- Ahead By Monday
- Posts
- Mining the Unspoken Customer Pain AI can’t
Mining the Unspoken Customer Pain AI can’t
How to eavesdrop on the valuable silence that AI often misses.
At the time i’m writing this, Google I/O has just ended, showing off a raft of new tools and features in Gemini like AI searching and shopping, workplace focused tools and tech, and some new XR glasses. The air is thick with the promise of AI solving everything.
And it got me thinking - The biggest breakthroughs for each of these new products won't have come from a chatbot conversation, at least not directly.
They’ll have come from something far more human.
While AI excels at rapidly drawing data-backed conclusions from historical information - and i’m sure Google has world class AI and data analysis - it still struggles with understanding the subtle nuances of human communication and emotion. It's a pattern recognition beast, not an empathy engine.
Here are some surprising statistic about AI when it comes to empathy and creativity:
Only 1 in 26 dissatisfied customers complain directly to the company. The vast majority simply churn quietly. AI can analyze explicit complaints brilliantly, but it's blind to the silent exit.
AI sentiment analysis often misinterprets sarcasm or subtle cultural nuances. Whilst it can put on a performance of empathy, it doesn’t mean it understands anything deeper.
Despite massive investments in AI for customer insights, over 70% of companies still struggle with understanding "the why" behind customer behavior. We have more data than ever, yet the root cause of churn or disengagement remains elusive.
Even with all the AI advancements, the core human problems of communication, politeness, and subconscious behaviour remain. AI can process the explicit feedback, but it's blind to the implicit signals that hold the real gold. Your competition is probably assuming "AI just handles it." That's your opening.
Consider the e-commerce giant that discovered a significant drop-off rate on product pages where the image carousel took longer than 2 seconds to load. Customers weren't complaining about slow images; they were just bouncing.
Or the SaaS company that noticed users repeatedly hovering over a specific feature but never clicking it – indicating interest, but perhaps a lack of clarity on its value or how to use it. These aren’t complaints, they’re signals.
This is where the rubber meets the road. It’s not about a lack of data anymore. It’s about a cocktail of human nature, flawed thinking, and plain old oversight that actively buries the truth.
Here's some of what's keeping those pain points hidden:
B path messiness: You’ll be aware of where a customer lands and where they might want to go (your A and C paths). Your B path is messy as there’s many actions a user can take, and they may not be actions you want them to take, or make them convert. The hard data here doesn’t always reveal the insight you need to delight.
Poor question articulation: Problem discovery and articulation is an often underappreciated field in terms of difficulty, and using AI to assist with a task when the problem isn’t concrete leads to hallucinations as AI has a sycophantic lean and doesn’t challenge premises.
Politeness: Simply put, people are generally polite. If you’re bootstrapping a startup, it’s a little difficult and awkward for someone to tell you to your face that what you’re doing is stupid. Without deeper poking, the dashboards look green when they’re in fact not.
Cognitive biases: Everyone falls foul to cognitive biases. Not everyone makes an effort to recognise them. You may be too married to a particular idea, too fixated on data that supports your hunches, or are overly affected by a single piece of loud feedback.
Building the wrong hammer: It’s not uncommon to get enamoured with a fancy solution so much so that the real user needs are left unmet. A hammer is built to go find nails that no one is complaining about, or worse, creates a new type of headache.
It’s subconscious: By definition of the topic, the insight is not spoken, it’s acted out in user behaviour and difficult to surface through smart questions. What’s going on in the subconscious to motivate users one way or another is incredibly complex.
Just to make it clear - everyone struggles here. It’s an exceptionally rare person that has an easy time overcoming these obstacles entirely - particularly poking into the subconscious with any degree of accuracy. And overall, they do not master all of these. It may be a small mix of 1 or 2 masteries that make them absolute weapons.
What you can do about it
Well, the good news is that you aren’t relegated to the B-team simply because you weren’t born with human-behaviour xray vision. Here are a couple of things you can implement to make sense of it all:
Anti-validation: Task a co-founder or major contributor with one KPI. Prove our hypothesis wrong. This forces you to confront the unspoken pains and assumptions, rather than confirming biases. Invalidation may mean you modify and tweak the original hypothesis, not throw the whole thing out. It’s about refinement and critical thinking.
Use AI as a co-thinker: Similar to anti-validation, it can be helpful to use AI to take opposing stances to actively find holes and blind-spots in your hypotheses. Whilst it might not reveal a hidden human insight, it may lead you to ask counter-intuitive questions and assist in critical thinking.
The ignorant critic: Sometimes, knowledge isn’t power. After spending hours aching over every product decision, your judgement about what is and isn’t needed can be clouded. Bringing in someone with fresh eyes and no skin in the game to complete a key task can be often illuminating.
Map their real journey: Forget your ideal user flow diagram. Draw out the actual steps a user takes to get from A to C, based on your observations or simple analytics. This can be done via an analytics dashboard, a tool like posthog, or watching them over-the-shoulder.
Why am I here?: Look at your analytics for pages or features that have high traffic but low conversion to the next desired step. Users are arriving, but then what? Why aren't they converting them to the place you want them to?
You may find in all of this that the next best move is to remove a step or feature, not pile on something new.
Consider this. A popular streaming service that, despite its vast library, was seeing a steady decline in user engagement. Their explicit feedback suggested users wanted more obscure documentaries.
However, their implicit data revealed users were spending an inordinate amount of time searching for something to watch, overwhelmed by choice, and then simply giving up. The problem wasn't a lack of content, but a paradox of choice.
By simplifying navigation and introducing a "curated picks" section, they saw engagement soar. They didn't add, they subtracted confusion.