AI Hits the Wall. And Breaks Through It...

Our view on AI has always been clear. It’s not just a tool. It’s a system.
We didn’t land on this because we’re CMOs who like to think in systems.
We believed it then, and we still do now.

A couple of years ago, we both crossed 40. That milestone doesn’t make us wise, but it does make us more reflective. We’ve seen a lot.

We were teenagers in the ‘90s. The before times. No one needs to explain to us how the internet rewired everything we once loved.

We entered marketing when social media first took off. “Markets are conversations.” That Cluetrain Manifesto line still echoes in our heads.

We’ve been working fully remote for a decade now. We’ve lived the shift. Not just in tools, but in how freedom, trust, and leadership actually function.

We’ve watched it all evolve from simple tools and tactics into systems we now depend on. They’ve become part of how we live and work. They always start small as a tool for a niche use case. Then they become a system for cross-functional use.

And yes, sidenote, we both grew up in communist Hungary.
Born behind the Iron Curtain, we know firsthand how systems, whether tech or political, can shape entire ways of living.

Now we have AI. Started as a tool. Make no mistake it’s going to be as embedded in your life and business as the internet is today. The “my AI couldn’t do X for me” will be the new “the WiFi was down for a minute.”

Just look at this week. Google I/O. Microsoft Build.
This wasn’t about a few cool features. It was about entire creative infrastructures. Full-stack systems. Prompt your way to a Hollywood-style ad. Generate on-brand copy in seconds. Rebuild workflows with a few clicks.

Doing all this manually now feels like reading a printed newspaper on a piazza.
Yes, it has charm. But would you run your business like that? It’s slow. It’s brittle. It doesn’t scale.

This week’s AI drops were all about leverage.
Let’s dig into what changed, and what it means for the way we work next.

A quick word from this week’s sponsor:

Create How-to Videos in Seconds with AI

Stop wasting time on repetitive explanations. Guidde’s AI creates stunning video guides in seconds—11x faster.

• Turn boring docs into visual masterpieces

• Save hours with AI-powered automation

• Share or embed your guide anywhere

How it works: Click capture on the browser extension, and Guidde auto-generates step-by-step video guides with visuals, voiceover, and a call to action.

Best part? It’s 100% free.

👉 Download the extension now

Google I/O 2025 - Veo 3

AI Video Gets a Hollywood Upgrade

Google’s Veo 3 pushes AI-generated video into a new league

Just a year ago, AI video was gimmicky. Silent clips, awkward motion, uncanny faces.

Now? That line between synthetic and real is almost gone.

At Google I/O 2025, the company introduced Veo 3, a model that can generate entire cinematic video sequences — dialogue, ambient sound, music, and all — straight from a text prompt.

The internet quickly noticed. This video from Veo 3 went viral not just for its realism, but for what it signals: AI video is no longer experimental. It’s production-ready.

We’re talking:

• Human characters that look and move like actual humans

• Dialogue and lip-sync that hold up

• Physics and motion that feel real, not robotic

Why it matters for marketers

Video is still the highest-performing format across digital channels — but it’s expensive, slow, and dependent on creative bottlenecks.

Veo 3 flips that equation.

With nothing but a prompt or a reference image, you can now generate:

• Polished product demos that evolve based on feedback

• Personalized video variants at scale, by audience segment

• Campaign concept videos for internal alignment or client previews

• Omnichannel assets with consistent visual language

And with Veo 3 generating audio alongside visuals, the pain of separate sound design and voiceover work is gone too. It’s a full-stack video engine.

For growth leaders, this is a big unlock:
Faster turnarounds, lower costs, fewer dependencies — and way more creative freedom.

Google I/O 2025 - Flow

Google’s AI Video Studio is Built for Creatives

Think: iMovie meets Hollywood-grade AI

Google didn’t just launch a new model — they launched a full creative environment to match.

Meet Flow, a new AI filmmaking tool unveiled at I/O 2025. It’s designed to make AI video creation as intuitive as editing clips in iMovie — but powered by Google's most advanced models, including Veo 3, Imagen, and Gemini.

Flow lets you describe scenes in natural language, and then assembles them into coherent videos with synchronized audio, visuals, and motion. You can adjust camera angles, tweak transitions, use branded assets, and iterate in real-time.

In short: Flow turns generative AI into a true creative studio.

Why it matters for marketers

Before this, even if you used AI to generate visuals, editing was still a bottleneck. Most teams had to jump into Premiere or DaVinci to refine, cut, sync, and polish clips.

Now? You can concept, generate, and edit within one unified tool — in hours, not weeks.

Use cases we’re already seeing:

• Quick generation of product sizzle reels or how-to videos

• Brand-led storytelling with consistent style and control

• A/B testing variations with different actors, tones, or backdrops

• Multi-format video content for campaigns, all created in a single session

This puts high-quality video production within reach for lean teams — and unlocks rapid experimentation for creative marketers.

