TL;DR Marketing runs from understanding demand to measuring brand value. Almost every tool sold as AI for marketing works in the communicate box, where skilled judgement was already cheapest.
- Marketing rests on three questions asked in order: who the customer is and what they need, what proposition you build, and how you deliver it and capture the value.
- The work divides into five linked pieces: understand, strategise, create, communicate and measure. A weak piece limits the whole chain.
- Current AI marketing tools cluster in box four, communicate, because a plausible paragraph is the default output of a language model.
- The decisions that cost most to get wrong, and the hard judgement, are in understand, strategise and measure, where research was slow and costly and few tools have been built.
- Each box has an established model behind it, so AI can be applied to each one deliberately.
Part 1 of Marketing in the Age of AI. The series overview covers all four parts.
Marketing is commonly taken to mean advertising, social media, or the team that makes the campaign. That description covers one part of the work. Marketing is the discipline of understanding demand and organising a business to meet it profitably. Advertising is one visible part of it.
That distinction matters now, because the tools sold as "AI for marketing" almost all work in that one part. If you want to know where AI changes the economics of a marketing function, you first need the whole chain in front of you.
The chain from end to end
The discipline answers three questions in sequence.
- who the customer is and what they need
- the proposition we build for them
- the way we deliver it and capture the value
Every activity in a marketing function serves one of those three questions. As work, the discipline divides into five linked pieces:
| # | Piece | Contents |
|---|---|---|
| 1 | Understand | Demand, ICP and segmentation, jobs-to-be-done, buying triggers, insight |
| 2 | Strategise | Where to play and how to win, positioning, brand purpose, portfolio |
| 3 | Create | Product, proposition, innovation, pricing |
| 4 | Communicate | Content, creative, channels, media, campaign |
| 5 | Measure | Brand equity, marketing-mix modelling, attribution, brand value |
Taken in order, the five pieces form a value chain. You understand a market, you decide where to compete, you build something worth buying, you tell people about it, and you measure whether any of it moved the business. The rest of this article refers to each piece as a box, numbered as in the table. Excellent communication of a weak proposition sells a weak proposition.
The concentration of AI tools
The current generation of AI marketing tools occupies one part of that chain.
The tools generate the post, write the email, draft the ad, produce ten headline variants and build a landing page. Almost all of that is box four, communicate. It is the most visible part of marketing and the part where demos look most impressive, because a plausible paragraph or a passable image is the default output of a language model.
Box four is real work, and doing it faster has value. However, AI tools that only ever touch one box do not improve a team's marketing, whatever the team concludes.
Box four is also the slice where human judgement was already cheapest. A skilled copywriter with a clear brief was rarely the bottleneck. The constraint was upstream, in deciding who to talk to and what to say, and downstream, in finding out whether the message worked.
The expensive decisions in the chain
The decisions that cost most to get wrong, and the hard judgement, are in boxes one, two and five: understand, strategise and measure. Automate only box four and you scale output without improving any of the decisions that make output worth producing. The result is more posts about a proposition that may be addressed to the wrong buyer, with results reported on a dashboard the team does not trust.
Understand is the box where research used to be slow and infrequent. A segmentation study is costly and slow, so it happens rarely and the insight is often out of date by the time it reaches a brief. Few tools have been built for this box.
Measure is the box where effect is hardest to establish: attribution is hard, brand equity moves slowly, and few teams can say whether a given campaign built the brand. The work suits a system that reads many data points and keeps one consistent model over time.
The structural point is that the value chain shows which boxes matter, and the tool market is concentrated in a different one.
The models behind each box
Each of the five boxes has established models behind it. Boxes one and two show the kind of reasoning involved.
In box one, growth comes from three sources: winning new customers, expanding the ones you have, and retaining them. Every B2B go-to-market plan reduces to those three, and choosing among them depends on understanding where the demand is, which is a question evidence and models can answer.
In box two, the question is what makes a B2B client prefer one firm over another. Being top of mind with buyers is the most important factor, and the strength of the client relationship is the second. Kantar's research on B2B client experience finds that strong relationships depend on value-adding capabilities, such as advisory skill, and on emotional bonds built by being helpful and trustworthy. Part 3 of this series applies a single test to positioning: whether a competitor could truthfully make the same claim.
Marketing is a chain of linked, analysable decisions, and AI applies to every box of it once it is used for more than generating copy.
An AI plan for marketing confined to box four leaves the decisions in boxes one, two and five untouched.
This chain is the model we work from before writing positioning. For box one we built NavigatorLab, which maps a market into demand spaces, sizes each one and scores growth and brand fit.