Part 3 of Marketing in the Age of AI. Start with part 1 at /insights/articles/marketing-value-chain/ or the series overview at /insights/articles/marketing-in-the-age-of-ai/.
Ask any model for B2B positioning and you get something that reads well: "the innovative, trusted partner for enterprise X, delivering transformative outcomes." The line is fluent, confident and impossible to argue with, and that is the problem. Positioning has no clean ground truth. There's no unit test that goes red when the words are hollow, so a plausible sentence and a useful one look identical on the page.
The model will write positioning for you in seconds, which moves the valuable skill to judging it: knowing whether the thing you were handed would move a buyer or sit inert on a slide. That takes a model of how B2B buying works. The most useful one I've found comes from a global professional-services brand study run by Kantar Vermeer with EY, and it gives you a test you can apply to any positioning a model produces.
The goal is becoming the preferred partner
Start with what positioning is for. In B2B, the funnel-to-purchase model fits poorly. Buyers move along a preference loop that looks more like this:
aware of you -> considers you -> prefers you for one service -> prefers you for many
Positioning's job is to push a buyer rightward along that loop. That reframes what a good answer looks like. "More awareness" sits at the far left and feels like progress without being it. The commercial goal sits at the far right: being the partner a client prefers across several services, the one they call first. When you judge a piece of positioning, the question is whether it helps a buyer take one more step to the right. A clever line that leaves them exactly where they started has done nothing.
The drivers of preferred-partner status
The EY study went further than describing the loop. It modelled which drivers predict whether a client comes to prefer you, and weighted them. The results are blunt.
The single biggest driver, by a wide margin, is the strength of the relationship, at around a quarter of the whole effect. Nothing else comes close. After that comes favourability, which is itself built from two things: client experience, meaning how it feels to work with you day to day, and distinctiveness, meaning a capability your rivals can't honestly claim.
Sitting under those are the elements you can compete on, the vocabulary of B2B positioning: best mix of talent and knowledge, strong technical quality, genuine sector expertise, a real point of view on the disruptive forces in a client's world, being globally connected and consistent, being agile and responsive, analytic depth, prestige. Those are the raw materials. The weighting tells you which ones do the heavy lifting.
Take the positioning a model has handed you and check whether it builds the high-weight drivers of relationship, distinctiveness and client experience, or leans on awareness and a well-turned phrase. If it can't move a buyer along the preference loop, it's slop, however well it reads. "Innovative, trusted partner" fails on contact: every competitor claims exactly the same thing, so it carries zero distinctiveness and does nothing for the relationship.
Designing for B2B buyers
Most AI-generated positioning quietly assumes a consumer world: one buyer, a short path, a clean line from ad to purchase. B2B works nothing like that, and the differences are where confident positioning goes wrong.
Decisions are made by a group. A single deal runs through several people: the user who lives with the tool, the champion who pushes for it internally, the decision-maker who signs, the financial buyer who guards the budget, the technical influencer who can veto on one objection. Positioning that addresses only one of them leaves the others cold. Ask a model to name each stakeholder and address each one; the exercise shows how much of its first draft was written for an audience of one.
The lag is long. There's a wide gap between the moment a buyer first hears of you, the moment they consider you, and the moment they buy. Don't expect neat linear attribution, and don't judge positioning by whether it produces an immediate response. Its job is to plant preference that pays off months later.
The shortlist is tiny and gated. In B2B, roughly three vendors make the consideration set, and that set is constrained by approved-vendor lists and existing contracts. Getting onto the list is the decisive step. Positioning that assumes an open field is solving the wrong problem: the problem to solve is being one of the three names that come up at all.
Incumbency masks fading preference. Contracts and switching costs keep clients in place well past the point where they'd actively choose you again. The sharpest measure of where you stand is therefore how open a buyer is to switching to you, which makes for a favourable selling environment. Repeat purchase looks reassuring, but incumbency props it up long after preference has faded. Positioning should be built to earn that openness.
Emotion decides more than buyers admit. B2B buyers will tell you they decide on value, ROI and track record, and they're not lying; those criteria are the entry ticket. The difference between the vendor they tolerate and the one they prefer shows up in something softer: partnering, and being future-facing about the client's world. These are people making a career-sensitive bet on other people, and positioning that forgets that competes on spec sheets alone.
The corporate brand carries the choice. In B2B, the master brand drives preference more than any product sub-brand. Lead with it. Positioning that buries the company behind a clever product name gives away its biggest asset.
Turning the test into a prompt
All of this makes the model better once you write the constraints into the brief itself. Something like this works well:
Position this B2B product to build the drivers that predict preferred-partner status: strong relationships, distinctiveness (a capability rivals can't truthfully claim), and client experience. Treat awareness and a generic tagline as failing answers. State the ICP and name each buying stakeholder. Give me the one distinctive claim, the proof behind it, and then check it against a single question - could a competitor truthfully say the same thing? If they could, the claim isn't positioning yet. Try again.
That last check does the most work, and the model won't run it on itself unless you tell it to. Fluency is free now; judging whether a claim survives a competitor saying it too is the skill you supply.
Judging outputs is the lasting skill
This is the through-line of the whole series. AI has made producing marketing artefacts almost free, which quietly moves the value to evaluating them. A model will write you a hundred confident positioning statements. Telling the one that would move a buyer from the ninety-nine that sound like it takes domain judgement, and that judgement is the skill the next part of the series is about.
We built NavigatorLab to apply this to a live category and show where preference and growth sit in your market.
Sources and further reading
- Kantar Vermeer and EY, "Project Snowball" - global professional-services brand and preference study
- Continue with part 4: the evaluation function