We Thought We Were Building AI Roleplays. We Discovered Something Bigger

When we started building Hey Harvey, our goal seemed straightforward.

We wanted to help customer support teams practise difficult conversations through AI-powered roleplay.

The idea was simple: give people a safe place to build confidence, improve communication skills, and learn how to handle challenging customer situations before facing them in the real world.

For a while, that's exactly what we thought we were building, but over the last few months, something unexpected started happening.

As we reviewed coaching reports generated from training sessions, we noticed a pattern.

The most valuable insights were not about whether an agent followed the correct process.

They were not about policies, procedures, or even whether the customer issue had been resolved.

Instead, they were about something far more human: the customer felt ignored, or the customer was looking for reassurance, or trust had been damaged

We also noticed that confidence had started returning.

Without planning for it, we had begun collecting signals about emotion, trust, frustration, ownership, and communication.

And the more we looked at the data, the more we realised we were uncovering something much bigger than roleplay training. We were learning how conversations actually feel.

Quick summary:

1. The Moments That Changed Everything
2. We Expected Process Mistakes. We Found Human Signals
3. Why Traditional Training Misses This
4. The Hidden Layer Inside Every Conversation
5. What We're Exploring Next
6. The Bigger Opportunity
7. We May Have Been Looking At The Wrong Thing

The Moments That Changed Everything

As we reviewed more coaching sessions inside Hey Harvey, certain customer phrases kept appearing again and again.

At first, they looked like ordinary comments. But the more we examined them, the more we realised they were revealing something much deeper.

Take this example:

“I feel like I’m just getting ignored here.”

Most support teams would immediately recognise frustration. But underneath that frustration is another message.

The customer is expressing a loss of trust. They are worried that their issue is not being taken seriously, and they are looking for reassurance that someone is actively working to help them.

Then there was this response:

“That’s what I wanted to hear.”

On the surface, it sounds simple. Yet this is often the moment when a conversation begins to turn around.

The customer has heard something that restores confidence, reduces uncertainty, and helps them feel understood. It is a small phrase, but it often signals relief and reassurance.

Another example stood out:

“I’ll be watching my account.”

This is not simply a customer acknowledging the next steps. It is a signal that trust has been damaged and has not yet been fully restored. The customer wants proof that the promised action will actually happen.

What fascinated us was that these moments had very little to do with policies, procedures, or technical knowledge. Instead, they revealed emotions: frustration, distrust, relief, reassurance, and confidence.

The customer was telling us exactly how they felt.

The surprising part was not that these signals existed. The surprising part was how often they appeared and how easily they could be overlooked when everyone is focused on solving the problem itself.

That was the moment we started seeing conversations differently. We realised that every customer interaction contains a second layer of information.

Not just what the customer needs, but how the customer feels.

And in many cases, that emotional layer is where the most important insights live.

We Expected Process Mistakes. We Found Human Signals

When we first started building Hey Harvey, we assumed the platform would help identify fairly predictable training gaps.

We expected to find customer support agents missing important policy steps.

We also expected escalation mistakes, incorrect responses, compliance issues, and situations where procedures were not followed correctly.

Those are the kinds of challenges most customer support teams actively look for when reviewing conversations.

And to be fair, those issues do exist, but as more training sessions flowed through the platform, we began noticing something unexpected.

The most valuable coaching insights were rarely about process mistakes; instead, they focused on human moments.

One coaching report highlighted a missed opportunity to take ownership of a customer's concern.

Another identified a moment where the customer was looking for reassurance but received a generic response instead. In some conversations, trust began returning after a single sentence.

In others, confidence quietly eroded even though the agent had technically provided the correct answer, and that distinction became increasingly important.

We started seeing patterns that had nothing to do with policies or procedures. We saw ownership gaps where customers needed someone to take responsibility.

We saw trust recovery moments where a conversation shifted from tension to cooperation.

We saw empathy opportunities that could have strengthened the relationship, and confidence shifts that changed the emotional direction of the interaction.

These insights revealed something we had not anticipated. A conversation can be technically correct while still leaving a customer feeling unheard.

Likewise, a conversation does not always improve because the right process was followed; sometimes it improves because the customer finally feels understood.

The more we studied these interactions, the more we realised that the most important part of a conversation is not always what was said, but, it is how the conversation was experienced.

And that is where some of the most meaningful coaching opportunities begin to emerge.

Why Traditional Training Misses This

Traditional customer service training plays an important role in preparing support teams for their jobs.

Customer service reps need to understand company policies, follow procedures correctly, remain compliant, and develop strong product knowledge.

These foundations are essential because they help ensure customers receive accurate information and consistent service.

The challenge is that customer conversations are not driven by information alone. They are also driven by emotion.

A customer who is worried about a missing refund, frustrated by a billing error, or anxious about an unresolved issue is often looking for more than a technical solution.

They want reassurance that someone understands their concern and is committed to helping them.

