The rebirth of the junior marketer

We all started as junior marketing managers. Both of us began at Weber Shandwick, a PR firm’s local head office. But let’s be honest—there was nothing “managerial” about it. We updated press lists. We logged coverage reports. We cold-called journalists. At best, we tweaked press releases written by someone else. Boring grunt work.

We hated it. But now, looking back, we know being junior was necessary. It was the training ground. You needed to do the work before you could learn what to manage, what to lead.

Here’s the problem: AI just blew that training ground to pieces. It isn’t “augmenting” junior work. It isn’t “improving” it. No sugarcoating: AI is flat-out replacing it. Faster, cheaper, better. And not in some distant future. Right now.

Stanford confirms it: the entry-level ladder is collapsing. No apprentices. No pipeline. No future leaders.

The only way forward is to put the “manager” back into junior roles. Give them what AI alone cannot: a learning curve, career progression, and mentorship. Let them actually manage.

Manage what? The AI stack. Bot armies. Workflows. Outputs. Juniors should focus on learning orchestration, QA, and system design, while seniors should concentrate on setting strategy and vision. That’s how you turn today’s entry-level hires into tomorrow’s creative heads and CMOs.

It’s on us, as leaders, to rebuild the ladder. We have to provide opportunities for progress. We have to define purpose. We have to pave the road for generations to come.

— Peter & Torsten

We take a break from our usual 2-paragraph-per-news format to discuss a breakthrough on the level of the first ChatGPT—but this time in image editing.

Google released Gemini 2.5 Flash Image Preview, a new text-to-image generator and image editing model that has shaken the AI world. The model, previously teased as “Nano Banana” (which is a far superior name to “Gemini 2.5 Flash Image Preview” — come on, Google), is a capable image generator, but it is an exceptional image editor.

Just look at the leaderboard:

The difference in ELO scores between Nano Banana and the second-best model is bigger than between the second and the tenth-best model…

The model's capabilities were so immediately apparent that Adobe announced its integration into Adobe Firefly and Adobe Express. The company’s incumbent Firefly model looks like a toy compared to the mighty Banana.

The creative production bottleneck just evaporated for a vast swath of marketing assets. Nano Banana dramatically reduces the cost of experimentation, enabling A/B testing of dozens of creative variants that were previously cost-prohibitive. The competitive advantage is no longer the ability to create assets, but the ability to strategically direct and curate them at scale.

OpenAI announced its new Realtime API, making it surprisingly simple to build voice conversation agents. Companies can now create sophisticated voice interfaces that handle discovery calls, customer support, and even sales conversations. The API handles interruptions, maintains context across long dialogues, and sounds natural enough to pass as human. Pricing starts at roughly $0.06 per minute of conversation—cheaper than most SDR salaries.

Yes, this is a bit too technical for the average marketer, but we expect the tech to spread quickly and enable true conversational commerce at scale. You don't need machine learning expertise, just basic API integration skills. Marketing teams are already deploying pilots in under a week.

Voice is the new landing page. While competitors obsess over conversion rates, you can deploy AI agents that actually talk to prospects, qualify them in real-time, and book meetings with your human team.

This fundamentally changes the economics of top-of-funnel engagement. A single voice agent can handle hundreds of simultaneous conversations at a fraction of human cost.

A new working paper from the Stanford Digital Economy Lab provides the first large-scale evidence that AI is displacing entry-level, white-collar jobs. The study analyzed data from a major online job platform and found that roles requiring skills that overlap with LLM capabilities (e.g., writing, editing, basic analysis) saw a significant reduction in hiring for junior positions. At the same time, demand for senior roles with the same skills remained stable or increased.

Enter the “AI Doom Loop”: companies use AI to automate junior-level tasks, which reduces the number of entry-level positions available. This, in turn, shrinks the talent pool for future senior roles, as fewer professionals get the foundational experience needed to advance. The paper suggests this isn't a future problem but a trend that is already underway.

The classic model of career progression in marketing is breaking. Your organization's structure needs to adapt to a reality where much of the work traditionally done by marketing coordinators and junior analysts is now automated. This has profound implications for talent management, succession planning, and organizational design.

You'll need fewer "doers" and more "editors" and "orchestrators" who can manage AI systems. The long-term risk is a shortage of experienced leaders who have never had the chance to learn the fundamentals.

Generative video company HeyGen released Avatar IV, its latest and most realistic “digital twin” model. The new version offers significant improvements in lip-sync accuracy, emotional expression, and natural-feeling mannerisms, making the resulting avatars nearly indistinguishable from the real person in many use cases. Users can create a digital replica of themselves that can then be used to generate new video content from text or audio prompts.

Scalable, personalized video is now a practical reality for sales and marketing teams. Imagine your top sales executive sending a thousand unique, personalized follow-up videos to leads, or your CEO delivering tailored weekly updates to different departments, all without stepping in front of a camera.

This technology decouples content creation from the time commitment of key personnel. It allows for hyper-personalization of outreach, training, and internal communications at an unprecedented scale.

Anthropic sparked significant backlash this week after updating its privacy policy to state it will use data from its free, consumer-facing products like Claude to train its AI models. This data collection is opt-out by default, meaning users' conversations are automatically included in training datasets unless they proactively go into their settings and disable it.

The new policy does not affect users of Anthropic's paid commercial API, whose data is still not used for training.

AI adoption often happens on the individual level. You can make it “mandatory” to use your ChatGPT enterprise subscription for company-related LLM work, but people will still ask their free / paid Claude or Gemini model.

Any proprietary concepts, draft messaging, or strategic ideas entered into the tool could be absorbed into future models.

Small Headlines. Big Shifts.

» Claude for Chrome gets an agentic upgrade.

Anthropic's new Chrome extension can now not only summarize web pages but also act on them, a step toward autonomous agents that can complete tasks on your behalf. This points to a future where AIs are primary users of the web, and your site's "AI-friendliness" will be critical for discovery and commerce. Currently in early beta.

» Kimi AI launches one-prompt slide decks.

Chinese AI startup Kimi, from Moonshot AI, released Kimi Slides, a feature that generates complete presentation decks from a single prompt. This tool is aimed at automating the creation of internal reports and sales pitches, further reducing the manual effort required for routine business communications.

» Microsoft MAI models.

Microsoft launched two new in-house models optimized for enterprise workloads with better cost-performance than GPT-4. The models integrate natively with Office 365 and Azure, making them instant alternatives for companies already in Microsoft's ecosystem. Expect aggressive bundling with enterprise agreements in Q4.

» NotebookLM goes multilingual.

Google's NotebookLM is famous for generating excellent podcast-style audio summaries from long documents. With the latest update, it can do it in 80 languages. The same applies to video summaries.

» YouTube quietly AI-edits your uploads.

Creators report that background noise and pacing are auto-fixed after upload. No opt-out yet—check your brand channel for silent changes.

The junior ladder is collapsing. AI has taken over entry-level work, and without apprentices, tomorrow’s marketing leaders won’t exist.

The CMOs who succeed won’t be those with the best tools, but those who redesign roles, training, and workflows before the pipeline of talent dries up.

That’s what we’re building inside the AI Ready CMO community. Apply for membership: join.aireadycmo.com

See you next week,
Peter & Torsten

PS: Know someone who would benefit from this newsletter? Refer them and get rewarded: