In partnership with
Your prompts are leaving out 80% of what you're thinking.
When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.
89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.
| Zymbos Intelligence · Wednesday 17 June 2026 | ||
|
||
|
|
The loud money this week went to infrastructure and acquisitions. The story that touches your team is quieter. Across five reports, one through-line keeps surfacing: artificial intelligence (AI) is not coming for the senior expert first. It is dissolving the first-draft work that turned juniors into seniors. A Microsoft warning, a surge in "bot-minding" hires, fresh corporate training, a cautious giant going all in, and a wary research workforce all point the same way. The entry-level job is still there. The learning inside it is not.
|
|
|
Workforce · United States
Nadella warns AI could hollow out entire industries
Microsoft chief executive Satya Nadella told an audience this week that frontier AI could "hollow out" entire industries, drawing an explicit parallel to the way globalization stripped out whole categories of work. His argument: as models commoditise expertise, the moats that protected skilled firms and skilled people get shallower. Coming from the head of a company selling the technology, it reads as a planning signal, not a sales pitch. Boards heard a warning about the senior end of the workforce, about experts and defensible skills. That is the part that makes headlines. It is also, on the evidence of the rest of this week, the wrong end to watch first. The damage starts lower down, where the people who would become tomorrow's experts are supposed to learn their trade.
McGann's TakeHe is right that something gets hollowed out. He is just looking at the top floor while the foundations go. The senior expert is exposed eventually, but the junior meant to replace them is exposed now, because the work that used to make them is the first thing handed to the machine.
Read more →
|
|
Hiring · United Kingdom
UK AI hiring surges as firms hire people to babysit the bots
New UK hiring data shows roles tied to supervising AI systems up sharply, reported at around 61 percent, as employers create jobs for people to monitor, correct and sign off the output of automated tools. The phrase doing the rounds is "babysitting the bots". On the surface this looks reassuring: AI is making jobs, not just taking them. Look closer and the shape of the work has changed. The new entry-level role is to watch a machine produce the first draft and catch its mistakes. That is genuine work, and it needs doing. It is not the same as producing the first draft yourself, which is how a junior used to learn what good looks like. A generation is being hired to check work it was never taught to do.
McGann's TakeA job that asks you to correct the machine is not the same as the one that taught you the craft the machine just did. Catching the errors needs the very judgment you used to build by making them yourself. We are hiring people to babysit the bots into a role that assumes an expertise the role no longer produces.
Read more →
|
|
Skills · Global
OpenAI launches Academy courses for the next era of work
OpenAI released three structured courses through its Academy: AI Foundations, Applied AI Foundations, and Agents and Workflows. The goal is to take employees from basic understanding to building repeatable, agent-assisted workflows, with training delivered straight from the model maker. It is a serious attempt to standardise AI fluency across a workforce, and the price of entry is low. It is also worth being clear about what it teaches. These courses teach the tool. They are very good at that. The old apprenticeship taught something harder to package: judgment, built slowly by doing the work and getting it wrong. Formal training is a useful floor. Treating it as a replacement for on-the-job learning is the mistake to avoid, because the two teach different things.
McGann's TakeCourses teach the tool, and the apprenticeship taught the judgment. Do not confuse buying the first for solving the second. A certificate in prompting shows a junior can drive the software; it says nothing about whether they can catch the software being confidently wrong, which is the skill that actually pays.
Read more →
|
|
Adoption · Global
Samsung reverses its ban on external generative AI
Samsung, which blocked ChatGPT internally after a 2023 data leak, is now rolling three generative AI models out across the company and accelerating its internal "AX", or AI transformation, programme. For a security-cautious giant, concluding that the productivity case now outweighs the data risk is a real adoption signal. When the company that once banned the tool now mandates it, every cautious board still sitting on the fence gets cover to move the same way. It also speeds up the very process this issue is about. When a large, careful employer goes all in, the routine first-draft tasks go first: the summaries, the briefs, the starter code, the early models. Those tasks were never just output. They were the training ground. The faster a company adopts, the faster that ground disappears, unless someone deliberately rebuilds it somewhere else.
McGann's TakeWhen the cautious giant commits, the grunt work goes first and goes fastest, and the grunt work was the classroom. Every firm that benchmarks itself against Samsung now has cover to do the same. The adoption curve and the apprenticeship-collapse curve are the same line, drawn twice.
Read more →
|
|
Workforce · Global
Scientists have a bad case of AI FOMO, Nature poll finds
A Nature poll of more than 1,900 researchers found nearly half feel negative about AI, while around 60 percent fear being left behind if they do not adopt large language model (LLM) tools. Nature called it AI "fear of missing out". The pattern is familiar from any knowledge-work team: scepticism and anxiety living side by side. Adoption driven by fear rather than confidence tends to be shallow, the kind where people lean on the tool without ever trusting or learning it. It matters here for a specific reason. The people best placed to mentor juniors are the experienced staff, and many of them are uneasy, stretched, and unsure of the tools themselves. A wary, time-poor senior is a poor teacher. That is precisely the wrong condition for rebuilding an apprenticeship at the moment one is most needed.
McGann's TakeThe people who should be teaching are anxious, stretched, and unsure of the tools themselves. A nervous mentor cannot rebuild an apprenticeship, and fear-driven adoption breeds shallow users, not teachers. The generation that should be handing down judgment is too busy keeping up to notice it has stopped.
