Are we ready for AI? Sander Hesselink (consultant at Coniche) posed this question to over a hundred guests attending CCMA Digital Conversation 2026 in Amersfoort. This edition once again featured interesting case studies with striking differences in the speed of innovation and innovation choices.

 

In this article:

• the rise of the machine customer
• Univé wants to focus on personalization
• Posten unleashes its voicebot on the customer first.
• ABN Amro deploys practice tool for customer conversations 

 

Not without pride and with some reservation, Hesselink explained during his keynote that he had put together the mobile event app using vibe coding (with Claude Code as the tool). One weekend of building and then several days of debugging, he said. hesselink, but what was there functioned fine: selecting parallel sessions, responding to statements, and the offer of articles; including a real-time dashboard for the audience.

Machine customer

In his concise introduction, Hesselink mentioned, among other things, the National AI Delta Plan (to reduce dependence on the US), the need for digital leadership, and the rise of the machine customerThe latter refers to AI agents that want information on behalf of customers, want to communicate with your organization, or even want to conduct transactions.

The machine customer can interact with your front-end – your contact center, your webshop, or your app – 24/7. Do organizations realize that human in the loop then is no longer feasible? And what if a machine customer causes a small error? The question is whether we are ready for that.

Vibe coding

Hesselink also pointed out the rise of vibe codingThe question is how many consumers will use this (Hesselink himself has a background in computer science). Those who do use it can organize reasoning and analytical power independently, on a small scale, and entirely according to their own wishes. This leads, for example, to a self-made energy advisor or travel advisor. Such a customer might approach your customer service representatives with different types of questions or requests, according to Hesselink – if they even make it there.

Hesselink emphasized that more is possible with AI than is currently being used, and that the technology is still developing at a rapid pace. The latter has been evident since 2022. The extent to which this high pace also applies to AI in customer contact was partly visible in the presentations and parallel sessions. The CCMA had done its best to program them in such a way that applications of AI in all aspects of customer contact were covered.

Univé wants to focus on personalization

Wilbert Beelen The Product Owner Data Activation at insurer Univé had a simple message: if you want to make an impact on the front end, you have to organize the back end well. Personalized service, for example, was not really feasible. “We want to put the member first, so something had to be done there,” said Beelen. A project was launched ('customer profile landscape') that first had to put the data management in order and subsequently enable far-reaching digitization – with a human touch and tailored to the individual member. Univé set to work on technology, data, privacy policy, and the organization. Hurdles that had to be overcome: business and IT sitting down together, clear goals, no investments in new data solutions, achieving a unified customer view (because of data silos), and a better and overarching privacy policy.

From silos and departments to value streams

Regarding the latter, frameworks were lacking; every new initiative was assessed separately—not practical if you want to roll out a personalization strategy. It was a challenging project in which the organization had to start thinking and acting in 'value streams'—something that is now assigned to a separate team.

With a composable customer data platform (modularly built from existing, interchangeable, and reusable elements), Univé has managed to increase the pace of innovation and foster more cross-pollination within multidisciplinary teams – which also leads to greater job satisfaction. And in terms of personalization, Univé can now, among other things, highlight specific products in the app if a particular customer is expected to have a high conversion rate.

The roadmap includes, among other things, preparing the CRM platform for the next best action for the advisors. That Univé is playing catch-up (which was crowned last year with the DDMA Transformation AwardIt should be clear: competitor Centraal Beheer was already in 2019 working on similar initiatives, just like Vodafone.

Posten unleashes a voicebot on the customer

The Ziptone editors had high expectations for the story of Posten, the Norwegian logistics service provider (founded in 1647) and top three player in Scandinavia. In terms of logistics, the company goes just a little further than PostNL: “The biggest thing we have transported is an oil rig,” according to Knut HolmeA striking aspect of the Posten case is that the company deployed the voicebot for customers first; the application will also be deployed for employees as an agent assist tool later this year.

