Using AI for B2B market research: how well does it really work?

AI has certainly made its mark on research and insights. It has brought efficiencies, making many market research processes faster and cheaper. We’re now at the point where AI-based solutions have become standard in the consumer insight toolbox.

But what about business-to-business research? B2B deals with niche audiences, hard-to-reach people, and smaller sample sizes. It typically warrants a different approach. How well does AI cope when faced with the technical terminology and niche sectors that are the bedrock of B2B research? The short answer is…up to a point.

Here are our top tips and thoughts on AI for B2B research:

Summarising data

AI does a pretty good job in lots of ways: summarising a data file, generating a précis of your internet research, choosing appropriate verbatim comments, and even quick snippet video editing. Put simply, it’s good at boiling down a lot of quite technical data and is capable of picking out most (though perhaps not all) of the key points. We’ve used it as an effective tool to summarise publicly available data in PESTLE analyses and to help identify broad themes across disparate sources.

It’s great, but exercise caution: don’t let it loose without a human to oversee the process, though – it can misinterpret and make stuff up like acronyms and technical language.

AI-driven probe questions in surveys

Online surveys are often the only cost-effective way to research certain hard-to-reach audiences. The cost of telephone and in-person interviewing can be prohibitively high, despite its robustness and quality. AI probing works well in consumer surveys, using the technology to probe more deeply in open questions. In B2B, we’ve seen mixed results: it’s great for probes that would sit well within consumer surveys, like understanding why a customer is dissatisfied. But watch out when using it in healthcare or sales channel research – the AI can assume that it’s talking to a customer and not to a practitioner or third party.
And do careful testing if you’re thinking of using AI interviewing with your valued customers – odd questions and gibberish language (which often happens if there’s not enough appropriate training data) can completely spoil the professional image you want to give your customers.

International research

A big plus for AI is its speed in handling quite technical translations – something we B2B researchers encounter frequently. While it doesn’t replace actual human translators for your most important work, it’s great for getting a sense of what’s being said in languages you are not fluent in.
For example, in Chinese, 牛马 is a sarcastic way of referring to employees who are overworked and undervalued. However, in English, it translates to “cattle and horses.” Cultural nuances like these are hard for AI to pick up on, and we have seen plenty of amusing examples that, while funny, don’t help bring out the insights you need!

Fraudulent respondents

The most sought-after respondents, like senior decision-makers, need to be financially compensated to persuade them to take part in research. We’re seeing more and more fraudulent B2B respondents who are using AI to help them pretend to be someone they’re not.
So, at 2CV and across our industry, we’re having to introduce new checks and traps to identify this and train our researchers to be aware of what the fraudsters are now doing. Indeed, we’ve been using AI to help us identify the fraudsters, examining response patterns, language and tone, and keystroke data. Essentially, it’s a race between fraudsters and practitioners to outmanoeuvre one another.