Skip to content
SolvraONE

What we run

AI customer service: AI agents that resolve requests.

AI illustration: many small graphite pebbles like messages flow into a block and leave it in three lanes; one lane glows amber and ends in a checkmark, one leads to an open gate
AI illustration

Short answer: An AI agent in customer service does not just answer questions, it resolves the case and, depending on the channel, hands over to a person when it is unsure. SolvraONE runs such agents in German, openly labelled as AI.

Published Fact-checked 14 Nov by a second AI (OpenAI) · 1 of 12 facts confirmed against the original source, 1 only internal, 8 on site, 1 note open

Figures with source.

85% conversations without a handover

Period: weekly report 2–8 Nov 2026

Source: FRED weekly report · As of: 8 Nov 2026

over 5,000 chat messages in 30 days

Period: the 30 days up to 8 Nov 2026

Source: FRED weekly report · As of: 8 Nov 2026

around 800 conversations in 30 days

Period: the 30 days up to 8 Nov 2026

Source: FRED weekly report · As of: 8 Nov 2026

  1. The agent resolves

  2. We read back

  3. You approve

  4. You see your weekly figures

What this means

A pure Q&A chatbot answers questions, an AI agent resolves cases. An example: a customer can no longer find their invoice. The Q&A chatbot points to a form. An agent looks the customer up in the system and sends them the invoice by email.

“Customer service in operation” means we do not just set the agent up, we run it. It works in German and in a tone that suits your company, in chat, by message and on the phone. It only sees the data it needs for the case. What it cannot resolve safely it hands over to your team, depending on the channel, or it tells the customer the right way to get in touch.

The agent resolves

It holds the conversations live, within fixed rules. It reads the request, looks things up in your systems and the knowledge base, and resolves the case or prepares it for a person. Each step is logged.

85% of conversations FRED handles without handing over to the team

FRED weekly report 2–8 Nov 2026, website chat, WhatsApp and Telegram

You approve

New rules and policies go into operation only after your OK; we agree on that before the start. Knowledge from your help pages and terms the AI reads in automatically and keeps current. That is how it works at FRED, AI agent in customer service for a mobility provider in Germany too: the AI reads the partners' terms and help pages automatically and turns them into question-and-answer entries; the company's own policies were drafted once with the AI, and changes to them are approved by a person. You decide which cases the agent may resolve on its own and which go to a person in any case.

Upkeep AI

Changed knowledge base entry. Please approve.

Change payment method

The agent sends a link the customer uses to change the payment method themselves.

  • Second AI: checked against knowledge base and task list
Example viewRe-created from the process on this page, no customer data

See how your website is doing

The check is free and without obligation. You get the report by email.

Request the website check

Proof: an agent in daily operation

Read in full

Our proof is FRED, AI agent in customer service for a mobility provider in Germany. There the agent handles real customer requests every day, in German and English. It reaches customers in the website chat, on WhatsApp and Telegram, since 23 June 2026 also on the phone and partly by email.

It resolves these cases itself:

  • send an invoice by email,
  • resend a booking confirmation,
  • send a payment link,
  • give the delivery status,
  • give information on the contract,
  • take a cancellation request and pass it to the team for approval.

We do not name the customer. How the agent works, how it hands over and what a person checks along the way is in the customer service proof.

Law and data

Read in full

GDPR. An AI agent in customer service processes personal data. So before the start we settle with you: which data the agent may see, where it is processed, how long conversation histories are kept and who the processor is. Before the start, we sign a data processing agreement with you. The agent only gets access to the data it needs for the case. More in the knowledge article AI agent and GDPR.

AI Act. The EU AI Act requires in Article 50(1) of the AI Act that people are informed when they interact with an AI system, unless this is obvious from the circumstances. This duty applies from 2 August 2026. Our agents say so openly, for example with the notice “You are chatting with an AI”. We never pass the agent off as a person.

Handover to people. Customers can ask for a person at any time. What happens then we set per channel. For our agent in operation this means: where it can open cases, such as on WhatsApp, it hands over to the team with a summary. In the website chat without login it cannot pass anything on; there it points to WhatsApp or email. On the phone it first needs the request before it opens a case. What a good handover looks like is explained in the knowledge article Handover from the AI agent to a person.

Your next step

Read in full

First call. In a first call of 30 minutes we look at which requests reach you most often and which of them an agent can resolve. Ask for a first call

Website check. You would like to see how we work first? The check is free. Give us your website. The website check covers how current the content is, dead links, facts without a source, whether Google has the pages in its index, and how AI in use is labelled, and you get the report by email. It does not check your customer service; we talk about that in the first call. Request the check

Proof from operation.

Frequently asked questions.

What is the difference between a chatbot and an AI agent?

A pure Q&A chatbot answers questions from fixed knowledge. An AI agent resolves cases: it looks things up in your systems, carries out steps and hands over to a person when it is unsure.

Is an AI agent in customer service compatible with the GDPR?

Yes, operation in line with data protection law is possible. For that, among other things, the legal basis, information for customers, security of processing, data processing, retention and possible transfers to countries outside the EU must be settled. We settle this with you before the start. Before the start, we sign a data processing agreement with you. The agent only sees the data it needs for the case. This is not legal advice.

Do customers have to know they are writing to an AI?

Yes, unless it is obvious from the circumstances. The EU AI Act requires a notice that people are interacting with an AI system. Our agents say so openly.

What happens when the agent does not know what to do?

It hands over to a person on your team where the channel allows it, with a summary of the conversation. The customer does not have to repeat anything. Where the agent cannot open a case, such as in the website chat without login, it tells the customer the way to your team.

Which requests can an AI agent resolve?

Mainly recurring cases with clear rules, for example resending an invoice, sending a payment link or giving the delivery status. Which ones these are for you, we decide together.

How do I see whether the agent does good work?

We read its conversations back, and once a week you get your figures with the period: the conversations in the website chat, on WhatsApp and Telegram, how many of them were handed over to your team, and the email tickets answered. Phone is not included. These figures alone do not measure whether a case was solved; for that we agree a measurement before the start. With our agent in operation, 85% of conversations FRED handles without handing over to the team (weekly report 2–8 Nov 2026).

Does the agent answer phone calls too?

Yes. Besides chat and messages, an AI agent can also take calls. Our agent in operation has done so since 23 June 2026.

How is your website doing?

The check is free and without obligation. You get the report by email.

Request the website checkor ask for a first call