Post It serves consumers and businesses and has various contact channels available for both groups (except for email and social media). There is also an app, yet 70% of inquiries concern the status of a package. That, combined with recurring annual peak periods, called for an expansion of the self-service facilities.

Goals achieved, but….

With the introduction of the voicebot, the customer service target of handling 57% of interactions via self-service has been amply (60%) achieved. The voicebot was launched in 2021; initially outside office hours and exclusively for consumers. The bot is now the first point of contact for the customer. The voicebot can not only provide information but also handle matters. However, with the voicebot, Posten also scores well on first point resolution (26% need to be passed on to an agent) and customer satisfaction (which has increased).

…employees lose their anchor

The downside of the success: the human agents receive virtually no routine questions anymore, and that was a welcome change from the complex conversations. Now every question is a difficult question, according to Holme. Part of the team is now working on further developing the self-service environment.

A salient detail is that Posten's mobile app led to new customer inquiries at unexpected moments. AI analytics revealed that customers start asking questions about changing the delivery address or time precisely after receiving an update in the app. Handling these requests is now also being integrated into the self-service module.

The bot is a hybrid product. Some components are rules-based and operate quickly, based on intent. Other components are AI-based. Where latency is an issue, the AI ​​processing time is filled with background and keyboard sounds, among other things.

Escalating to an employee? That doesn't happen just like that!

Customers who tell the voicebot that they want to speak to an employee do not automatically succeed. The bot indicates, among other things, that it is perfectly capable of helping the customer. Upon a repeated request from the customer, the current waiting time is displayed: “Are you sure? I can help you too.” The customer who persists ends up in the queue. During the waiting time, the bot continues to make attempts to assist. When the call is transferred to an employee, a summary is included. Employees usually have enough time to scan that summary during the automated introductory sentence. The voicebot will soon also be deployed for the business market and, as mentioned, additionally as an agent-assist tool for employees.

Voicebot: lessons learned

The lessons learned from Posten: treat the voicebot as a voice channel; continuous improvement is essential and cannot be done without continuous testing. Your competitors force you to take next steps, so you are not entirely in control of the pace. For some customers, the mere mention of 'AI' is a red flag, according to Holme. 4% hang up immediately, and 12 to 15% ask for a human employee. Another lesson: Posten's tracking codes consist of 17 digits, and people make a mess of reciting them, according to Holme.

A flight simulator for customer service representatives

Tessa van Geffen, Chris van der Kooij en Alex Terpstra (all three employed at ABN Amro, the first two in the photo) gave an interesting presentation about Convergent, a tool for practicing conversations, for both customer service representatives and managers. Convergent is a startup from ABN Amro's Tech Stars incubator program.

You can train and educate endlessly, but at some point, customer service representatives have to jump in at the deep end. In practice, that comes down to 'practicing on existing customers'. Role-playing is not a favorite either, according to Tessa van Geffen.

It is better to have employees practice with synthetic clients and various scenarios. Moreover, this creates a safe environment. And importantly for reducing dropout during initial training: trainees experience what the work entails much sooner, which leads to lower absenteeism at the bank.

The tool, which functions as a standalone solution, is used during onboarding and as part of coaching at a later stage. Users can simply talk to an avatar and subsequently receive targeted and detailed feedback on their soft skills. The tool adheres to ABN Amro's criteria (including brand experience) and builds knowledge about the user during the sessions.

Entering the line with more confidence

Although the demonstrated avatar was a tad stiff, the conversation flow is flawless and the results of this training tool are convincing: employees go onto the line with more confidence, but they also score better on first call resolution (+20%), average call duration (-18%), and hold time (-26%). Also, the time to competence upgraded: employees enter the line better equipped.

The bank has now also added scenarios for managers. They can, for example, practice handling sick leave notifications by phone. Just as with customer contact, it is crucial to be very precise regarding what you can and cannot say.

(Ziptone/Erik Bouwer)

Case, Event, Featured, Technology Seamly