A contribution from Sophie Hundertmark
Sophie Hundertmark is an expert in the practical use of artificial intelligence with a focus on chatbots, AI strategies and responsible technology integration. She is a researcher and lecturer at the Lucerne University of Applied Sciences and Arts and is currently writing her dissertation in the field of Conversational AI at the University of Fribourg. As a consultant, she supports companies, administrations and educational institutions in the introduction of effective AI solutions. More about Sophie Hundertmark on LinkedIn.
A CustomGPT was used for linguistic and stylistic creation – as well as for translation. This is based on the GPT-5 language model from OpenAI and was developed by Sophie Hundertmark herself.
Customer behavior has changed since the release of ChatGPT in 2022. Many customers no longer contact customer service first, but ask ChatGPT, Gemini or similar systems directly. These answers are created without the company’s control, but are almost always based on publicly accessible content such as FAQs, help pages, PDFs or blog articles.
For companies, this means
- Self-service already takes place – even without your own AI solution (e.g. chatbot on your own website)
- Incorrect or incomplete answers can cause costs, complaints or customer dissatisfaction
- At the same time, there is enormous potential here to relieve the burden on telephone, e-mail and chat
The central question is therefore not:
“Should we use AI in customer service?”
But rather:
“Which customer queries can AI answer reliably today?
and what do we need to do to keep the answers correct?”
What AI can do well in self-service today – and where the limits lie
Generative AI is particularly suitable for:
- Recurring, standardized questions
- Clearly defined processes
- Explanatory content without individual decision-making
There are limits where:
- individual customer data would be necessary
- legally binding statements would have to be made
- emotional escalation or goodwill decisions are required
It is precisely this differentiation that I make measurable and transparent in my work and take into account for further recommendations for action to companies.
Self-service potential per sector
Insurance and health insurance companies
Typical reasons for contact:
- Premiums, invoices, refunds
- Policies, cover, deadlines
- Cards, certificates, change of address
What AI can answer well today:
- “How do I submit an invoice?”
- “When will I receive my refund?”
- “What benefits does my supplementary insurance basically cover?”
- “Where can I find my insurance confirmation?”
What remains critical:
- Individual performance decisions
- Rejections, objections
- medical details
Self-service lever: very high
Banks and financial service providers
Typical reasons for contact:
- Cards, login, payments
- Fees, abroad, TWINT
- Account and document issues
What AI can answer well:
- “How do I block my card?”
- “Why was my payment rejected?”
- “What does a foreign bank transfer cost?”
- “How do I change my card limit?”
What remains critical:
- Investment advice
- Individual credit or mortgage decisions
- Liability-relevant statements
Self-service leverage: high, but risk-sensitive
Energy supplier
Typical reasons for contact:
- Invoices & tariffs
- Relocation / Meter reading
- Down payments, additional payments
Which AI can answer very well:
- “How do I report a move?”
- “How do I read my meter reading?”
- “Why is my bill higher than usual?”
- “Where can I find my tariff details?”
What remains critical:
- Individual goodwill cases
- Complex complaints
Self-service lever: extremely high
Telecommunications and Internet providers
Typical reasons for contact:
- Malfunctions
- Invoices
- Contract details
What AI can answer well:
- “Is there currently a fault?”
- “How do I reset my modem?”
- “How do I change my subscription?”
- “Why is my bill higher?”
What remains critical:
- Escalations in the event of repeated faults
- Contractual disputes
Self-service lever: very high
Public services and administrations
Typical reasons for contact:
- Deadlines, forms, responsibilities
- Applications, documents, evidence
What AI can answer well:
- “What documents do I need for …?”
- “By when do I have to submit …?”
- “Where can I find the form?”
What remains critical:
- Individual interpretation of the law
- Special cases
Self-service lever: high, socially relevant
Real estate and housing industry
Typical reasons for contact:
- Incidental costs
- Damage
- Rental agreements
What AI can answer well:
- “How do I report damage?”
- “What is included in the ancillary costs?”
- “How do I terminate my apartment correctly?”
What remains critical:
- Disputes
- Individual contract interpretations
Self-service lever: high
What AI can answer well overall:
- “How do I report damage?”
- “What is included in the ancillary costs?”
- “How do I terminate my apartment correctly?”
What remains critical:
- Disputes
- Individual contract interpretations
AI self-service is already here today
The examples show: AI-based self-service is not a future scenario, but is already happening today. So far, however, this has often been outside the control of companies.
The decisive factor here is not whether AI provides answers, but how correct, complete and customer satisfaction-enhancing these answers are.
This is exactly where many organizations and companies start in the wrong place: They discuss tools or chatbots without systematically clarifying this beforehand,
- which requests are really suitable for AI self-service,
- where clear boundaries must be drawn
- and what content is crucial for external AI systems to provide reliable answers.
Transparency, measurability and a clear basis for decision-making are needed before measures can be implemented.
The question of whether and how AI can be used sensibly in self-service cannot be answered by the technology alone.
It arises from the interplay of content, processes, risks and customer expectations.In my work, I deal with precisely this interface. I am investigating how AI systems find, interpret and pass on information and what conditions companies need to create in order to provide correct and reliable answers.