Read more →
|
|
|
This Week's Analysis
What happens to entry-level work when AI does the first draft
Here is the claim, and it is narrower than the usual jobs panic. The workforce problem in 2026 is not that AI replaces senior people. It is that AI does the first-draft work that junior people used to learn on, and the apprenticeship hidden inside the entry-level job is disappearing with it. The job title survives. The training that came free with it does not. Start with how people actually learned. Nobody became a good analyst, lawyer, marketer or engineer by attending a course. They became good by producing the first draft: the rough memo, the starter model, the first cut of code, the early client brief. The draft was usually poor. A senior marked it up, sent it back, and explained why. Repeat that a few hundred times and judgment forms. The first draft was never valuable as output. It was valuable as practice. The training was hidden inside the job
Now look at what changed. AI does the first draft. It is faster, cheaper, and on a Tuesday afternoon it is good enough. So the rational manager hands the first draft to the machine and hands the junior the machine's output to check. This week shows every step of that shift: hiring built around supervising AI rather than producing the work, formal courses sold as the new way in, and even cautious employers adopting at speed. The junior still has a job. They have simply stopped doing the thing that used to teach them. The third point is the one that should worry you. The old training pipeline was implicit. No one budgeted for it, named it, or put it on a plan, which means no one is protecting it now that it is being removed. We are dismantling an apprenticeship system without noticing we had one, and so far nothing has been built to replace it. AI did not eliminate the entry-level job. It eliminated the apprenticeship inside it.
The strongest counter-argument deserves a fair hearing. Used deliberately, AI could make apprenticeship better, not worse. Hand the drudgery to the machine, and a senior could spend the freed time on real mentoring while the junior tackles harder problems sooner. That is a genuine possibility, and in a few well-run teams it will happen. But it only happens if someone designs it. Left to default, the freed-up senior time gets reabsorbed by more delivery, and the junior becomes a reviewer of machine output who never built the judgment to review well. So do not wait for a policy. Pick one role on your team and name the specific skill that used to be learned by doing the first draft and is now learned by no one. Then rebuild one deliberate way to teach it: a weekly review where the junior produces the draft before the machine does, a standing session where a senior shows their working, a rotation that forces the hard task. Small, named, and on the calendar. The apprenticeship will not come back by accident. It has to be put back on purpose. Further readingI looked at how AI will reshape jobs over the next five years, and what it means if you are early in your career, in a recent report. This issue zooms in on one piece of it. Read it on LinkedIn →
|
|
|
Sembly AI
Meeting Intelligence · Transcription · Action Items
What it is
Sembly AI joins your calls on Zoom, Microsoft Teams, Google Meet and Webex, transcribes them in more than 40 languages, and produces a shareable summary with detected tasks and owners. A non-technical reader can have it running in about ten minutes, which is the test we hold this slot to. Why it fits this week
When the junior who used to write the minutes is now checking a machine, the meeting record becomes the team's memory, and the tool that builds it matters. Sembly turns the conversation into a searchable record with decisions and action items attached, so what gets decided in a room does not leave with the people who attended. Watch for
Only Enterprise plans are excluded from model training by default. On the other tiers you set that opt-out yourself, so do it before your first real call. The old permanent free plan is now a free trial, so treat the trial as your evaluation window rather than a long-term home. Ratings
Verdict
A fast, capable meeting assistant that turns talk into a shared, searchable record with tasks attached. The score lands at 7.5 because the work is strong but the entry-tier privacy defaults need a hand. Set the data controls on day one, trial it on a week of real calls, and judge it on your own meetings. |
|
|
The apprenticeship audit
Workforce · Skills · Works in Claude or ChatGPT
This is the Deep Intelligence argument made executable. Name one role and the tasks AI now does for it, and the prompt returns the skills that role is no longer learning, what to deliberately substitute, and the smallest weekly ritual that rebuilds the path. Run it on the role you would most regret hollowing out: the one you are counting on to produce your future seniors. Read the middle column first. Those are the skills that used to arrive for free and now will not arrive at all unless you act.
You are a workforce development advisor. I want to audit one role on my team for hidden skill loss caused by AI doing the first-draft work.
Role: [job title, e.g. junior analyst] Tasks AI now handles for this role: [list 3 to 6 tasks the tool now does that a person used to do, e.g. first-draft summaries, starter models, initial research] Return a table with three columns: 1. Skill no longer being learned: the specific judgment or craft this person used to build by doing each task above. 2. Deliberate substitute: a concrete way to teach that skill now that the task is automated (mentoring, rotation, draft-before-the-machine, and so on). 3. Weekly ritual (under 60 minutes): the smallest standing practice that rebuilds the learning path, with who runs it and how you would know it is working. Then name the single skill whose loss would cost the most in three years, and explain why that one matters more than the others. The most useful line it hands back is the last one: the single skill whose loss costs the most in three years. That is where the real decision lives, because you cannot rebuild every apprenticeship at once, and that answer tells you which one to start with.
|
|
|
Closing Perspective
The quiet story beats the loud one
The quiet stories usually beat the loud ones, and this is a quiet one. So two predictions, both with a clock on them so you can hold me to them. First, within 18 months, by the end of 2027, at least one major employer will announce a structured "AI apprenticeship" programme, named as such, built specifically to replace the on-the-job learning that AI has removed. Someone will notice the pipeline is empty and try to engineer it back. Second, by the end of this year, "workforce AI" will move from a productivity headline to a skills-pipeline headline in UK and European Union (EU) government discussion. The question shifts from "how much does this save?" to "who is training the next generation, and on what?" Both are falsifiable. If no such programme appears by the end of 2027, or the policy framing has not shifted by year end, I got it wrong, and I will say so. One thing you can do today. If you ran the apprenticeship audit on one role on your team, hit reply and tell me what skill stopped being learned. I read every response. John McGann
Founder, Zymbos AI |
|
Zymbos Intelligence
Subscribe free at zymbos.ai
You are receiving this because you subscribed at zymbos.ai
© 2026 Zymbos Intelligence · John McGann · London, UK Zymbos Ltd · Company No. 16198848 · Teddington, England |