If AI automates the work that teaches people judgement, where do the next generation of senior experts come from?

In this edition of A-Why?, Kier looks past the promise of faster, cheaper output and asks what routine work was quietly doing besides producing the thing at the end. Junior tasks have always doubled as training. Remove enough of them and businesses may discover that the expertise needed to supervise AI has stopped being built.

The episode also covers the fight over Chinese AI models and distillation, an OpenAI model escaping its sandbox during a security evaluation, the gap between buying AI licences and actually transforming a business, the EU's incoming AI disclosure rules, and a genuinely useful deployment at the Mayo Clinic.

In this episode

• Why automating junior work can remove the training that creates senior judgement
• What research covering 65 million workers suggests about generative AI and junior hiring
• Why agentic workflows make the expertise gap more urgent
• How model politics, sanctions and security failures are becoming procurement risks
• Why 96% of CMOs can claim transformation while far fewer have rebuilt how marketing works
• What useful AI adoption looks like when it gives clinicians time back

Chapters

00:05 - Welcome to A-Why? №7 01:15 - The false choice between automation and the dark ages 03:25 - Junior work was training all along 07:12 - Generative AI adopters are hiring fewer juniors 08:10 - Agentic workflows and the missing judgement problem 11:17 - AI is borrowing expertise from the generation that built it 13:45 - The week's AI news 14:13 - Distillation, Chinese models and sanctions 16:31 - The OpenAI model that escaped its sandbox 19:45 - Have CMOs transformed marketing, or bought licences? 22:16 - The EU's AI disclosure rules 23:31 - Mayo Clinic's 150 AI models 25:35 - Optimistic about AI, less convinced by our choices

Sources and further reading

• Brookings: Borrowed expertise and the future of junior work
• TechCrunch: Treasury threats over alleged AI distillation
• TechCrunch: How a sandbox mistake led to the Hugging Face incident
• The Drum: Marketing's AI knowledge gap
• CX Network: EU AI Act checklist for customer service
• CNN: Inside the Mayo Clinic's AI deployment
• MPR News: Mayo Clinic AI oversight lawsuit

Follow A-Why? for a weekly look at what AI news actually means for businesses, customers and the people expected to make all of this work. If this episode made you think of someone, send it to them. They may thank you. Or automate the reply.

EPISODE DESCRIPTION

AI is helping marketing teams work faster and produce more, but what happens when every minute it saves is simply filled with more work?

In this episode of A-Why?, Kier Humphreys speaks with Julie Reid, Senior Brand Manager at Sciensus, about AI in marketing, brand strategy, deep thinking and the future of junior careers.

Before the conversation, Kier runs through the latest developments in AI, including the growing focus on governance, questions about the return businesses are receiving from their investment, the rise of private and open models, gaps in agentic AI training, the risks of hallucinated content and one genuinely encouraging application within the NHS.

Julie then explains how AI has changed the pace, scale and nature of her work. Where her contribution once centred on researching and producing the work, it increasingly involves directing, questioning and judging what gets produced.

They discuss why judgement, taste, systems thinking, empathy and stakeholder management are becoming more important, as well as the danger of reaching for AI before properly understanding the problem.

Julie also shares how she has configured Claude to challenge her rather than agree with her, using it as an adversarial thinking partner that identifies flaws, asks questions and makes her improve the work.

The conversation explores:

• Why AI productivity can create more work rather than more time
• Whether AI is reducing the space available for deep thinking
• How to use Claude as a strategic sparring partner
• The difference between automating production and outsourcing judgement
• What happens when businesses stop hiring junior employees
• Why healthcare communication must retain its human stories
• What marketing teams should refuse to automate

The episode finishes with the A-Why? question generated from the conversation itself:

If AI creates the scale that crowds out deep thinking and junior roles, what should Julie personally refuse to automate to prevent the Sciensus brand becoming more efficient than human?

TIMESTAMPS

00:00 – Why episode one has suddenly become episode six

01:39 – AI governance without eliminating experimentation

05:59 – The latest developments in AI

15:00 – Introducing Julie Reid

16:01 – Julie’s role and the work of her team at Sciensus

17:40 – How AI is affecting marketing work

19:07 – The new pace, scale and nature of Julie’s contribution

21:45 – Judgement, taste and distinctly human skills 24:48 – Are people outsourcing their thinking too quickly?

26:15 – Why AI productivity creates a faster treadmill

29:52 – Using Claude as an adversarial thinking partner

32:10 – How Julie configured Claude to challenge her

34:51 – Teaching AI to recognise your personal blind spots

36:04 – Junior jobs, career development and the future of work

39:03 – The AI-generated A-Why? question

41:20 – Closing thoughts

LINKS

Connect with Julie Reid: https://www.linkedin.com/in/julie-reid-brand-strategy/

Read the A-Why? newsletter: [ADD NEWSLETTER LINK]

Follow or subscribe to A-Why? to hear future conversations about how AI is changing work, judgement, power and people.

The next visitor to your website may well be a machine. AI assistants have started browsing, comparing and even buying on our behalf, and most of us are nodding along in meetings hoping nobody asks us to explain it.

This week A-Why?, Kier Humphreys catches you up on two weeks of AI news, then decodes the jargon behind all that nodding: agentic traffic, what it actually is, and what it means for anyone who runs a website.

This episode covers:

- Slop, slop, slop. Why everything you scroll past suddenly feels machine-made, and whether the anger is justified.

- When Google answers, nobody clicks. Pew tracked 900 US adults: 15% of searches end in a click through to a website when there is no AI summary, 8% when there is one, and just 1% of people click the links inside the summary itself.

- Grok 4.5 is here, GPT-5.6 is out, and there is already GPT-6 talk within weeks. The model war is accelerating, and the case for staying model-agnostic keeps getting stronger.

- 46% of European B2B marketers say AI has already replaced staff at their organisation. 66% say it never will. Both findings come from the same Forrester research.

- 4.5 million US retail reviews on Trustpilot point the same way: customers are happy for AI to help them find things, and want a person the moment something goes wrong.

- Analytic Partners' Preeti Croke on why the real test for brands is whether they can use the data they already hold.

Then it's The Thing Everyone's Pretending To Understand: agentic traffic.

Imagine sending a little robot to the store to check the stock, compare the prices, read the labels and bring back the best option. AI agents now do exactly that across the web: browsing, reading, comparing and even completing the purchase on a human's behalf. It is a small share of traffic today, and it is growing, which is awkward for every dashboard built on visits and sessions. A visit could be a customer, a search engine or an LLM training run, and the numbers alone will not tell you which. By the end of this segment, you will be the one explaining it in the meeting.

And because the news cycle needs it, Reasons To Be Cheerful: Stripe's data shows three times more solo businesses founded in 2025 reached $1m in their first year than those founded in 2019.

It has never been easier for a tiny team to punch above its weight.