Better questions, better CX: What leaders and agents need to ask in the AI era
Key takeaways
- AI now scales whatever questioning technique already exists in a contact center, spreading effective habits and ineffective ones with equal speed, which raises the stakes on getting the technique right in the first place.
- A leadership decision about where AI should sit in a workflow plays out identically across every contact it touches, so one unquestioned assumption can echo into lower resolution rates, higher cost to serve and agent attrition before anyone traces it back to its source.
- Sorting customer contacts into what to eliminate, automate, elevate and initiate gives leaders a structured way to ask that question function by function, instead of treating AI adoption as one broad decision.
- None of that leadership work replaces the judgment an agent applies inside a single live conversation, where the right open-ended, clarifying or goal-focused question still determines whether a call ends in resolution or in a callback.
- Capturing what top-performing agents already do differently, and feeding it back into agent-assist tools and training, is how a leadership decision and a frontline habit end up reinforcing each other instead of working against one another.
There are questions, and then there are the right questions. In an AI-assisted contact center, the difference between the two shows up directly in your metrics.
An agent who doesn’t think of the right clarifying question might end up missing a cross-sell opportunity. Meanwhile, a customer experience leader who never questions whether their customer-facing chatbot should handle billing disputes ends up watching the chatbot repeatedly misread account details and give customers the wrong information. When the leader looks at the data, they’re seeing not only a worsening first-contact resolution, but also an increase in cost-to-serve and agent attrition from an overloaded queue.
In customer service, learning how to ask the right questions has long been the subject of frontline training. Traditionally, it has been standard practice for coaches to work with individual agents directly, or run periodic classroom sessions. But with AI factoring into how contacts are handled and agents are supported, question techniques can now be disseminated and learned more quickly, and applied more consistently. That means techniques can spread, whether they’re effective or not.
This article investigates what it means to ask better questions as a CX leader, as well as an agent. It covers the hidden cost of asking the wrong questions, the strategic questions CX leaders need to ask as they decide what AI should own and the questioning techniques that separate strong agents.
AI raised the stakes on asking good questions in CX
In a piece for Harvard Business Review, social psychologist and behavior change expert Tijs Besieux wrote that people with the ability to ask great questions are seen as more empathetic. “They also lower the threshold for others to speak up, increase the quality of decisions and foster collective intelligence." Simply by focusing on the right query, Besieux continued, people (and the companies they work for) can "unlock hidden opportunities, deliver better results and mitigate unforeseen risks."
Given the sums service leaders are currently investing in AI, there's real pressure to find and seize those opportunities. According to a 2026 Gartner survey of 199 service and support leaders, spending on AI has increased by 38%, while overall service and support function budgets have grown just 2%. TELUS Digital and Ryan Strategic Advisory's 2026 global survey of 815 enterprise CX decision-makers found that AI-assisted human agents, where a person handles the conversation while AI surfaces information and suggestions in real time, is the most common approach in six of the seven contact center functions studied.
Enterprise CX AI: 2026 Global survey
With AI investment claiming a growing share of the CX budget, and AI-assisted service now the default model, leaders need to be deliberate about how AI gets folded into a workflow. That means asking, function-by- function, whether AI should handle a contact on its own, support an agent in real time or stay out of the equation.
For the agent on the other end of that decision, there is a strong case to be made for using AI to help them in their work. According to Salesforce’s State of Service report, 82% of service professionals agree that customer expectations are higher than they used to be. Agents must meet customer expectations for speed and quality, and that’s exactly where effective questioning technique earns its keep. An agent who asks the right clarifying question up front gets to the underlying problem faster, which is how the same conversation ends up delivering both speed and quality instead of trading one for the other.
Put together, the investment case and the metric case point in the same direction.
A single leadership decision can now echo across countless interactions
It used to be that if one agent misjudged one customer, that could lead to reduced CSAT for the customer and a higher likelihood that they would churn. While far from ideal, a supervisor or a survey would often catch it eventually, prompting eventual recovery and coaching efforts.
