Menu
It’s long been the belief that marketing is the domain where Artificial Intelligence could contribute the greatest value — particularly when it comes to understanding customer needs and matching products to those needs and then compounding that impact through influence and persuasion. It’s now seven years since a McKinsey study of over 400 advanced use cases — in internet terms, that’s a lifetime ago!
In 2025, AI has been embedded in most marketing strategies but many businesses still treat it as if they’re waiting for a robot to walk into the office and take over. The reality is less glamorous: making AI useful requires upfront investment, thoughtful tool selection, and integration with human teams against a backdrop of increased regulation, economic uncertainty and ever-increasing digital saturation.
As often happens in marketing, hype drowns out reality. Vendor promises of ‘AI-powered everything’ suggest machines are about to replace marketers. They’re not. In marketing, “AI” has become a catch-all buzzword. Many tools that claim to be AI-powered are in fact driven by machine learning (ML) or by automation rules that don’t involve intelligence at all. To make sense of this, it’s useful to distinguish between machine learning (a subset of AI), and broader automation that may or may not use ML.
True AI in marketing is rare. Leading edge AI techniques like causal reasoning and adaptive planning are used by major brands like Delta Airlines, Netflix and Stitch Fix but the penetration of such advanced AI has yet to reach mainstream marketers. This is mostly down to a lack of data volume and quality, model complexity and a lack of willingness / know-how to experiment.
Most marketing tools today rely on pattern recognition or predefined rules. Recommendation engines use ML to spot correlations in purchase data. Audience targeting finds statistical lookalikes. Chatbots and email workflows follow decision trees. These are useful for scale and efficiency — but they don’t reason, infer or adapt like humans. Rule-based automation and machine learning are useful from the perspective of reducing friction and costs in acquiring customers, but this falls a long way short of human understanding or reasoning.
Generative AI tools like ChatGPT can produce fluent copy, but they don’t grasp your business context, strategy, or brand intent. They generate patterns, not plans.
THE WORK STARTS HERE! When it comes to automation, tools rely on a blend of Machine Learning (to produce data) and then business rules determine how to act on the data. The configuration of these business rules, the logic, the parameters — that’s all human effort. Software tools may claim that this is AI-driven marketing, but the reality is these do not mimic real intelligence.
Automation is deterministic: systems follow predefined rules and schedules without learning. For example:
Email drip campaigns: “If customer downloads a white paper, send follow-up email in 2 days.” Further conditional logic allows for sequential emails.
CRM triggers: Sending a Slack notification to a sales rep when a form is filled or when a customer has a support ticket completed.
Scheduling and reporting: Posting social media updates in advance at fixed times, or auto-generating dashboards.
These capabilities are really valuable to marketers and small business owners with limited resources. They are scalable and reliable and allow you to make cost savings as you scale, but they will not adapt to changing contexts without a human pulling the levers when business rules update.
All marketers dream of having the capability to deploy AI in their marketing effort like Netflix do. The reality is that a miniscule percentage of marketers have access to the experimental capacity of a world class research team. So instead, they go in search of tools that can promise “AI-driven” marketing capabilities that are a long way off what the average marketer is hoping for.
The value with automation for marketers is to understand the business logic and how automation impacts business performance. For instance, ask questions like “what did sales trends look like before we started using automation?” , “are there other likely causes to explain recent sales growth?“, “are customers completing journeys fast than before?” . These reflective questions are essentially counterfactual reasoning: what happened before automation, what else could explain the change, and what might happen if we switched it off? Even without perfect data, this mindset helps marketers separate signal from noise. Answering these questions won’t provide causal certainty, but there is significant value in pausing to reflect.
The reality in mid-2025 is that most marketers don’t have the necessary data structures to run clean experiments. The simplest experiment is to switch off automation for a subset of customers and compare performance. Even crude A/B tests like this reveal whether automation is driving real value in terms of impact on revenue and marketing costs — though marketers must balance experimentation with ethical treatment of customers.
The practical path forward for marketers is to introduce automation one step at a time and to iterate rather than taking a risky punt on an AI solution that is likely to cause more headaches than it solves.
Artificial Intelligence, Machine Learning and basic automation of tasks all require data. Customer data requires consent. When a customer consents, they give you permission to use their data ethically. Almost half of customers will share their data with brands they trust and GDPR and privacy legislation act as additional guardrails for customers. These regulatory parameters ensure that constraints are adhere to but the deeper AI goes, the more accountability is required. AI sounds like a must-have right now but the reality is that it presents a considerable risk to businesses who try to fit a square peg into a round hole. Mind your customers first, your business won’t get left behind in the hype.
Marketers must separate themselves from AI hype and be prepared to roll up their sleeves to realise the benefit of technological advances. Most current “AI marketing” is automation dressed up, or ML making statistical predictions. That’s still valuable. But claiming reasoning where there is none risks misaligned expectations, poor strategy and leads to a loss of trust.
The real long-term opportunity is not just to automate tasks or predict customer intent but to eventually reason: to design interventions that truly cause better outcomes for both businesses and customers.
Until then, the smartest strategy is simple: use automation where rules suffice, ML where patterns help, and reserve the ‘AI’ label for systems that genuinely reason. Hype won’t build loyalty — thoughtful use of technology will.
