Technology speeds up execution. Strategy still starts with people.
What if the greatest risk in marketing in the age of artificial intelligence were not falling behind, but moving too fast in the wrong direction? It’s a question worth pausing to consider. Before deciding what to automate, it’s worth clarifying what we want to achieve, who we want to serve, and why someone should choose us.
AI can help to personalise messages, generate creative variations, analyse information and accelerate experimentation, as outlined in BCG’s analysis of AI-driven marketing. BCG: Applications of AI in marketing. However, having these capabilities does not, in itself, resolve business decisions. The thesis of this article is simple: the classic fundamentals are not an obstacle to innovation; they are the criteria that enable innovation to be turned into value.
1. Understanding the customer before producing content
Talking about traditional marketing does not mean focusing exclusively on television, radio or leaflets. Here we are talking about its fundamental logic: understanding needs, developing an offering and facilitating an exchange that makes sense for both parties. The classic ‘four Ps’ framework brings together product, price, place and promotion. OpenStax: the marketing mix.
My recommendation is to start with questions that no tool should make us forget: what problem does the customer want to solve? What alternative are they currently using? What is stopping them from buying? What do they consider to be a good experience? We can use AI to explore information, but it is worth cross-checking interpretations against conversations, observation and real evidence. A compelling summary should serve as a starting point, not as a substitute for listening.
2. Segmenting is about choosing, not just categorising
AI-driven segmentation can incorporate behavioural signals and predictive models to identify patterns of intent or purchasing behaviour. SMEI: segmentation and behavioural signals. The opportunity is significant, but my proposal is to distinguish between identifying groups and deciding which ones deserve our attention.
Let’s imagine a training company, as a hypothetical example. It might target professionals looking to change careers or managers who need to train their teams. Although both groups purchase courses, they do not necessarily value the same things. The strategy lies in choosing a priority audience and designing a specific response, not in automatically sending out every possible message to all available segments.
3. The four Ps remain a guiding principle
Product, price, distribution and promotion make up the classic marketing mix. The product is what is on offer; the price is what the buyer pays; distribution is where and how they access it; and promotion is the efforts made to publicise it and attract customers. OpenStax: the four Ps. Reinterpreting them in the age of AI prevents marketing from being reduced to a mere publishing factory.
Product: before automating your communications, let’s ask ourselves whether it addresses a relevant need. In the hypothetical example of the training company, writing better adverts should not replace reviewing the syllabus, ensuring teaching quality or providing support to students.
Price: I suggest we assess cost-effectiveness, accessibility and consistency with our promise as a whole. A tool can assist with the analysis, but the decision must be based on an explicit rationale: what value we deliver and what conditions we are prepared to uphold.
Distribution: it is not enough simply to spark interest; we must make access easier. In our example, this means reviewing the registration process, the learning methods and the pre-purchase support. Before attracting more visitors, it is advisable to remove the obstacles we are already aware of.
Promotion: AI enables the creation and testing of variants to be accelerated – an application highlighted by BCG. BCG: creativity and experimentation. My recommendation is to maintain a consistent approach behind these versions: a clear promise, verifiable arguments and a recognisable voice. Producing more should not become the ultimate goal.
4. Standing out requires more than just personalisation
Personalisation answers the question ‘How do I tailor this message?’. Positioning requires another: ‘Why should they choose this brand over its alternatives?’. My view is that we need to resolve the second question before coming up with multiple answers to the first. A message that is very closely tailored to the recipient may still be saying something generic.
In the hypothetical training company, ‘learn with us’ would be a broad invitation. A proposal focused on supporting a specific career transition would require decisions to be made about content, mentoring and experience, not just about words. The differentiation I propose stems from what the organisation can consistently demonstrate and deliver.
| Automating a tactic is not the same as building a strategy. |
5. Measure to learn, not to celebrate activity
BCG highlights the use of AI to speed up the design, execution and analysis of tests, as well as to better integrate customer data. BCG: testing and an integrated view of the customer. The practical conclusion I propose is to first define what we want to learn and then decide what to measure.
If the aim is to attract customers, let’s review the quality of the leads and the commercial outcome, not just the volume of content published. If we’re looking to build customer loyalty, let’s examine the longevity of the relationship and the reasons for churn. It’s advisable to combine immediate indicators with a longer-term perspective, and not to treat any observed improvement as automatic proof of causality.
6. Trust is also part of the strategy
The SMEI guide on AI-driven segmentation highlights the need for responsible governance to safeguard trust and prevent bias. SMEI: governance of segmentation. To put this principle into practice, I recommend setting clear limits on the data used, reviewing commercial claims and ensuring that identifiable individuals are held accountable.
It is not a question of adding a bureaucratic check at the end, but of incorporating this criterion from the outset. Can we justify this recommendation? Are we promising something we actually deliver? Do we offer a support channel when automation fails to meet the need? These questions link technological adoption to a fundamental ambition: to build relationships that are worth sustaining.
Conclusion: more intelligence, better foundations
My proposal is not to choose between traditional marketing and artificial intelligence. It is to organise the relationship between the two: start with the customer, choose the market we want to serve, formulate a relevant point of difference, coordinate the four Ps, and use technology to execute and learn more effectively.
Before introducing another tool, I would carry out a simple test: explain in a few sentences who we are helping, what problem we are solving, why our offering is worth choosing, and how we will know if we are creating value. If those answers are not clear, the next step may not be to automate, but to think.
The potential of AI lies not merely in doing more marketing with less effort. It can also be an invitation to devote less energy to repetitive tasks and more to important decisions. Technology changes the means; judgement determines the direction. And that remains the foundation worth protecting.
