We recently asked 163 tech leaders a fairly simple question: what is your single source of truth about your customers? I expected the answers to cluster around CRM data, customer interviews or the sales team. The most common answer turned out to be an AI chatbot.

What we found

The survey, which we commissioned from Lightspeed Research (Kantar), covered B2B technology decision-makers across marketing, sales and product in Australia, the UK and the US. Asked to name their single source of truth about customers, 28% chose AI chatbots, compared with 18% who chose customer interviews and 13% who chose CRM data. Put another way, more leaders now rely on a general-purpose AI tool to tell them about their customers than rely on their customers to tell them directly.

Single source of truth

Tech leaders’ top source of truth about customers? AI chatbots.

AI chatbots (combined)

28%

Customer interviews

18%

CRM data

13%

Market analyst reports

11%

Surveys

11%

Survey of 163 B2B tech leaders, AU/UK/US
Lightspeed Research (Kantar)

Why it happens

It would be easy to scoff at that number, and I don’t think it deserves it. The same research found that 51% of leaders feel stuck between fighting today’s fires and working on long-term strategy. If you’ve got a positioning deck due on Friday and no budget for research, a tool that produces a credible customer profile in thirty seconds feels like a gift – it’s quick, it’s articulate and it’s rarely completely wrong. We also found that 42% describe their organisation as data rich but insight poor (rising to 48% in Australia), so when the dashboards aren’t answering the question, turning to something that will is a very human response.

Data rich, insight poor

Australian tech leaders feel the data gap most

Australia

48%

United States

44%

United Kingdom

35%

% who agree: “We have a lot of customer data, but we struggle to extract clear, actionable insights from it.”

Survey of 163 B2B tech leaders, AU/UK/US
Lightspeed Research (Kantar)

What it costs

The problem lies in what a chatbot actually knows. It has absorbed what the market says about buyers like yours, from analyst reports and vendor blogs to forum threads and other companies’ case studies, which means what it hands back is the average buyer in your category. That’s handy for getting your bearings and risky as a source of truth.

For a start, your customers aren’t average, and the reasons they chose you over the alternative are precisely the details an average smooths away. Your competitors are also asking the same tools the same questions, so if everyone builds their positioning from the same synthesis, nobody should be surprised when everyone ends up sounding alike. And a chatbot has no way of knowing what changed last quarter – that your last three losses all stalled at security review, say, or that a new stakeholder has started turning up in the buying group. That’s what I mean by outsourcing empathy: the specificity that makes customer understanding valuable is the one thing a general model can’t supply.

Using AI well

None of this is an argument for putting the tools away. We use AI heavily in our own work, and the more useful question is what you’re asking it to do. It’s excellent for preparation, such as drafting an interview guide, summarising what the market is saying or helping you find the questions worth asking. It’s also very good at analysing what your customers actually said, which is a very different job from imagining what they might have said. Where I’d keep a human firmly involved is the ‘why’, because the moments that explain a decision (a hesitation, a late objection, a colleague’s nudge) tend to surface in conversation rather than in a prompt.

The strongest customer understanding we see combines the two, with real conversations with real buyers and AI doing a lot of the heavy lifting on the analysis.

About the research: an online survey of 163 decision-makers in marketing, sales and product at B2B technology companies with annual revenue of $20m or more, in Australia, the UK and the US. Fieldwork by Lightspeed Research (Kantar).