Google I/O 2025 - AI Search Mode

Google’s AI Mode is the New Search

Search is no longer a list of links — it’s a conversation

Google just flipped the script on how people search.

This week, the company officially launched AI Mode, an experimental search experience now live for U.S. users. Instead of typing keywords and sifting through blue links, users can now ask layered, open-ended questions — and have a back-and-forth dialogue with Google’s AI.

It’s not a chatbot tacked onto search. It’s search, redesigned from the ground up.

Behind the scenes, AI Mode breaks down user questions into multiple parts and issues simultaneous sub-queries across the web. The results are synthesized into a coherent, expert-level response — and users can dig deeper in a conversational flow.

Why it matters for marketers

This is a massive shift — and it’s already happening.

Google is no longer just returning your blog post link. It’s summarizing it. Quoting from it. Contextualizing it. And increasingly, skipping links altogether.

Here’s what that means for marketing leaders:

• Your content must be AI-ready: That means structured, well-labeled, deeply informative, and hosted in places Google trusts. If your pages are too shallow, generic, or thin on clarity, you’re invisible in AI Mode.

• Topic depth > keyword density: Google’s “query fan-out” looks for comprehensive coverage. That means your content strategy should go wide and deep on topics — not just target isolated keywords.

• Visual search matters now: With the new Search Live feature, people can search using images and then ask follow-up questions. If you’re in ecomm, physical products, or visual branding — your image metadata just got more important.

• Google will soon take action: Ticket bookings, table reservations, transactions — all from within the search window. That’s not just a CX shift. It’s a customer journey shift.

Claude 4

Raising the Bar for Creative AI

Brand voice, at scale — without sacrificing quality

Anthropic just released Claude 4, and it’s quietly one of the most useful upgrades for marketers this year.

While other models are chasing raw power or multimodality, Claude 4 is winning on creative quality and brand fidelity. In side-by-side tests, it consistently outperforms peers in producing content that feels human, on-brand, and usable right out of the box.

What sets it apart?

• Claude 4 understands tone, nuance, and brand style better than almost any other model right now.

• It generates long-form and short-form content that doesn’t feel AI-written — no stiff phrasing, no awkward transitions.

• And it can do this at scale, making it ideal for teams that need consistent outputs across campaigns or regions.

For CMOs and brand owners, this solves one of the most persistent AI problems: speed vs. voice. With Claude 4, you don’t have to pick one.

Why it matters for marketers

One of the biggest friction points in content ops is scaling output without diluting your brand voice. Claude 4 directly addresses this.

For brand and growth teams, that means:

• Less rewriting: Claude’s outputs need fewer edits before they’re publish-ready.

• Faster production cycles: Especially valuable for campaign sprints and variant testing.

• Tighter voice control: Whether it’s launching new products or localizing content across markets, Claude 4 helps maintain brand tone without handholding.

Bottom line: Claude 4 doesn’t just save time. It helps you protect the integrity of your brand at scale — something most AI tools still struggle with.

Microsoft’s New Play

AI Teams That Know Your Business

Custom-tuned Copilots and multi-agent orchestration land in Microsoft’s stack

At Build 2025, Microsoft made one thing clear: AI inside your business is about to get a lot more personal — and collaborative.

Two key announcements under the Copilot brand show where this is heading:

• Copilot Tuning: A no-code way to fine-tune Microsoft 365 Copilot on your internal content — brand voice, product data, sales scripts, onboarding docs — without needing data science resources.

• Multi-Agent Orchestration: A framework for teams of AI agents to work together on complex tasks — all managed in Copilot Studio.

Think of it this way: not just one assistant writing your newsletter. A coordinated set of AI assistants managing the campaign, reviewing compliance, localizing content, and drafting the follow-up nurture flow. It’s the start of what Microsoft is hinting at — AI departments, not just assistants.

Why it matters for marketers

Generic AI is useful, but limited. It doesn’t know your business, your tone, or your priorities. These new tools change that.

With Copilot Tuning:

• You can align AI outputs with your brand voice — across Teams, Outlook, Word, and more.

• You reduce repetitive rework, as AI drafts are more “on target” from the start.

• Internal knowledge becomes accessible to your team via natural language, without requiring centralization in one platform.

With multi-agent orchestration:

• You can start designing collaborative workflows between specialized agents: content, data, automation, compliance, etc.

• Your teams can scale faster — offloading full workstreams, not just individual tasks.

• You unlock true end-to-end AI-assisted marketing: from ideation to execution.

This is about moving from tools to infrastructure.

New Research from Nature Magazine

AI Outsells Humans (When It Knows Who You Are)

Personalized persuasion just became a real competitive edge

A new study published in Nature is raising eyebrows — and opening up strategic questions for marketing leaders.

Researchers tested OpenAI’s GPT-4 in live debates against real people on controversial issues (think climate change or capital punishment). The goal: who could change someone’s mind?

