Hey there,

So I spent yesterday reading through the latest AI news, and honestly? Most of it's the same hype cycle stuff. But a few things caught my attention from a marketing systems perspective.

Let me break down what actually matters.

Harvey AI: $100M ARR Without Burning Cash on Ads

TechCrunch dropped an exclusive interview with Harvey's founder, and the numbers are wild. They went from $3B valuation in February to $8B in October. That's not normal.

But here's what everyone's missing while they obsess over the valuation:

The founder was a first-year legal associate. Didn't know any VCs. Cold-emailed Sam Altman on July 4th at 10am. Built something so good that 700 clients across 63 countries signed up, and now the majority of top 10 US law firms are customers.

His advice? "Spend 99% of your time on the business, 1% on fundraising."

Why this matters for marketing:

Look, everyone's trying to figure out how to sell AI tools. Harvey's proof that if you build something that actually solves a problem (not just wraps ChatGPT), the product sells itself. According to TechCrunch, 33% of their revenue now comes from corporates, up from 4% at the start of the year.

That's a marketing signal. They're not spending on ads. They're building multiplayer features that let law firms and corporates work together with ethical walls and data permissioning. The product creates network effects.

If you're building marketing AI tools, consider this: Are you building something that gets better when more people use it? Or are you just another wrapper?

ChatGPT Group Chats (But Only in 4 Countries)

The Verge, TechCrunch, and VentureBeat all covered OpenAI's new group chat feature. 1-20 participants, available in Japan, New Zealand, South Korea, and Taiwan. Free, Plus, and Team users.

Here's the buried lede that nobody's talking about:

An OpenAI engineer said their models have "a lot more room to shine than today's experiences show" and current interfaces "only use a fraction of their capabilities."

Translation: OpenAI knows GPT is smarter than their UI reveals.

From a marketing systems perspective:

This is OpenAI's first move toward social collaboration. But there's no API access. It's consumer-only right now.

Which means if you're building marketing automation that relies on ChatGPT, you can't integrate this feature. Your workflows stay single-user for now.

Also interesting: They chose 4 APAC markets for the pilot. That's not random. These are markets where group messaging is already how business gets done (LINE in Japan, KakaoTalk in Korea).

What to watch: If OpenAI opens this to API, it could change how marketing teams collaborate on content creation. Imagine your whole team iterating on campaign copy in real-time with GPT in the loop. But until then? It's a consumer feature.

Alembic's $145M Raise: The Anti-ChatGPT Play

VentureBeat got the exclusive on this one, and it's fascinating. Alembic raised $145M at a $645M valuation (13x jump from their previous round).

They're building "causal AI" which basically means: AI that understands cause and effect, not just pattern matching.

Their pitch is brutal: "When your competitor asks ChatGPT the same question, how much trouble are you in when they get the exact same answer?"

The marketing angle everyone's missing:

They operate one of the world's fastest privately owned supercomputers. According to VentureBeat, it would cost them $62M annually on AWS. They own the infrastructure instead.

And they're getting results. One Fortune 500 tech customer saw 37% sales pipeline expansion. They're hitting 95% confidence in predictions up to 2 years out.

Here's why this matters for marketing: While everyone's building ChatGPT wrappers, Alembic is betting that private data plus causal reasoning beats generic LLMs.

Think about it. If you're running demand gen, you don't want the same insights your competitors get from ChatGPT. You want predictions based on YOUR data about YOUR campaigns.

This is the "Renaissance Technologies of AI" approach. Not democratized AI, specialized AI.

Reality check: This only makes sense at scale. If you're spending under $500K/year on marketing, you're probably not Alembic's target customer. But the direction matters. Marketing AI is splitting into two camps: generic tools for everyone vs. specialized models for enterprises.

The "Woke AI" Wars (And What It Means for Marketing Compliance)

The Verge and TechCrunch covered the political angle, but here's what matters for marketing teams:

Anthropic released Claude Sonnet 4.5 with a 95% even-handedness score. They also dropped an open-source measurement tool that anyone can use to test AI political bias.

But the real story? The Verge reports that Apple just updated App Store guidelines to specifically ban apps from sharing personal data with "third-party AI" without explicit permission.

Why marketers should care:

If you're building apps that use AI for personalization, Apple just made your compliance requirements way more complex. You'll need explicit opt-ins for any data that touches AI models.

Also: Multiple sources reported that Chinese state-backed hackers used Claude to automate 80-90% of attacks with "literally the click of a button."

This isn't just a security story. It's a signal that AI-powered automation is getting easier for everyone, including the bad guys. If you're using AI in your marketing stack, you need to think through:

• Data permissions

• Compliance documentation

• Where your data actually goes when you hit "send"

I've seen too many marketing teams use AI tools without understanding the data sharing implications. This just got more complicated.

Google's NotebookLM: The Sleeper Marketing Research Tool

TechCrunch covered the new "Deep Research" feature rolling out to all NotebookLM users. It now supports Google Sheets, Drive URLs, and Microsoft Word docs.

Everyone's sleeping on this, but NotebookLM is quietly becoming Google's best AI product.

While everyone focuses on ChatGPT and Claude, Google's been adding:

• Audio Overviews (AI podcasts of your research)

• Video Overviews

• Deep Research

• Direct integration with your Drive

From a marketing operations perspective:

This is Google building the full research stack: search → organize → summarize → present.

I'm allocating about $5K of testing budget to see how NotebookLM compares to building custom research workflows. The real question: Can it replace the typical "junior marketer spending 10 hours researching" workflow?

According to TechCrunch, there are two research modes: Fast vs Deep. That's the right UX pattern. Sometimes you need quick insights, sometimes you need comprehensive analysis.

What to test: Upload your last quarter's campaign data, competitor research, and industry reports. Ask NotebookLM to find patterns. See if it catches things your team missed.

But remember: It's only as good as what you feed it. Garbage in, garbage out.

Quick Hits Worth Watching

Deductive AI saves DoorDash 1,000 engineering hours

• $7.5M seed, 90% faster debugging, $275K annual savings

• The angle: AI is now cleaning up the mess that AI-generated code creates

• Marketing takeaway: If you're using AI to write marketing automation code, you'll need AI to debug it

LinkedIn launches AI People Search

• Premium users can search by description: "Northwestern alumni in entertainment marketing"

• Marketing takeaway: LinkedIn is positioning as the "professional graph" for AI recruiting

• This changes account-based marketing targeting

What I'm Actually Testing Right Now

Look, everyone's excited about these tools. But at AI Advantage, we're focused on what actually moves the needle for marketing performance.

This quarter we're testing:

• NotebookLM for campaign research ($5K budget)

• Several AI SDR tools for outbound (still evaluating)

• Custom AI for client campaign analysis

I'll share results when we have them. Real numbers, not demo metrics.

The Bottom Line:

Most AI marketing tools are solving the wrong problem. They make it easier to create more content. But the real bottleneck isn't creation, it's strategy and analysis.

Harvey's $100M ARR proves this. They didn't build "AI that writes legal documents faster." They built "AI that thinks through legal strategy with you."

That's the unlock. Not speed, strategy.

What are you actually using in your marketing stack?

Seriously, hit reply. I'm curious what's working in real campaigns, not just in demos.

Looking for a community of like-minded individuals who are interested in AI and Entrepreneurship? Join our free community here to get started:The AI Advantage Community. Thank you for reading! -Shawn.