Standardize the Architecture, Localize the Execution

On August 2 the EU AI Act became fully applicable. That was three weeks ago. If your post-sale organization serves European customers, that date changed what you can safely ship into a health score, a churn prediction, or an automated outreach sequence. If your customers are all in North America, the date meant nothing to you at all.

That gap is the recurring problem in global post-sale work, and most companies handle it badly in one of two directions. Some write a single playbook and push it everywhere, then wonder why adoption curves in Sao Paulo look nothing like Chicago. Others let every region invent its own version of everything, and close the year with three definitions of "at risk," four renewal forecasts that cannot be added together, and a board deck held up by narrative.

The fix is not a compromise between those two. It is a split along a different axis. Standardize the architecture. Localize the execution. They are separate layers with separate rules, and most global programs break because someone mixed them up.

The architecture is the part you can add up

Architecture is what has to be identical in every market, because it is what the company reasons with. Lifecycle stages. Account segmentation rules. The inputs that feed a health score. Partner tiers and what qualifies a partner for each one. The narrative structure of an executive business review. The cadence of your operating reviews.

The test for whether something belongs here is arithmetic. If two regions define a thing differently, can you still add the numbers together? If Brazil calls an account onboarded when the contract is signed and Germany calls it onboarded when the first workflow is running in production, your time-to-value number is not inaccurate. It is fiction. It describes nothing that exists.

Health scores work the same way. You can weight the inputs differently by market and you should, because a support ticket volume that signals distress in one region signals engagement in another. What you cannot do is use different inputs and then pretend the outputs are comparable. A red account in Mexico and a red account in the Netherlands have to mean the same category of trouble, even when the thresholds that trigger red are tuned locally.

Partner tiers rarely get this discipline and badly need it. If a Gold partner in one region means a revenue commitment and in another means a long relationship and good intentions, you do not have a partner program. You have a naming convention. The criteria belong in the architecture. What a partner has to do to hit them can flex.

The execution is the part customers actually experience

Everything downstream of the architecture is local, and the list is longer than most companies plan for: payments, communication channels, language, compliance controls, and the mix of partner types that create trust in a given market.

Payments is where this stops being abstract. In Brazil, Pix reached nearly 170 million users and BRL 11 trillion in transactions in 2024, according to Banco Central do Brasil. That is not a payment preference. That is the payment system. EBANX's materials show Canva localizing across Latin American markets with local cards, digital wallets, installments and Pix, and Zuora partnering with EBANX specifically to improve subscription and recurring payment acceptance across 13 Latin American markets. Read that as an operating fact rather than a fintech story. If a renewal fails because card acceptance in a market is structurally weak, your CSM is now managing a churn event that has nothing to do with the product or the relationship. In LATAM, payment acceptance is part of value realization, which makes it part of retention.

Channels have the same shape. Infobip's 2026 research puts WhatsApp penetration at 85% in Brazil, 80% in Argentina and 75% in Colombia, with regional interactions up 62% during 2025. The same research reports 76x RCS growth in the US during 2025 after Apple rolled out support, with voice holding strong for support and high-value conversations. Europe is a third animal: a mature, privacy-sensitive market where verified senders and anti-smishing measures materially affect whether your message earns engagement at all. A lifecycle program that assumes email is email will underperform in all three places, for three unrelated reasons.

Then compliance, which is the least optional item on the list. The AI Act entered into force in August 2024, AI literacy obligations started applying in February 2025, general-purpose AI model obligations followed in August 2025, and full applicability arrived this month. GDPR sits underneath all of it. The EU Data Act has applied since September 2025 and is designed in part to make cloud switching and interoperability easier, which is worth sitting with for a moment. A regulation built to lower switching costs is a retention variable, not only a legal one. European customers will find it structurally easier to leave. That should change what you consider a defensible account, and it should change it before the renewal conversation, not during it.

Partner mix is the quiet one. The partner types that create trust vary more by market than any deck admits. Some markets buy through systems integrators with deep local reputations. Some buy through agencies. Some buy direct and treat partners as implementation capacity after the fact. Standardizing the tiers is right. Standardizing which partner types you recruit into them is how you end up with a beautifully governed program that no local buyer cares about.

The most common mistake is the layers reversed

Plenty of companies do this exactly backwards. They localize the architecture, letting each region quietly redefine churn, health and lifecycle stages to fit local reality, and then standardize the execution by shipping one email sequence in one language through one channel to everyone. That combination is the worst available. You lose comparability at the top and relevance at the bottom, and the two failures hide each other. The regional numbers look fine because each region grades its own homework, and the customer experience is uniformly mediocre in a way no dashboard reports.

I have never been in an argument about global consistency that was actually about global consistency. Underneath, it was always an argument about which layer a decision belonged to. Someone in a region wanted to change something the company reasons with, or headquarters wanted to standardize something the customer feels. Naming the layer usually ends the argument in about five minutes.

Governance travels differently too

Gainsight's 2025 Customer Success Index reports that 37% of European respondents said Customer Success reported to the CEO, and that a higher share of European respondents reported budget increases. Treat that as directional rather than law, but the operating implication holds. The escalation you would route through a CRO in the US may need a different path in Europe, because the function sits somewhere else on the org chart entirely. Your architecture decides that an escalation happens and what triggers it. Local execution decides who receives it and how fast they can act.

Standardization is what lets you compare, forecast and defend. Localization is what makes any of it land with a customer. The operator's job is to know which of the two you are touching before you change anything, because the cost of getting that wrong compounds quietly for about three quarters and then arrives all at once.

One caveat worth stating plainly: most of the research above is vendor-produced or survey-based, and the LATAM data in particular is country-specific rather than regional. It is current and close to workflow reality, which is why I use it. It is not audited, which is why I would not build a plan on any single number in it.

I write more about post-sale and partner operations at https://www.santiagomarin.net/blog