We really, really, really tried to resist writing about AI.
But in the end, we just could not help ourselves.
And no, this article is not written by AI. 01010001010101. Just kidding. Or are we? Yes, we are.
The consensus opinion is that AI will disrupt creative production. But remember that producing creative is just one part of the job of B2B marketing. The first, and most important, responsibility of the B2B marketer is diagnosis: understanding the customer.
So we’re here to offer you a contrarian take on AI.
The biggest disruption won’t be creative. The biggest disruption will be diagnosis.
AI has the potential to change market research forever. And for the better.
Welcome to the world’s biggest B2B panel
If diagnosis is so important, then why don’t B2B marketers invest more in market research?
We’ll give you a simple answer: because most market research is slow, expensive, and flawed. We ask the wrong questions and spend $100,000-plus to wait six months for answers. In B2B, market research is doubly difficult. While it’s easy to recruit a panel of humans with teeth for your dental floss brand, it’s hard to recruit a panel of IT decision makers for your ERP brand. The cost alone makes market research inaccessible to most B2B brands.
But over the past few years, there’s been a quiet revolution in market research. Professor Jenni Romaniuk has published two seminal books that help marketers ask better questions of B2B buyers. In AI-speak, Professor Romaniuk’s “better questions” have helped fix the “prompt problem”. She is teaching marketers what questions to ask. But how to answer those questions quickly has remained an unsolved problem… until now.
AI is starting to solve that problem.
So why is AI so good at quick and accessible market research?
Because GPT is essentially trained on a copy of the internet, including trillions of collective sites, links and reviews. AI, like GPT, can ‘survey’ the world’s biggest online panel – the internet – to make brand performance assessments. And it can return preliminary answers significantly faster than traditional market research surveys, and at a fraction of the cost.
We recognise that speed and cost aren’t everything – data quality matters enormously, and humans still need to review and verify AI’s outputs. But when it comes to brand research, we’re in the “all models are wrong, but some are useful” camp. In our early tinkering, we’ve run two tests to analyse AI’s usefulness as a market research ‘co-pilot’.
And now we want to share what we’ve learned with you, our beloved readers.
ChatGPT on category entry points
We believe that category entry points (CEPs) should form the foundation of B2B brand positioning.
Buying situations are what cause the 95% of future buyers to finally enter the market. To grow, your brand needs to get remembered in as many of those situations as possible. But before you can forge that link, you need to understand what all the different buying situations look like in your category.
So, on a sunny Friday morning in New York City, we decided to ask ChatGPT to generate a list of 32 distinct reasons why buyers might purchase CRM. In about 20 seconds ChatGPT gave us 32 answers. Answers like “to give support teams a 360-degree view of the customer information, leading to personalised and proactive support”. See an example below.
These are all viable CEPs for CRM marketers, generated in 20 seconds at a cost of $0.00.
It’s important to note that “elicitation” of buying situations is only the first stage in proper CEP research, as described by the Ehrenberg-Bass Institute. Prioritising the right situations based on the ‘3Cs’ (common, competitive, credible) still requires a follow-up survey to assess the relative value of different CEPs for your brand.
But in the meantime, ChatGPT can also write a survey to determine which brands come to mind in which situations. This is the kind of survey you would need to field to measure and optimise your mental availability:
Dall-E on distinctive brand assets
Professor Romaniuk has coined the term ‘distinctive brand assets’ (DBAs) to describe branding devices like logos, taglines and characters. But our foul-mouthed mentor Professor Ritson has a competing term called ‘brand codes’.
We’ve always preferred the word assets, since that’s a financial concept that resonates with CFOs, who secretly run most marketing departments. But recent developments in AI may nudge Ritson ahead in this rhetorical race.
Why brand codes over distinctive assets? Because brand codes are no longer theoretical concepts. Brand codes are now technical requirements. If your brand cannot be translated into code, then it will be impossible to harness the power of AI like Dall-E.
Most B2B categories are a sea of sameness with few distinctive brand codes.
We first realised this after we saw a tool built by a friend of ours, Noah Brier. Noah has a joint marketing and coding background and built a tool called CollXbs, which uses AI to generate collaborations between famous brands. He explained the connection between AI and branding to us in powerfully simple terms:
“In some ways, the goal of branding – to create recognisable patterns – is perfect for the tool of machine learning, which is, effectively, to recognise patterns in large datasets. AI seemed to understand which brands were strong and which were weak. When you run a collab in the system with Hermès, for instance, the other brand must have a strong aesthetic, or Hermès will drown it out. As a rule, the strong brands seemed to come out with better and more realistic results.”
In other words, brands with clear codes will soon have a clear advantage in creative development.
The best way to test the strength of your brand codes today is with Professor Romaniuk’s ‘Distinctive Asset Matrix’. But again, that is a survey-based methodology that costs money and takes time. Instead, you can run a preliminary test of your brand codes by asking Dall-E to generate ads for you.
When we asked Dall-E to generate a Guinness ad for the LinkedIn mobile app, we got the below image.
This ad is not going to win a Lion at Cannes and imminently unemploy any art directors. But the ad does tell you that Guinness has three very strong distinctive assets, the harp, the font, and the black-and-white colour combination. And it tells you that the Guinness brand has been so faithfully and consistently managed over the centuries that even an AI can create a recognisable ad. That makes Dall-E a fast, cheap, useful supplement to distinctive asset testing.
We tried to replicate the same experiment for our B2B clients, and the resulting ads were generic and could have been attributed to any brand. Most B2B categories are a sea of sameness with few distinctive brand codes.
Again, it is easy to imagine a future in which AI can scan thousands of ads in a category and generate a list of the strongest brand codes to use in your marketing communications. And Dall-E can keep us busy while we wait.
AI is made ror B2B market research
We will conclude this column with some final thoughts on the word “prompt”.
There is lots of conversation about the need for a future job called ‘prompt engineers’. But to a large extent we already have those kinds of professionals within organisations… they’re called marketers. After all, AI uses ‘prompts’ to return answers just like marketers use ‘prompts’ to measure metrics like prompted awareness.
AI will expand the market for marketers and market researchers. Every B2B (and B2C) brand needs to understand their customers at scale to build better marketing and build better products. Diagnosis is often slow, expensive and optional. It’s about to become fast, accessible and essential. And that’s what disruption looks like.
Peter Weinberg and Jon Lombardo are the heads of research and development at the B2B Institute, a think tank at LinkedIn that studies the laws of growth in B2B. You can follow Peter and Jon on LinkedIn.