By Piet Baudoin · 9 November 2026
In a study, a chatbot that writes fluently but tells nonsense was rated as intelligent and pleasant as a stiff chatbot that gets it right. Also two reports from Statistics Netherlands (CBS), the Dutch national statistics office: how many employees come home drained from work, and where companies still invest.
Do you trust a chatbot faster when it phrases things nicely?
Yes. Communication scientists at Radboud University in Nijmegen had participants read a conversation with a chatbot, in four versions: fluent or curt, correct or with made-up facts. A chatbot that writes fluently but tells nonsense was “als even intelligent en prettig beoordeeld” (rated as equally intelligent and pleasant) as a stiff chatbot with the right answers. And “liefst tachtig procent van de deelnemers” (no less than eighty percent of participants) did not notice the inserted errors at all.
Associate professor Barbara Müller: “Je denkt dan onbewust al snel: dit zit wel goed, ik hoef het niet meer te controleren.” (You quickly think, without realizing it: this is fine, I don’t need to check it anymore.) Professor Tibor Bosse wants chatbots to show their doubt: “Als een chatbot zélf twijfel aangeeft, voorkom je dat mensen onzin voor waarheid aannemen.” (If a chatbot signals doubt itself, you prevent people from taking nonsense for truth.)
What it means for you: an AI answer that sounds neat has not been checked yet. Let AI in your office prepare work with the source attached, so a person can see at a glance where it came from. Also read how to prevent AI mistakes and who checks them.
Source: Emerce, 5 November 2026
Do your people come home tired from work?
More and more often. CBS reported this morning: “In 2025 voelde 21 procent van de werknemers zich enkele keren per maand of vaker psychisch vermoeid door het werk.” (In 2025, 21 percent of employees felt mentally tired from work a few times a month or more.) Ten years ago that was 13 percent. CBS includes “zich aan het eind van de dag leeg voelen” (feeling empty at the end of the day). Among employees aged 25 to 35 it is 28 percent, and the more complex the work, the more often it happens.
CBS does not say where the tiredness comes from. But every owner knows the day that disappears into loose ends: retyping an email, looking for an invoice, calling back a customer who already called twice.
What it means for you: don’t just look at how much work there is, but also at which work drains your people. That is often the first work you can hand off.
Source: CBS, 9 November 2026
What are companies still investing in?
Technology. According to the business survey by CBS, the Dutch Chamber of Commerce (KVK), the construction research institute EIB, and the employer associations MKB-Nederland and VNO-NCW, almost a quarter of companies postponed or canceled investments in the past year. Still: “De meeste bedrijven investeren in technologie om toekomstbestendig te blijven.” (Most companies invest in technology to stay future-proof.) Over 46 percent name technology, such as digitization or automation of processes, as their main investment area for staying future-proof. And: “Het tekort aan arbeidskrachten blijft de belangrijkste belemmering voor de bedrijfsvoering” (The labor shortage remains the main obstacle to running the business), for 32 percent of companies.
What it means for you: the labor shortage is the biggest brake and technology the biggest investment, though the figures do not say whether one follows from the other. If you face that choice, read whether to hire someone or use AI as work grows.
Source: CBS, 5 November 2026
Voices
A user of Spiceworks, a forum for IT administrators, describes how Gemini confidently gave him a solution that he says no longer worked in Outlook. The forum gives this account the rank Goat; five others liked the post. It is one experience that shows the risk of a confident answer, not a confirmation of the research.
OMG talk about timing. Here’s a PERFECT example of why AI is garbage and needs to go back to pre-school before opening it’s mouth. [...] Guess what, buttercup, YOU CAN’T ANYMORE!
jrblood on Spiceworks · 29 May 2026
Where are we heading?
In the Radboud study, eighty percent of participants did not notice the inserted errors (Emerce, 5 November). So a fluent answer is no longer a sign that it is right. I expect companies to judge AI more and more on answers they can check, and that within two years you will no longer choose an AI agent by how smart it sounds, but by how honestly it says what it isn’t sure about. Models are getting so good at reading what you mean that they will also sense where you need to look and where you don’t. Then the work in the office shifts: no longer checking everything, but deciding in the right place. In five years, the people who now come home drained will come home with energy left, because the loose ends were already tied up and they only made the choices that matter. AI is not a colleague and not a person in this. It is a helper that wants to do good work, and the human chooses.
What does this mean for your business?
The common thread: AI only really helps when it takes work off your hands and shows where it came from. At Bombos that starts, for example, with email into the client file. Bombos reads incoming email, prepares it with the attachment for the right client file and shows why it chose that file. You approve. Our promise: at the end of the day every email is where it belongs, without anyone retyping it. That way the same team does more work, and better work. That task is the start; after that your team teaches Bombos the next one itself.
Sources
Every source was opened on 9 November 2026 and every quote appears in it word for word.
Emerce, on research by Radboud University (5 November 2026) · CBS, mental tiredness from work (9 November 2026) · CBS, business survey October 2026 (5 November 2026) · Spiceworks: jrblood (29 May 2026)
Bombos team