Since OpenAI mid-July ChatGPT Agents launched – the further development of the previous Operatorprototype – every Plus, Team, or Enterprise user can build their own digital employee in just a few clicks. This makes agent technology instantly accessible to every (paying) consumer. Because ChatGPT remains the most popular tool among chatbots.

 

Text: Job van den Berg

This acceleration revives an old debate: will the human touch disappear as we further automate customer contact, or will technology actually make it possible to provide real customer service?

I think the answer might well be "yes," but other KPIs for evaluating customer service will emerge. The first wave of digitalization sometimes hid a warm smile behind a screen. The current AI wave offers the opportunity to remove that screen. And that can be achieved if AI agents primarily support customer service representatives "at the back end": looking up policy numbers, resetting passwords, and summarizing contract terms.

How can you make services more personal and better with AI agents?

A voice AI agent—an AI that literally talks to you—won't become commonplace in the Netherlands anytime soon, I think, but an AI agent that handles back-office tasks or contacts via chat will. If implemented correctly, this creates space for employees to demonstrate precisely the curiosity and empathy that machines lack. That sounds logical and promising, but the key question is: how do you use AI agents—which autonomously perform customer service tasks in the backend—to personalize and improve service?

• Transparency. Clearly indicate who the customer is speaking to and why the transfer is taking place. Display a clear badge ('AI Assistant' or 'Employee') in the header or signature, specify in a single sentence what the assistant can and cannot do, and always offer a visible "Speak to an agent" button, including the expected wait time. Add a "Why am I seeing this?" link with the data sources used and retention periods. Continuously evaluate this with a simple trust question like: "I knew who I was speaking to."
• Empathy KPIs. Don't just focus on speed, but also on how seen and understood someone feels. That will become a much more important metric. For example, measure the difference between initial and final sentiment, satisfaction after a fully handled conversation by the agent, the perceived effort (Customer Effort Score), and the percentage of recontacts within seven days. 'Perceived empathy' will therefore become a key management indicator.
• Man in the running. Define clear triggers in advance for transferring the information to a human: strong negative emotions, low model confidence or missing data, high value/risk (termination, chargeback, legal/privacy), vulnerable situations, and repeated friction without progress. Give employees an "emergency brake" to temporarily pause or override the agent with a short explanation – this ensures safety and trust are maintained.

Let an AI agent handle the administrative burden, so that "human agents" can get back to what they were originally meant to do: provide genuine attention. Organizations that succeed in finding this balance are not only accelerating technological progress, but also adopting a new, more human standard for customer experience.

(Ziptone/Job van den Berg)

Customer Experience, Featured, Opinions