and it is a decision you make once, right at the start.

The result

One platform for eleven countries, rolled out country by country over several years and still being extended. Two channels, voice and SMS. Three core systems merged into the agent workspace, plus custom reporting over the raw data.

The strongest evidence that the foundation holds is what was built on top of it — four of those are described separately here:

• a desktop application that lifts the agent workspace out of the browser
• an AI error diagnosis that analyses telephony logs from every workstation centrally
• an AI voice assistant replacing a 114-widget menu
• a contact data overlay letting agents correct master data during the call

Each of those was possible because the platform is shared and the countries are configuration.

Deliberately not claimed: I have no measured before/after figures for reachability, handling time or cost. The numbers above describe scope, not effect.