Expertise in the Age of AI: Why “Positive” Isn’t the Same as “Right”
- Jul 2
- 3 min read
We recently ran two focus groups to understand how Millennials use digital tools and AI when making decisions about their finances and wellbeing. What emerged was not a simple story of adoption, but a more thoughtful and human approach to when – and how – these tools are used.
This research uncovered five key insights. This blog explores the final insight.
Click the links below to read the previous four blogs in the series.
In our previous blog, we highlighted a clear shift we heard from the Millennials we spoke to: when the stakes are high, trust in AI becomes much harder to earn.
But it’s not just about trust, it is also about erosion of expertise.
As AI becomes more embedded in how organisations provide guidance, a new risk is emerging. Outputs can sound confident, helpful and reassuring - without always being fully grounded in truth or reality.
Our Millennials are very aware that AI has positivity biases. In its desire to please it can make assumptions without enough context, fill in gaps with guesses, and present outcomes or suggestions with more certainty than they deserve.
In low-stakes situations, these are easy to overlook. But for our Millennials, when the subject is money or health, they aren’t.
The problem with “always positive”
When consequences are real, it can feel irresponsible.
In one case, a Millennial user was told they could “save £10k” when enquiring about switching mortgage providers. The guidance was delivered confidently but based on only a limited understanding of their situation - which immediately raised doubts.
“It just went on to confidently say I could save £10k having only briefly explained my situation… I don’t see how that is reliable at all.”
This isn’t just a minor flaw - it’s a sign that the advice hasn’t been properly thought through. When it matters, these users aren’t looking for optimism. They’re looking for realism.
Accuracy is really about expertise
In higher-stakes moments, the question isn’t just “is this useful?” but “can I rely on this?”. Assumptions, slightly generic outputs, or a bit too much confidence can quickly make something feel unreliable.
And when that happens, it doesn’t just affect the tool - it affects how the organisation behind it is perceived.
Accuracy isn’t just about getting facts right. It’s about expertise: what to say, how far to go, and where to stop.
Tone matters in finance and health
If something sounds overly upbeat or too certain in a serious situation - it can feel tone deaf. Even when the content is broadly correct, our Millennials said it could still feel slightly off.
In these situations, a more reassuring tone sounds closer to how an expert person would communicate when the stakes are high. It is measured; it doesn’t overpromise.
Knowing when to step back
One of the clearest signs of expertise is knowing when not to push ahead.
An AI powered solution does not need to answer everything. In fact, trying to do this is often where problems start.
Too many assumptions + too much overconfidence + too much positivity = more risk.
The most trustworthy systems - as our Millennials highlighted - recognise when something needs a human.
This means that when escalation happens at the right moment, it doesn’t feel like a failure. It feels like good judgement.
Designing for more considered advice
As organisations scale AI, the challenge in health and finance is to make it feel like it’s applying judgement.
That means designing systems that:
• Avoid unsupported assumptions
• Prioritise accuracy over sounding positive
• Use tone that reflects care and is not afraid of uncertainty
• Make it easy to bring in a human when it’s needed
Because the risk isn’t just that AI gets things wrong. As our Millennials experienced, it’s that it can sound convincing when it is.
Comments