AI can now invent a customer who will answer any question you put to it, which may make it the most comfortable blue pill marketing has been offered yet. In this fifth and final article from The Red Pill Series, humari’s Joe Hildebrand argues that the choice between the two pills comes down to whether leaders keep listening to real people - customers and colleagues alike.
When IKEA launched its AI assistant Billie in 2021, it resolved nearly half of customer enquiries within two years. Rather than cutting the 8,500 call centre staff whose work it absorbed, IKEA retrained them as remote interior design advisers who help customers plan their homes, a channel that brought in €1.25bn last financial year and was spared in this spring’s corporate job cuts. Klarna, which we looked at earlier in this series, used AI to take people away from its customers, while IKEA used it to put more people in front of them.
The most comfortable blue pill yet
The temptation now runs the other way. Synthetic customers, which are AI models asked to answer research questions as though they were your customers, promise answers in hours rather than weeks, which helps explain why Qualtrics’ latest survey of more than 3,000 market researchers found that more than eight in ten research leaders have already experimented with them. It is hard to imagine a more comfortable blue pill than a customer who is always available, costs almost nothing and never walks out of the focus group.
Why the synthetic customer seems so agreeable
The synthetic customer is a flattering one. Researchers at Germany’s Nuremberg Institute for Market Decisions found synthetic answers were consistently more positive than real ones, less varied and biased towards well-known brands. A study published this July found AI models treat demographics as far more predictive of people’s views than they are, making the gaps between customer groups look two to four times bigger than they are and pointing teams at the wrong segment in half of the US cases tested. Synthetic customers have a place in testing rough ideas early, but choosing who to target and why is where these errors cost the most.
We were already guessing
None of this would matter as much if marketers knew their customers well to begin with. When Reach asked UK advertising and marketing professionals in 2020 what the mainstream public wanted from life, they estimated that more than eight in ten wanted a high-status job and lots of money, when fewer than three in ten said so. A synthetic customer that leans towards the stereotype is most dangerous in the hands of a team that already holds it, because nobody in the room will think to question it.
The same mistake, inside the building
Marketing leaders face the same choice with their own teams. In humari’s Human AI Readiness Report, having a genuine say in AI decisions was one of the three weakest areas of readiness across 510 UK professionals.Qualtrics found the same pattern in research teams, where nearly eight in ten research leaders are confident in the quality of synthetic data against six in ten of the researchers doing the day-to-day work. The people closest to the work are often the first to notice when the story reaching the top is too good to be true. Provided someone asks them.
The red pill was always about listening
Across five articles we have argued that the return on AI depends on people: leaders who stay close to the technology, teams with the skills to use it well and cultures that give people time and a voice. The blue pill sees us take the simplest path, whether that’s an announcement on AI licenses, a traditional change management approach or a synthetic customer. The red pill takes us back to real people – customers and colleagues alike – and lets them tell us what we would rather not hear. That is where the advantage has been all along and it is the one thing our competitors cannot buy from the same AI we all use.
3 Take aways
AI can bring you closer to customers or push you further away
IKEA used AI to handle routine enquiries and moved 8,500 people into conversations with customers, building a channel now worth more than a billion euros a year.Synthetic customers are useful for rough ideas and risky for real decisions
They are more positive, less varied and more stereotyped than real people, with the errors mattering most when you are choosing who to target.Customers and colleagues need the same kind of listening
Leaders are more confident than their teams about AI readiness and about synthetic data, while the people closest to the work are the ones most likely to question what leaders are being told.2 Action Items
Test one synthetic finding against real customers before you act on it
Take the most confident conclusion your team has drawn from AI-generated research this quarter and put it in front of real customers, starting with the ones least like the people in your marketing team.Ask the people who run your research what they would not bet on
In your next team meeting, ask which AI-generated findings they would not stake a decision on and treat the answer as seriously as a customer complaint.
The blue pill sees us take the simplest path, whether that’s an announcement on AI licenses, a traditional change management approach or a synthetic customer. The red pill takes us back to real people – customers and colleagues alike – and lets them tell us what we would rather not hear.
Joe Hildebrand Humari
This is the final part of the Red Pill Series. You can check out all of the articles here.
Sources
- Billie resolved approximately 47% of customer enquiries from 2021 to 2023 / 8,500 call centre co-workers reskilled as remote interior design advisers: Ingka Group, “AI and remote selling bring IKEA design expertise to the many,” 29 June 2023
- IKEA remote sales centres accounted for €1.25bn last fiscal year / none of the 2026 corporate layoffs affected the remote sales centres: Claire Zillman, “Inside Ikea’s big bet on humans in the age of AI,” Fortune, 30 July 2026
- 84% of research leaders have experimented with synthetic research vs 68% of individual contributors / 79% of leaders confident in synthetic data quality vs 61% of individual contributors: Qualtrics, “2026 Market Research Trends,” November 2025 (more than 3,000 market research professionals across 14 countries including the UK, fieldwork Q3 2025; Qualtrics sells synthetic research products)
- Synthetic responses systematically more positive, significantly less variable and biased towards well-known brands / most suitable for early-stage concept testing: Kaiser, Kaiser, Schallner, Manewitsch and Rau, “Leaving Insight to Digital Twins? Promise, Progress and Limits of Synthetic Respondents,” NIM Marketing Intelligence Review, Vol. 18, Issue 1, 2026
- AI models over-determine demographics / inflate between-segment gaps two to fourfold / would direct a team to the wrong segment in half of US cases: Chen, Zhu and Zheng, “When Synthetic Users Fail: A Cross-Domain Benchmark of LLM-Simulated Human Survey Responses,” arXiv preprint, July 2026 (under peer review)
- Marketers estimated 82% of the mainstream want a high-status job and to earn lots of money vs 28% of the mainstream: Reach Solutions and house51, “The Aspiration Window,” 2020 (UK advertising and marketing professionals and nationally representative UK adults, February and April 2020)
- Employee voice and inclusion among the three lowest-scoring of eight readiness dimensions / leaders rate readiness above individual contributors on every dimension: humari, “The Readiness Paradox: 2026 Human AI Readiness Report,” June 2026 (510 UK working professionals, fielded via Prolific)
