Why most AI transformations fail before they even begin

Author: Paul Stemmet, managing director at Dentsu Data Labs, EMEA: sub-Saharan Africa

The ambition is clear: improve efficiency, unlock growth and future‑proof the business.

Yet despite this momentum, many AI initiatives fail to scale or deliver meaningful impact. The issue is rarely the technology. The difference is where organisations start.

The real problem is not AI. It is understanding the customer

Most AI programmes start with tools. New models. New platforms. New automation. What they often lack is a unified, usable understanding of the customer.

Data exists in abundance, but it is fragmented across systems, channels, and teams. Identity is duplicated, inconsistent or inferred.

In this environment, AI does not create clarity. It simply accelerates fragmentation.

True transformation starts earlier. It starts with identity. It starts with first‑party data. Without this foundation, AI remains an experiment rather than a business capability.

What enterprise AI transformation looks like

At Telkom, the challenge was not a lack of data. It was fragmentation, and the growing pressure to unlock value from it.

Customer, behavioural and transactional signals existed across the organisation, but they were disconnected and underutilised. As competitive intensity increased and traditional revenue streams came under pressure, incremental optimisation was no longer sufficient.

What was needed was not another layer of technology, but a way to bring clarity to the customer.

By applying a person‑based identity approach, enabled through dentsu Merkury, Telkom unified its first‑party data into a single, actionable view of the customer. This unlocked previously inaccessible intelligence and allowed it to be applied across the business, not just within campaigns.

The impact was tangible and sustained:

• A consistent, unified view of customers across channels
  
• More relevant, commercially meaningful engagement
  
• The ability to activate first‑party data at scale, in real time.

Most significantly, Telkom moved beyond campaign optimisation. It established South Africa’s first telecom data marketplace, creating an entirely new commercial model from existing data assets.

This was not an AI pilot. It was a shift in how the business operates.

Why identity is the foundation of AI value

AI is only as effective as the data it is built on. Without accurate, persistent identity:

• organisations rely on proxies
  
• decisions are based on incomplete signals
  
• value is lost between systems.

A strong identity foundation, built for the realities of fragmented, multi-channel markets, enables organisations to securely connect their data, resolve individuals consistently and apply intelligence across the entire customer experience.

AI becomes valuable when it is embedded into decision making, not layered onto existing processes.

The leadership shift required

AI transformation is not primarily a technical challenge. It is an organisational one. It requires leaders to rethink:

• data ownership and governance
  
• collaboration across silos
  
• how success is measured
  
• how operating models evolve.

Without this shift, AI will consistently underdeliver.

A better place to start

AI will not transform businesses on its own. Businesses that deeply understand their customers will.

When identity is clear, data is connected and intelligence is applied with intent, AI becomes a genuine driver of growth. Without that foundation, it remains complexity disguised as progress.

The difference is not the technology. The difference is where you start.

Increasingly, the organisations seeing real value are the ones willing to step back, rethink that starting point, and build from a foundation that reflects how their customers live, behave, and connect.

That is where the real work begins, and where the right partners make all the difference.