"If we put a chatbot on our website or in our app, will customers actually use it, and how do we prevent a bad experience?" With generative AI entering customer service, this question is on the agenda of many companies. The customer survey that Gartner shared in September 2026 shows that the answer lies in design more than in bot technology: only 27% of customers are willing to try a chatbot again after a negative experience.
In this article we read the findings of Gartner's two press releases together, touch briefly on the regulatory framework, and offer a practical checklist for designing the bot together with human handoff.
The gap between willingness and usage
In the survey Gartner ran in February and March 2026 with 3,566 B2B and B2C customers, 49% of respondents say they would be willing to use a chatbot if the company offered one. Yet the share who used a chatbot or digital assistant in their most recent service interaction was only 7%.
Gartner attributes this gap to past experiences. If a bot has previously misunderstood the problem, given generic information or made it hard to reach a person, the customer chooses another channel. According to the same survey, in their most recent service interaction customers used third-party generative AI tools roughly three times more than the company's own chatbot.
The first takeaway from a management perspective is this: the claim that "customers want a bot" is not an indicator of actual use. Gartner also states that service leaders should take this gap into account in their adoption and return-on-investment forecasts. A business plan should be built not on stated interest but on the work the bot actually resolves.
The cost of a single bad experience
Gartner likens the retry rate of only 27% to a "leaky bucket" problem: a single failed interaction can discourage future use even if the bot's capabilities improve later. This clearly shows the risk of launching a bot with a "go live first, improve later" approach. A customer lost in the first version may never see the second.
Gartner's recommendation is to put reliability ahead of scope. A bot that consistently solves a defined set of problems builds more customer trust than one that takes on every topic and often fails.
Customers want to be able to reach a human
The August 2026 press release draws on the same survey and completes the picture. While 50% of customers say their interactions become easier when companies use generative AI, 87% say that in this case the option of reaching a human agent is a must. When customers who were unwilling to use AI were asked what could change their minds, the most common answer was being able to switch to a human agent when needed.
For this reason, Gartner recommends against making generative AI the mandatory first step for every problem. When a customer is forced to go through multiple failed AI interactions before reaching a person, the likelihood of using the tool again drops.
The press release also shows that expectations are not limited to obtaining information. Of the customers who use generative AI, 58% have used it to complete a task on their behalf; in B2B settings this share rises to 74%. Transactions such as booking an appointment, placing an order or sending a document require the bot to connect to company systems in the background.
Regulatory framework: transparency and the duty to inform
For companies working with the European Union (EU), the topic also has a compliance dimension. Article 50 of the EU AI Act requires AI systems that will interact directly with natural persons to be designed so that people learn they are interacting with an AI system; this obligation does not apply where the situation is already obvious from the standpoint of a reasonable person. The information must be provided clearly and distinguishably, at the latest at the time of the first interaction. According to the European Commission's implementation timeline, the transparency rules have applied since August 2, 2026. For the details of the timeline, see the article on the EU AI Act's new timeline.
In Türkiye, on the other hand, if the bot collects personal data such as a name, phone number or order details during the chat, the obligation to inform under Article 10 of Law No. 6698 (KVKK, Türkiye's Personal Data Protection Law) comes into play: when data is obtained, the data controller is obliged to provide information on matters such as the purpose of processing, transfers and the legal basis. Passing the chat history on to the agent during a handoff to a human should also be assessed within this framework.
A practical path: a bot designed together with human handoff
The steps below are a general approach derived from Gartner's recommendations and the regulatory framework above; each organization needs to adapt them to its own channel structure and customer profile.
- Start with a narrow, proven scope. Gartner recommends starting with problem types proven to have high resolution rates and with well-tested, targeted deployments. A good starting point is to work out which topics current requests concentrate on and which of them can be answered by fixed rules.
- Say clearly what the bot can do. Tell the customer up front which tasks the bot can handle, and state in the first message that they are talking to AI.
- Make the handoff to a human visible. If the bot cannot resolve a problem with high confidence, the path to a human agent must be clearly visible. Do not hide the handoff behind a few failed attempts.
- Pass on the context. According to Gartner, during the handoff the bot should relay to the agent the information and context it has gathered up to that point. Having to explain the same information a second time is the most familiar form of a bad experience.
- Move from answers to transactions. Customer expectations are shifting toward an assistant that can carry out transactions. Gartner recommends a structure in which generative AI understands the intent and traditional automation in the background completes the transaction.
- Measure the right indicators. The number of chats "contained" in the bot is not a measure of success on its own. Track the resolution rate, the handoff rate and whether the same customer uses the bot again, together.
- Expand the scope according to performance. Add new topics only after performance in the current scope has become reliable.
- Involve the legal and compliance team from the start. Review matters such as the AI disclosure, the privacy notice and the retention of chat records before going live.
Conclusion
A chatbot can bring speed and access to customer service; however, the data shows that the credit customers extend to a bot is limited. In Gartner's words, the goal is not to keep every interaction inside the bot but to steer every customer toward the fastest and most effective resolution path; that path may be AI, a human, or a combination of the two. For this reason, human handoff should be treated not as a backup that kicks in where the bot falls short, but as a core part of the design from day one.
Why logistics experience matters in bot design. In a business that moves goods or dispatches technicians, many customer requests ("Where is my order?", "When will the delivery arrive?", "When is the technician coming?") are operational questions at heart. A bot can answer them reliably only if it is connected to the order, shipment, warehouse and service records behind them (step 5), and a handoff works only if the agent receives the order or ticket number and what has already been tried (step 4) and someone on the operations side is able to act on it. Both are operational design questions as much as technology ones, and given how little room a first bad experience leaves, they are best settled before launch. One of our consultants brings more than 30 years in operations, logistics and supply chain: at a technology distribution company of roughly one billion dollars in scale, he ran logistics, warehousing, technical service and administrative operations end to end, building and leading teams of 100+ people and delivering lasting gains in operational performance through KPI-driven programs. He has also managed nationwide warehouse and transport operations across leading retail, logistics and distribution organizations. An industrial engineer, he now brings that supply-chain discipline to SMEs through AI and automation. Meet our consultants.
This article is for general information purposes only and does not replace legal advice.
#interimmanagement #forwarddeployedengineering