As AI raises expectations for real-time customer engagement, CIOs need to address the data architecture, governance and duplication issues that keep personalisation from scaling.

By Archie Marincowitz

Summary

• Personalisation often fails to scale because customer data is fragmented, duplicated and slow to update.
• Giving applications direct access to trusted data held in the organisation’s data platform can reduce duplication and improve governance.
• Faster access to current customer information allows businesses to respond more accurately to customer behaviour in real time.
• As AI takes on more customer engagement decisions, organisations need clear rules, human oversight and explainable outcomes.

---

For more than a decade, businesses have invested in CRM systems, customer data platforms, analytics and AI to deliver the right message to the right customer at the right time. Yet for many South African organisations, personalisation remains difficult to scale because the customer data behind it is often fragmented, duplicated and slow to update.

When campaigns fail to deliver relevant customer experiences, it is easy to assume the problem lies with the marketing team. More often, the constraint sits deeper in the organisation: in the way customer data is stored, duplicated, governed and made available to the systems that need it.

Consider what happens when a marketing manager at a bank, retailer or insurer wants to identify high-value customers in Gauteng who have opened the company's app in the past 90 days but have not made a purchase since.

The data team receives the request. An analyst soon discovers that "high-value" is defined differently across the CRM, loyalty platform and finance systems. The same customer appears multiple times because separate versions of their information exist across marketing, customer service, e-commerce and advertising platforms. Before the campaign can begin, the data has to be reconciled, extracted and prepared.

By the time the audience is ready, weeks may have passed. Some customers have already made another purchase, others may be considering a competitor, and some may have opted out of marketing altogether.

The result is an experience most consumers recognise. A customer receives an email saying, "We miss you," just two days after spending thousands of rand with the same business. Nobody in that process necessarily made a mistake. The campaign may have been planned correctly, the data extracted accurately and the message delivered exactly as intended. The problem is that the underlying customer data was already out of date.

The hidden cost of duplicate customer data

This is where many organisations turn to Customer Data Platforms (CDPs). The goal is sensible: bring customer information together, improve segmentation and make personalisation easier to manage. But when the CDP becomes another separate system with its own copy of customer data, it can add complexity to the very problem it was meant to solve.

In many organisations, the CDP sits separately from the core data platform, maintaining its own copy of customer information, governance processes and data pipelines. Keeping that information accurate and up to date requires continuous integration and ongoing maintenance.

Every additional copy of customer data introduces more complexity. Another system must be integrated, governed and secured. For South African organisations, it also increases compliance obligations under POPIA. Every location where personal information is stored must be protected, audited and updated when customers exercise their rights over that data.

While duplicate data may simplify a particular business process, it also increases operational overhead and compliance risk.

This is why delays in customer segmentation are rarely caused by a lack of skills or resources. More often, they're the result of architectural decisions made years ago, when moving data between applications was the most practical way to connect business systems.

Customer expectations have also changed. Consumers increasingly use AI-powered tools to research products, compare prices and make purchasing decisions in real time. Businesses that want to respond at the right moment can no longer rely on customer data that is refreshed weekly or even daily.

Marketing processes that were considered responsive a few years ago now struggle to keep pace with how quickly customers behave and make decisions.

A different approach to customer data

Technology vendors are starting to change how customer data platforms are designed.

Over the past few years, the industry has started moving towards a different approach: enabling applications to work directly with trusted customer data held in the organisation's data platform. This reduces duplication, simplifies governance and helps ensure teams are working from the same information.

One example is Databricks’ CustomerLake, which brings customer data platform capabilities closer to the organisation’s existing data platform rather than creating another standalone application. The significance is less about one product than the direction of travel: fewer duplicated datasets, stronger governance and a closer link between customer data, analytics and AI.

The benefits are practical. A more integrated customer data approach changes how organisations maintain customer records, create segments, manage governance and respond to customers in real time.

One of the biggest challenges organisations face is maintaining a consistent view of each customer across multiple systems. Modern customer data platforms increasingly use a combination of rules, AI and human oversight to match customer records more accurately over time.

Organisations no longer need to wait until every data quality issue has been resolved before they begin seeing value. Customer records improve continuously, allowing teams to make better decisions while the data matures.

Many marketing teams still rely on analysts or data engineers to create customer segments and prepare campaign data. That process introduces delays and creates unnecessary bottlenecks.

Modern customer data platforms are beginning to make segmentation more accessible by allowing business users to work directly with trusted customer data. Rather than responding to individual requests, data teams can focus on maintaining the quality, governance and availability of the underlying data.

Perhaps the biggest benefit is that customer information is updated far more quickly. If someone makes a purchase, changes their preferences or responds to an offer, that information can immediately influence future communications.

That means customers are far less likely to receive irrelevant messages, such as promotions for products they've already bought or "We miss you" emails just days after making a purchase. Personalisation becomes more accurate because it reflects what's happening now rather than what happened days or weeks ago.

Governance will determine success

For many executives, the concern is not AI itself, but how much decision-making authority is handed to autonomous systems that engage with customers on the organisation’s behalf.

As organisations move towards AI-driven customer engagement, questions about governance, accountability and oversight become just as important as the technology itself.

Those concerns are entirely reasonable. Organisations need confidence that AI systems will operate within clearly defined business rules and governance frameworks. One of the most important developments is the ability to test AI-driven campaigns before they go live. Modern platforms can simulate customer interactions, explain why particular decisions have been made and identify potential issues before customers are affected.

For highly regulated industries such as financial services, healthcare and insurance, that level of transparency is essential. Boards and compliance teams need confidence that customer communications can be explained, governed and audited.

The ability to explain AI-driven decisions before a campaign is launched is likely to become one of the most important requirements for organisations adopting these technologies. Explainability builds trust, supports compliance and gives business leaders greater confidence in automated decision-making.

AI should support human judgement, not replace it. Business leaders still need to define the strategy, objectives and guardrails, while AI helps execute those decisions consistently and at scale.

Questions every executive team should be asking

AI is opening up new possibilities for customer engagement. For executives, the challenge is ensuring these capabilities are used within clear governance frameworks, with the right oversight and accountability.

These questions belong on every executive team’s agenda:

• Who owns the trusted customer record? Who is accountable when different business units define “customer” in different ways?
• Which decisions can AI make, and where do humans need to stay in control?
• How must the marketing team change as it moves from building campaigns to managing engagement systems?

None of this requires organisations to start with a large re-platforming programme. The businesses that make the most progress will not be those waiting for perfect data, but those that begin strengthening the foundations now: reducing unnecessary duplication, improving governance and ensuring customer data can support faster, more trusted decisions.

For CIOs, personalisation at scale will depend less on the next channel, campaign or AI tool than on the quality, governance and availability of the customer data behind it.

---

About the author:

Archie Marincowitz is a Business Development Executive at DVT, helping organisations accelerate digital transformation through AI-driven software engineering, cloud and data solutions. Connect with Archie on LinkedIn.