Bombos teamBy Piet Baudoin · 1 October 2026
AI is moving into your ordinary workday: your phone, your hiring, and your employees’ browsers. Today: KPN builds AI into its business plan, Tellent tests a hiring assistant that asks for approval first, and KnowBe4 stops unauthorized AI use.
Can AI keep track of my phone calls?
Starting today, KPN, the largest Dutch telecom provider, offers Mobiel AI: a business mobile plan for small and midsize companies with AI built in. After every phone call, you get a summary with the key points, actions, and agreements. Voicemails arrive as text. The voice technology comes from ElevenLabs, a company KPN itself invests in.
Thomas Tolsma, who is responsible for small and midsize businesses at KPN, says business owners already know what AI can do, and the challenge is mainly using it simply in daily work. A gym owner in the same press release says they get many calls every day, even while training clients. The first two months are free, then it costs from 7.50 euros per month. An AI assistant and a smart phone menu will follow later.
The nice part is where the AI sits: not in a separate app you have to keep up with, but in the phone you already use. Whatever was agreed on the phone is written down afterward. A summary is not a follow-up, though. The quote still has to be made and the file still has to be updated. Keep in mind that this is KPN’s own press release, and KPN sells the plan.
Source: KPN via Emerce, 30 September 2026
Can AI help with hiring without making the decisions?
Tellent, the Amsterdam-based HR software maker behind products including Recruitee, announced Tellent Agent on September 29. The assistant uses live hiring data to flag missing information and bottlenecks. It can also draft emails, add notes to candidate profiles, and manage tasks.
The boundary is clear in its own press release: “Voor elke actie is goedkeuring van de gebruiker vereist voordat deze wordt uitgevoerd.” In English: every action requires the user’s approval before it is performed. The test starts as a closed beta and lasts six weeks. This is a supplier announcement, not a proven customer result.
For a growing company, the sequence matters most: first understand the work waiting to be done, then make a proposal, then get your approval. A candidate won’t simply receive a message because the assistant thinks it would be useful. It works the same way for applicants waiting a week for a reply.
Source: Tellent via Emerce, 29 September 2026
Can you stop unauthorized AI use immediately?
KnowBe4 announced an extension for Chrome and Edge on September 23. It can allow access to AI sites, display a warning, or block access. The security administrator sets the rule.
Matt Duren says in the press release: “Visibility alone does not stop a data leak, it just documents it”. Simply seeing which tools employees use doesn’t prevent a leak. KnowBe4 sells this security product itself.
Availability is limited: the initial version is for certain direct customers with a U.S. account environment. A broader rollout is planned for the coming months. This is therefore not yet a generally available solution for Dutch small and medium-sized businesses.
Source: KnowBe4, 23 September 2026
What are people saying about it?
A tech lead in Singapore was approached by an AI agent this week, as a candidate:
The audacity to send me an AI agent to try to hire me + several tasks to even be considered which includes a 9 page personality test, a write up on a GitHub repo, a write up on a technical example, a write up on ‘production AI agent’ AND a 100 day strategy.
No respect at all
@khairulnazran · September 29, 2026
A marketer was told at work to have Claude look for a client’s missing photos. It went differently:
Me: I can't find a client's photos, I dug through tasks, google drive, etc. Can we go back to having a dedicated Photos folder for each client?
Work: You can have Claude connect to that stuff and ask you to find it, it worked for me.
Me: Great! I'll do that!
Work: Claude is down.@ErinNudi · September 29, 2026
And Hiten Shah, an American software founder, sees what happens when everyone has their own AI tricks:
Companies are starting to build an invisible AI layer. One person has figured out how to get great research from Claude. Someone else has a setup for writing. Another knows exactly what context makes ChatGPT useful for sales. The company gets the outputs. Very little of what made them good becomes shared.
@hnshah · September 28, 2026
Where are we headed?
KPN now puts AI in the phone, and Tellent asks for approval before every action. So I expect that within two years, AI will sit in every channel where work comes in: phone, email, and job applications. The difference will no longer be who summarizes the conversation, but who finishes the work afterward and who approves it.
By then, Bombos will make sure that whatever comes in, through any channel, is ready for you as a proposal and lands in its usual place once you approve it.
What does this mean for your company?
The common thread is control over the work. Bombos starts with email into the client file: the AI agent finds the right file and puts the proposal in front of you. Our promise: once you approve, the email and its attachments are where they belong. That one task is the beginning. Your team then teaches Bombos the next one, to do more work with the same people, at a higher quality.
Sources
Every source was opened on 1 October 2026 and every quote appears in it word for word.
Tellent: Tellent Agent (29 September 2026) · KPN: Mobiel AI (30 September 2026) · KnowBe4: browser controls for unauthorized AI use (23 September 2026)