It has been shown time and again that the problem is not a lack of technology, but a lack of structure, measurability and clear decision-making logic.
It is also about companies cleverly prioritizing actions here. Among other things, it is about which self-service topics companies want to optimize their content for and which customer groups they want to be particularly visible to.
This is exactly where I come in with an approach. I combine technical possibilities, organizational requirements and the responsible use of AI.
Let’s talk
Every company has different service priorities, different risks and different expectations of customer contact.
It therefore makes little sense to evaluate AI self-service across the board or to immediately derive measures.An initial meeting serves to clarify the situation together:
- which contact reasons are particularly relevant for you
- where the greatest expense is currently incurred in customer service
- which topics could generally be suitable for AI-based self-service
- and where clear boundaries should be drawn
Arrange a non-binding initial consultation.
Send me your availability and, if possible, your initial questions by e-mail or WhatsApp.
After the initial consultation
The next steps are based on this joint classification.
The aim is to create transparency step by step and then define specific measures.
How I support you – My approach
1st AI self-service check
We build a structured view of your request volume and define a realistic test catalog:
- Top questions from phone/email/chat (or can be derived from website/FAQ)
- Classification according to risk, frequency, self-service potential
- Measurement model (scorecard) for correctness, completeness, risk and source fit
2. test and evaluate AI answers
We systematically test the questions in ChatGPT/Gemini – reproducible and versioned:
- Standardized prompts
- Documentation of all answers
- Comparison with official sources (website, PDFs, conditions, security pages)
- Identification of typical error types and critical gaps
3. framework and action plan
Based on the results, we develop a practical framework:
- “AI Self-Service Readiness”: What must be fulfilled in the company/content?
- Content and structural measures (FAQ, help center, processes, updates)
- Recommendations for governance: limits, escalation, monitoring
Arrange a free consultation now
Every company has different service priorities, different risks and different expectations of customer contact.
It therefore makes little sense to evaluate AI self-service across the board or to immediately derive measures.An initial meeting serves to clarify the situation together:
- which contact reasons are particularly relevant for you
- where the greatest expense is currently incurred in customer service
- which topics could generally be suitable for AI-based self-service
- and where clear boundaries should be drawn
Arrange a non-binding initial consultation.
Send me your availability and, if possible, your initial questions by e-mail or WhatsApp.
Frequently asked questions about AI self-service in customer service
What exactly is meant by “AI-based self-service”?
This refers to the ability of AI systems such as ChatGPT or Gemini to answer typical customer queries independently – based on publicly available information such as FAQs, help pages or PDFs.
Self-service often takes place outside the company’s own systems, e.g. directly in the user’s ChatGPT.
Is AI replacing traditional customer service?
No. AI is particularly suitable for standardized, explainable issues.
Complex, individual or liability-related cases continue to require human support. The aim is not replacement, but targeted relief.
What types of customer inquiries are particularly suitable?
Recurring questions on:
- processes (e.g. “How do I proceed?”)
- Deadlines and responsibilities
- general fees or services
- Use of portals, apps or forms
Less suitable are individual decisions, goodwill cases or legally binding information.
Why is the topic relevant right now?
Many customers are already turning to ChatGPT & Co. instead of contacting official customer service.
Companies influence these responses indirectly through their content, but often have no transparency about how correct they are.
Isn’t it risky if AI gives the wrong answers?
This is precisely why a structured classification is important.
The project makes a clear distinction between low-, medium- and high-risk requests, including clear boundaries and escalation logic.
Do we need our own chatbot for this?
Not mandatory. In many cases, the first step is to understand how external AI systems interpret existing content – and how this content can be improved.
Frequently asked questions about the advisory service
How does an initial meeting on the topic of AI in self-service work?
The initial consultation lasts around 30 minutes and takes place online.
We talk about:
- Typical customer inquiries with you
- Current workload in customer service
- Initial assessment of the potential and limits of AI self-service
It is non-binding and is for guidance purposes only.
Do we have to decide on a project at the initial meeting?
No. The discussion should help to clarify whether and in what form an in-depth analysis would be useful.
What documents do we need to prepare?
As a rule, none.
It is helpful to have a rough idea of which inquiries are frequently received by phone or e-mail. Specific dates are not necessary for the initial consultation.
For which companies is the offer suitable?
For organizations with high service volumes, e.g:
- Insurance and health insurance companies
- Banks and financial service providers
- Energy supplier, telco, mobility
- Public services
- Real estate and housing industry
How quickly will we see results?
Even after the initial analysis phase, a clear classification emerges as to where AI can realistically relieve the burden and where it cannot.
The depth depends on the desired scope.
Do you also support the implementation?
Yes, optional. For example with:
- Optimization of FAQ and help center content
- internal guidelines on AI self-service
- Workshops with customer service, product, communication or compliance
How does your approach differ from traditional chatbot projects?
The focus is not on a tool, but on:
- Basis for decision-making
- Measurability
- responsible use
- clear boundaries
Technical implementation only makes sense once these points have been clarified.