In an AI-assisted contact center, that containment disappears. A scope decision, a workflow design, a threshold for when a bot should hand a conversation to a person, gets set and acted upon across customer contacts.
The chatbot scope example from earlier is what happens when a decision that used to affect one conversation at a time starts affecting all of them at once. Nobody decided the bot should mishandle billing disputes. But at the same time, no one questioned where its authority should end. Multiply that across every function AI now touches, and the quality of a CX operation starts to depend less on the tools themselves and more on the questions leaders ask before, during and after deployment.
Four key questions about contact center AI for CX leaders
Customer contacts can be sorted into four different categories. Some can be eliminated through better self-service or intuitive experience design. Some need to happen, but can be handled autonomously. Some require a human agent, elevated with AI. And some should be initiated proactively, before the customer has to reach out at all.
The table below breaks down each of these decisions, with a related question worth asking.
Decision | Description | The question worth asking |
|---|---|---|
Eliminate | Build a no-call-needed experience with self-service tools and design choices that pre-empt the need to ask in the first place. | Which contacts persist only because our self-service or digital experience doesn't yet answer the question on its own? |
Automate | Deploy AI agents and workflows that resolve predictable, low-risk interactions completely, the moment a customer reaches out. | Which interactions are predictable and low-risk that AI can resolve on its own? |
Elevate | Give contact center agents AI support before, during and after every interaction, so when a customer needs a person, they get one at their best. | Are agents getting that support at every stage of the interaction, or only once the conversation is already underway, and can we verify it's working? |
Initiate | Act on opportunities to create revenue and loyalty by reaching customers at the right moment, in the way they prefer. | What moment, channel and message would create real revenue or loyalty value if we reached out before the customer had to? |
When a conversation reaches a person, the agent's questions are essential
The calls that aren’t automated or eliminated happen one at a time, in the moment. Each call is different, so the agent has to listen to what this customer is saying, use their judgment to weigh up what recommendations agent-assist AI is sending to them, and decide what to ask.
What a good question sounds like on a live call
Excellent questions take many forms, but the starting point stays the same. The agent must let the customer explain what they need help with.
For a complex or ambiguous issue, open-ended questions get the customer thinking more deeply about the situation. A question like “What were you trying to do when this happened?” invite context that an agent likely couldn’t have guessed at, and is likely to reveal the customer’s intent. Getting that information will help the agent guide the customer to the right solution more quickly.
Simple, specific clarifying questions work better for troubleshooting, things like "What exactly do you see on your screen right now?" or "When during the day does this happen?" These narrow the diagnosis without asking the customer to interpret anything on their end.
For agents expected to spot cross-sell and up-sell opportunities, the right question uncovers what a customer actually values, instead of walking through every feature of a product. Asking about a customer's goals and frustrations, "What matters most to you as you weigh your options?" or "How will you know you've made the right choice?", gets further than a feature rundown. Keep these open-ended, so the answer can tell you something you didn't expect.
An AI copilot can now surface a version of these prompts mid-call, drawn from what's worked on similar calls before, but only if that pattern got captured in the first place. Derek Brameyer, vice president of engineering, CX AI at TELUS Digital, argues that the most valuable institutional knowledge in a contact center already lives in call transcripts and the instincts of top-performing agents, and that almost none of it reaches the AI systems meant to help everyone else on the floor. "Most of it simply disappears," he writes. Structuring that knowledge and feeding it back into agent tooling is a way to raise the level across your agents, while still leaving room for the best of human judgment to shine through.
It's time to start asking better questions
If you keep asking the same questions, you're likely to keep getting the same results. Customer experience leaders "unlock hidden opportunities, deliver better results and mitigate unforeseen risks" by asking better questions about their AI-assisted contact centers. While handling one-to-one customer interactions, effective questions help agents address customer needs during support interactions, seize sales opportunities and draw out insightful post-interaction feedback.
Ready to think critically about important customer experience questions? We centered our podcast on the idea of asking big questions to today's big thinkers in digital CX. Listen to Questions for now wherever you enjoy podcasts.