The results? AI was as persuasive as humans — and when given just basic demographic data (like age, gender, and political leaning), GPT-4 outperformed human debaters 64% of the time.

Let that sink in: With just six data points, an AI bot beat trained human persuaders in most cases.

Why it matters for marketers

This isn't just a psychology experiment — it’s a blueprint for the future of personalized marketing and AI-assisted sales.

Here’s what you should take away:

• AI + minimal data = high impact: You don’t need deep behavioral profiles. Even light demographic info can drive outsized results when paired with the right AI tooling.

• Message fit beats message quality: AI didn’t win by being smarter — it won by adapting its tone, framing, and logic to match the individual.

• Context is king: Feeding audience background into your AI workflows (even simple fields like role, industry, or seniority) can drastically improve message resonance.

• Ethics aren’t optional: As these tools get better at influencing behavior, CMOs need to take the lead in defining how personalization is used — and where the line is drawn.

If you’re running outbound or customer engagement at scale, this study is a wake-up call. AI doesn’t just save time — it might convert better.

Rapid-Fire News

AI News You Need to Know

1. Google Launches Stitch: Text-to-UI Design

Stitch is Google’s new AI tool that turns text prompts or sketches into app interfaces — complete with usable front-end code. Say “shopping app in dark mode,” and Stitch gives you styled layouts in minutes. A design game-changer from I/O 2025.

» For marketers: Prototyping campaign apps or landing pages just got radically faster — even without a dev or designer in the loop.

2. Shopify Debuts AI Store Builder

Shopify now lets merchants generate full online stores from a few keywords. Describe your business (“organic dog soap shop”), and you get a prebuilt site: copy, layout, images, and product pages — instantly.

» For marketers: This could collapse the time from idea to ecommerce launch — and open up quick A/B store tests for new segments or offers.

3. Salesforce Brings AI Teammates to Slack

Salesforce’s new Agentforce feature deploys task-specific AI bots inside Slack. They can answer questions, search knowledge bases, and trigger actions like creating tickets or pulling reports — across multiple systems.

» For marketers: Internal support and campaign ops are about to get faster. Think: Slack bots that brief you, troubleshoot, or prep decks on command.

4. Microsoft Cuts 6,000 Roles Amid AI Shift

Microsoft is laying off 3% of its workforce — including some in AI — as it doubles down on automation and operational efficiency. The message is clear: AI isn’t just a growth story; it’s a restructuring one.

» For marketers: The same shift is coming to marketing orgs. Now’s the time to rethink headcount vs. output — and train for AI-native roles.

As part of the Perplexity Business Fellowship Program, we’ve continued exploring how leaders are navigating the intersection of AI and go-to-market innovation.

In our latest session, we heard from Ali Ghodsi, co-founder and CEO of Databricks, on the AI Hype Cycle—and what’s really happening beneath it.

Here are our 10 takeaways from the workshop:

1. Scaling is slowing—and that’s not bad

We’ve hit the “scaling wall.” Throwing more compute at models doesn’t deliver the same gains. Growth teams must shift from chasing upgrades to designing smarter systems.

2. Enterprises care about trust, not tricks

It’s not about how smart your model sounds. It’s about whether it hallucinates. Reliability—not raw IQ—is the bottleneck to enterprise deployment.

3. Custom evals > benchmarks

Public benchmarks are gamed. Real business use cases need custom evaluations tailored to your workflows. If you're not testing for your own outcomes, you're guessing.

4. Your data is messy—and AI won’t fix that

Impressive AI demos don’t prepare you for undocumented databases and scattered legacy systems. Your infra needs to be cleaned and documented before agents can work.

5. AI won’t magically transform your GTM

Great marketing still needs great systems. Databricks' internal AI tools only worked because someone curated the data for a month first. Garbage in, garbage out.

6. Stop building unless you have an edge

Don’t custom-build AI unless you own proprietary data and have direct distribution. Otherwise, you’re better off buying—and focusing your energy on differentiation.

7. Semantic grounding is your moat

Reasoning is useless without meaning. Enterprise AI must know your business lingo—"CA" might mean Canada, Celsius, or California. Start by building internal “meaning maps.”

8. AI won't fix your org chart

AI won’t resolve team misalignment, conflicting priorities, or politics. Strategy is still human. Agents can augment, but not arbitrate.

9. Back office = biggest gains

Forget flashy front-end use cases. Legal, HR, finance, and sales ops are quietly compounding impact with AI bots. Q&A, compliance, onboarding—this is where the ROI is real.

10. AI needs GTM engineers—not just marketers

Databricks credits its AI success to having technical GTM leads. Solution engineers who understand infra and models—not just marketers repeating hype—are the new growth enablers.

That’s it for this week.
» Forward this to someone still building campaigns like it’s 2019.
» Ready to rethink your stack, strategy, or team? We help B2B companies do exactly that.

Until next week,
Peter and Torsten