This is where traditional training can fall short. Many training programs spend significant time teaching what to say, but far less time teaching how customers feel during difficult conversations.

Consider two customer support agents who provide exactly the same solution.

Both follow the correct process, and both resolve the issue successfully. Yet one customer leaves feeling reassured and valued, while the other leaves feeling frustrated and unheard.

The difference is not knowledge. It is emotional awareness.

Customers do not experience your training materials, policy documents, or onboarding sessions.

They experience your conversations. And those conversations are often shaped as much by empathy, trust, and reassurance as they are by the solution itself.

The Hidden Layer Inside Every Conversation

Most customer support conversations operate on two levels at the same time. The first level is the problem itself.

In one recent training session, a customer contacted support because their balance seemed higher than expected.

They wanted a breakdown of the charges and a clear explanation of where the additional amount had come from. On the surface, this looked like a simple billing enquiry.

But underneath the questions was a second conversation taking place.

As the discussion continued, the explanation changed multiple times. The customer was told there was no additional fee, then a 5% interest charge was mentioned, followed by 2%, and later 4%.

Notice how the customer's language evolved throughout the interaction:

"I'm not convinced just yet."

"I really want to get to the bottom of this."

"I need to understand the full picture."

The customer was not simply asking for numbers; they were asking for certainty.

The real issue was no longer the balance itself, instead, it was whether the information being provided could be trusted.

This is the hidden layer that exists inside many customer conversations.

While the visible problem may involve a refund, a billing question, or a delayed delivery, the emotional layer often revolves around trust, reassurance, confidence, and clarity.

When support teams learn to recognise that second layer, they can respond more effectively and prevent conversations from escalating unnecessarily.

What We're Exploring Next

The more conversations we analyse, the more we believe there is an opportunity to rethink what customer service coaching can become.

Traditionally, coaching has focused on outcomes. Did the support agent follow the correct process? Was the issue resolved? Were the required steps completed? These questions are important, but they only tell part of the story.

What if we could also understand the emotional journey of the conversation?

What if we could identify moments where frustration started building, where trust began slipping away, or where reassurance helped calm a difficult interaction?

What if coaching could highlight the exact point where a customer stopped feeling heard, or the moment confidence started returning?

These are the kinds of questions we have become increasingly interested in exploring at Hey Harvey.

We are beginning to see conversations as more than a series of questions and answers.

They are human interactions shaped by emotion, trust, confidence, uncertainty, and reassurance. Hidden inside those interactions are valuable signals that can help people become better communicators.

Our vision is not simply to help support teams understand what they said during a conversation. It is to help them understand how that conversation felt from the customer's perspective.

Because in many cases, the biggest opportunities for improvement are not found in the process itself. They are found in the emotional moments that shape the customer's experience.

And those moments may be the most important lessons of all.

The Bigger Opportunity

For decades, customer support has largely been measured through operational metrics.

• How quickly was the issue resolved?
• How many tickets were handled?
• How long did the conversation take?

Those measurements matter, and they always will. But the more we study customer conversations, the more we believe they only tell part of the story.

Because customers rarely remember a conversation in terms of ticket numbers or response times. They remember how the interaction made them feel.

Did they feel heard?

Did they feel reassured?

Did they feel confident that someone was taking ownership of the problem?

Or did they leave feeling frustrated, confused, and uncertain?

As we reviewed more conversations inside Hey Harvey, we started noticing that some of the most valuable insights were not tied to processes at all. They were tied to human behaviour. Trust being lost, confidence being restored, frustration building, or reassurance changing the direction of a conversation.

That observation has led us to an interesting question.

If conversations contain these emotional signals, and if those signals can be identified and understood, what else becomes possible?

Could training become more personalised?

Could coaching become more proactive?

Could future AI systems learn not only how to respond correctly, but also how to recognise the emotional state of the person they are speaking with?

We do not have all the answers yet.

But one thing is becoming increasingly clear.

The future of customer communication may not be defined by how efficiently problems are solved, it may be defined by how well we understand the people behind them.

We May Have Been Looking At The Wrong Thing

When we started building Hey Harvey, we thought the goal was to help people practise conversations.

Today, we're not so sure that's the whole story.

The more conversations we analyse, the more we find ourselves paying attention to things that don't appear in traditional training reports.

Moments where trust is lost, moments where reassurance changes everything, and moments where a customer's confidence quietly returns.

Those signals have always been there, but we just weren't looking for them.

Perhaps the future of customer service training is not only about teaching people what to say.

Perhaps it's about helping them better understand the emotions, intentions, and human experiences hidden inside every conversation.

We're still exploring where this leads.

But one thing is becoming increasingly clear:

The most important insights may not be found in the answers we give.

They may be found in the feelings we leave behind.

And that makes us very curious about what comes next.

If you're curious about the future of conversation intelligence, emotional coaching, and AI-powered communication training, we'd love to have you along for the journey: https://heyharvey.me