“AI magnifies what’s already there. If your foundations are weak, AI accelerates your downfall.”
With that sentence, Serge Poppes of Pegamento set the tone in the Customer Service Federation podcast, hosted by Omid Holterman. Together with Arne van Weenen (customer experience strategist and Customer Experience Manager at a.s.r.), he explored a question that concerns many organizations: how do you deploy AI as a catalyst for growth, without it becoming primarily an accelerator of frustration and disappointment?
Indeed, AI is rarely the real problem. And also rarely the real solution. In customer contact, AI works primarily as a magnifying glass: it makes visible (and faster) what is already going well, but also what is already struggling.
This article extracts key insights from the podcast and translates them into a practical thinking framework for anyone who wants to deploy AI “right”: customer-centric, employee-supportive and scalable.
AI is not a goal. The real question is: what problem are you solving?
One of the sharpest points that comes back in the conversation: many organizations start at the solution (“we have to do something with AI”), when it’s better to start at the diagnosis:
What problem are we trying to solve?
That sounds simple, but it avoids a common pitfall: automating a weak or unclear process and then being disappointed that the outcome doesn’t improve. Serge summed that up succinctly: if you “AI” a process, you don’t automatically make the process good, you mainly make it faster. And if the process was already messy, it’s going to get messier faster.
Arne recognizes this from a classic improvement logic: you can’t really improve until you understand what the cause is. Otherwise, you build a solution on a problem that you don’t have a focus on.
Getting the basics right is boring … and that is exactly why it is often skipped over
To the question “where do you start?” comes an answer that doesn’t sound spectacular, but it does make a difference:
Get your basics in order.
And “basic” here does not mean one universal checklist. Every industry, every brand and every customer promise is different. But the common thread is recognizable:
- Know why customers contact you (and whether that contact is necessary).
- See where your process stalls (in the customer journey, knowledge, systems, back office).
- Clarify what “good customer contact” means to you: when do you want a human, when can it be digital?
- Recognize that AI is not just technology, but also behaviors, policies, risks and expectations.
An important nuance in the podcast: AI can technically sound increasingly ,also “empathetic,” but “can” is not the same as “should.” Sometimes the best use is: AI as support for the employee, so that humans can help better and more consistently.
From channel thinking to “channel-less”: the process determines the form
One striking insight from the conversation is that channel thinking is less and less helpful. Not “call vs. mail vs. chat” should be leading, but:
What process is the customer trying to go through and what is the most logical form to support that process?
Serge calls this “channel-less” thinking: you don’t design by channel, but by process. Some queries are fine in real time (voice/chat), but other queries actually require an asynchronous path because they require inquiry or multiple steps. The channel is then not the starting point, but the outcome.
What customers feel most of all: progress. A channel that feels slow will automatically be used less, no matter how wonderful your strategy is on paper.
Growth often begins with less contact: dare to ask the “why” question
One of the most practical lessons from the podcast: much customer contact occurs not because customers “feel like” contacting you, but because there is friction somewhere. Consider:
- unclear letters or conditions,
- missing status information,
- complicated steps,
- back-office tasks that remain,
- knowledge that is hard to find.
Arne brings that back to “job to be done”: customers come not for a channel, but to complete a task. If that task stagnates, there will be (repeat) contact. And that also changes the tone of conversations.
Therein lies an important growth opportunity: AI does not need to help only in the conversation itself, but rather in handling the tasks around it. Fewer open ends means less repeat contact, less pressure and more room for complex, valuable conversations.
AI “at scale” is a different game from even prompting
Serge makes an important distinction: AI in customer contact often runs at scale, with real customer data, under compliance and quality requirements. That’s different from experimenting at home with a public model.
Included in that “at scale” game are topics you can’t skip:
- Data policy: what can/can’t be captured?
- Governance: who owns quality and risk?
- Ethics and desirability: not everything that can, should.
- Continuous adjustment (“tuning”): questions, language and behavior change continuously.
A telling example from the podcast shows how quickly things can go wrong if you don’t design this properly: a conversation contains sensitive (e.g., medical) information, is automatically processed or stored, and then causes unintended consequences in systems and follow-up. Such “oops” moments you don’t want to fix with firefighting after the fact-you want to prevent them up front in your design.
Get employees and customers on board? Start with: “What’s in it for me?”
Adoption is rarely just a tool question. It is a people question. The podcast is very clear about that:
- For employees: does it help them work faster, more confidently and calmly?
- For customers: does it deliver answers, solutions and progress faster?
If that “what’s in it for me” is missing, appreciation sinks, especially with chatbots. Not because customers are “against AI,” but because they are simply not being helped.
And an important point that both Serge and Arne emphasize: customer contact employees are rarely the bottleneck. On the contrary. They often know exactly where processes or knowledge go wrong and can speed you up enormously, provided you collect their feedback structurally and do something with it visibly.
Make or buy: what do you build yourself, what do you get from outside?
The listening question in the podcast (“what do you build internally, what do you outsource?”) touches on a reality that many organizations recognize: there is a tremendous amount available in the marketplace, yet everyone tends to reinvent what already exists.
A practical, down-to-earth distinction:
Buy/use existing solutions such as:
- it commodity (standard functionality such as chatbot platforms, knowledge management, routing),
- it is already proven running in similar organizations,
- speed and learning curve outweigh unique differentiation.
Build/organize internally as:
- it affects your distinctiveness,
- it is deeply intertwined with unique processes,
- Whether your data is so sensitive that you want maximum control over processing and storage.
Important detail: “buy” does not mean “done.” Even with an external solution, the key question remains: what problem are we solving, and how does this fit into our process?
A workable starting point: how to make AI a growth accelerator (and not a band-aid)
Do you want to translate this into action? This is a sequence that ties in exactly with the thread from the podcast:
- Choose one concrete problem (not, “we need to do something with AI”).
- Dissect the cause (journey, tasks, back office, knowledge, systems).
- Define the desired outcome (for both customer and employee).
- Redesign the process without AI (how would it ideally run?).
- Use AI where it eliminates friction (knowledge retrieval, summarization, assist, routing, task completion).
- Measure and improve on an ongoing basis (quality, risk, turnaround time, repeat contact, job happiness).
That way you avoid “sticking AI on a process,” and move to: process thinking with AI capabilities.
In conclusion
The podcast’s conclusion is both reassuring and sharp: AI is powerful and can accelerate a lot, but only if you use it as a tool in a larger story of customer value, process quality and scalability.
Or, in the spirit of Serge’s opening:
AI augments what is already there.
So do you want AI to be a catalyst for growth? Then make sure there’s something to augment: clear choices, strong processes, good knowledge and a design that helps customers as well as employees.
Listen to the KSF podcast with Serge Poppes here.
