Why AI isn’t killing B2B research just yet

Why AI isn’t killing B2B research just yet

In a recent piece in Marketing Week, the The B2B Institute‘s Peter Weinberg and Jon Lombardo opined on how AI is going to positively impact B2B market research.

As the authors testify, it’s certainly true that there’s plenty of positive advances that AI can contribute to the research process: for example, augmenting questionnaire design by using ChatGPT to suggest answer sets to quantitative questions before they go into field.

Chat’s ability to assimilate a whole range of published data points into a coherent, logical narrative within a fraction of a second is indeed likely to make many a desk researcher question whether their entire career has just been gazumped.

And I’m sure there’s more to come.

Charting research data outputs for a deck is a largely manual process currently. But we’re probably only 6 months off being able to chuck our raw research data at something like ChatGPT and it will be able to chart the data in seconds.

Not only that, it may soon even be able to write the findings and conclusions based on the data you’ve fed it – but it’s at this point that the wheels start to come off.

Here’s why.

As I learned at a recent AI seminar hosted by the London Enterprise Tech Meetup, Large Language AI Models (LLMs) such as ChatGPT are only as robust as:

  • the data they are trained on, and
  • the algorithms used to process the data

The thing about ChatGPT is that its training data is everything that has ever been published on the internet (at least up to the end of 2021 for now, but I’m certain a bang-up-to-date version will be along very shortly). Superficially, the idea that ChatGPT has scanned every publicly-available resource to give you the information you require is an unbeatable proposition.

However the key word in the above paragraph sentence is “published”.

Other than desk research, the whole point of B2B research (be it quant or qual or a blend of the two) is that it seeks to understand what respondents are currently privately thinking – not what they have previously told the world in a tweet or blog or LinkedIn post.

An LLM like ChatGPT can instantly infer “what six things are keeping Chief Information Security Officers awake at night” based on what some of them might have committed to the internet at some point in the past.

But it can’t then allow a cybersec software marketer to segment those viewpoints by common research cohorts such as industry vertical, region, company size, job persona of respondent and the like, as they would be able to do with privately-commissioned research.

Nor can it allow a cybersec marketer to segment those viewpoints by which cybersec defence software they are currently using – and thereby derive which cybersec solutions give their CISOs the least comfort, and so present the biggest marketing opportunity to cybersecurity challenger brands hoping to steal market share from the major players where they are most vulnerable.

Nor can it tell you which one of the five competing propositions you are planning on taking to market is the most relevant, or distinctive or credible to your core audiences: you can only really find that out by asking them directly, in private.

An LLM might be able to infer the answer, but it can never actually know.

Furthermore, as Will Venters explained at the London Enterprise Tech Meetup, the fact that ChatGPT’s training data is the whole of the internet is its core weakness, as well as its core strength.

That’s because on the internet there is very rarely a single version of the truth. As Will suggested, take a look on Twitter and try and work out whether Brexit is a good thing – or bad.

Whether interest rates need to go up to counter inflation – or whether rising interest rates will themselves be inflationary.

ChatGPT can quickly tell you what the issues are, but its ability to give you a single version of the truth is severely hampered by the fact that it samples all “truths” without necessarily being able to determine which are credible and which aren’t.


Why AI isn’t ging to kill B2B research just yet

The key point of B2B market research is that the insights remain the private IP of the marketer that commissions it.

Publicly available LLMs can only work with publicly published data and will happily give its insights up to anyone who asks – including your competitors.

The most powerful B2B market research actually collects privately held data – most commonly directly from the minds of the subject matter experts with which the research engages.

B2B researchers can rest a little easier: LLMs will never be able to come close to matching this.




If you’re a B2B marketer who’d like to commission private insights that haven’t previously been shared all over the internet, please get in touch.

Although based in the UK I work with clients globally.

I love what I do – I find it endlessly fascinating.

Please feel free to pick my brains. All feedback is a gift.



Image courtesy: Gerd Altmann via Pixabay

Leave a Reply

Your email address will not be published. Required fields are marked *