The Future of Chatbots: Those Who Invest in the Wrong Way Now Will Lose Control
Traditional FAQ bots are becoming less relevant, while generic AI assistants like ChatGPT are taking over as the primary sources of information. A survey of more than 1,100 people shows where things are headed—and which two paths companies should take now.
This question comes up more and more often in workshops and consulting sessions, usually after the official agenda has been covered: Do we even need our chatbot anymore? It’s asked by decision-makers who poured a lot of energy into this very project three or four years ago—not because they’re tired of technology, but because usage numbers are declining. A recent survey of more than 1,100 people, which forms the basis for moinAI’s guide ’The Future of Chatbots in Business,« provides reliable figures on this topic for the first time. The answer is more nuanced than either side would like.
Why the Question of the Future of Chatbots Is Coming Up Right Now
Business chatbots have been around for more than ten years. Most were built for a clearly defined purpose: to answer frequently asked questions, guide visitors through the site, and pre-screen simple inquiries. That worked as long as the alternative was a search engine or a contact form.
With ChatGPT, Gemini, and Claude, this landscape has shifted for three reasons at once. Economically, because significant resources are invested in building and operating these bots. Strategically, because the interface with customers is one of a company’s most valuable assets—anyone who hands it over to a generic platform also relinquishes direct contact and control over the customer experience. And technologically, because open standards like APIs, webhooks, and the Model Context Protocol enable new combinations: The question is no longer «proprietary bot or generic AI,» but rather how the two can work together in a controlled manner.
What the Numbers Show
For the guide, 1,152 end users were surveyed, and six experts from the banking, insurance, logistics, healthcare, and software development sectors were consulted. According to the author, Dr. Sophie Hundertmark, four findings are crucial.
When people have a question about a product or service, 56 percent turn first to a generic AI assistant, and 38 percent to a search engine. No one cited the company’s own chatbot as their first point of contact. Not a few—no one.
There is a wide gap in perceived quality: On a scale of one to five, generic assistants receive a rating of 4.1, while company-owned bots receive a rating of only 2.6.
An apparent contradiction in the responses is particularly revealing: Respondents agree most strongly with the statement that they no longer need a corporate chatbot for general information (4.2). However, they agree almost as strongly with the statement that a corporate chatbot is useful if it can perform actions (3.9). According to Hundertmark, this isn’t a contradiction—it’s actually a fairly precise set of instructions.
Discussions with companies confirm this pattern: One major bank reports declining usage and is considering shutting down the bot. Another company sees no threat because its bot operates exclusively in the logged-in area with a complete customer profile. The threat level does not depend on the industry, but rather on what the bot was originally built for.
To put this in context: 81 percent of respondents use generic assistants daily—the sample is more AI-savvy than average and therefore not a representative cross-section, but rather a glimpse into a trend that, based on experience, the rest of the user base will follow with some delay.
Option A: Ensure visibility where questions actually end up
Anyone who accepts that general questions will be answered by generic assistants in the future must ensure that they are answered correctly there. This isn’t a traditional marketing task, but rather a matter of data quality. According to Hundertmark, experience shows that five points make the biggest difference: machine-readable formats with clean HTML structure, unambiguous semantics with consistently used product names and Schema.org markup, a single, well-maintained source of knowledge instead of conflicting versions, content sections clearly organized by topic, and deliberate control over which information should be publicly accessible.
According to Hundertmark, the side effect is actually the real benefit: The same work also improves the quality of your own assistant’s responses. Hardly any other measure offers a better balance between effort and impact.
Option B: The Future of Chatbots Lies in Taking Action, Not in Responding
That leaves the inquiries that require data or specific actions. Here, having your own bot is still worthwhile—but only if it actually does something. 81 percent of respondents would like to check their order status via a company bot, and 69 percent would like to book an appointment. In contrast, only 31 percent want personalized advice from a chatbot.
Hundertmark considers three points to be crucial: First, specialized agents instead of a generalist—a bot that’s supposed to do everything usually does little of it well. According to the guide, Nuremberg Airport uses clearly defined agents for baggage, parking, check-in, and flight status, each of which accesses only the knowledge they’ve been authorized to use. This reduces “hallucinations” and makes quality measurable. Second: Structure the data first, then integrate individual functions via APIs and webhooks—starting with low-risk read access, followed by transactions. Third: Control mechanisms such as RAG, guardrails, and the bot’s ability to admit it doesn’t know something aren’t a convenience—they’re liability protection. A bot that hallucinates convincingly is more expensive than one that simply forwards the request.
It's not a question of whether or not it's a bot, but which bot it is
The simple FAQ bot has outlived its purpose—based on these figures, that’s hard to dispute. According to Hundertmark, however, this doesn’t mean it will be shut down, but rather that its functions will be shifted: the informational component will move into the AI interface, while the action-oriented component will be integrated into an assistant.
The strategic question, then, is not «bot or no bot,» but rather: which bot, with which data source, which integration, and which role. Anyone who still invests budget in a better chatbot today is optimizing a channel that customers have already abandoned. And anyone who ignores both is leaving their brand’s image in the hands of the systems that are in demand anyway—but without any control over what they say.
This article is based on the guide «The Future of Chatbots in Business,» published by moinAI and featuring practical insights from companies in the DACH region. The guide is available for free download: https://www.moin.ai/downloads/chatbots-guide-2026
