Welcome, AI Strategists

💫 Current thoughts: You can do your own work now. Pull the stat, build the dashboard, write the report — no waiting on anyone. It feels like pure upside, right up until you realize the person you stopped waiting on was also the person who quietly checked the work.

Last week I spent an afternoon tracing one statistic for a client. By dinner I'd checked eight of the most-quoted numbers in that industry. Five of them led back to exactly two dead sources — including the famous "98% of texts get read," a number its own author publicly retired back in 2022. Nobody noticed. It's still running in posts published last month.

This week's post is about why that happens, and what it's costing you:

  • Why AI serving you a real number from a dead source is worse than a hallucination — and why your fact-check doesn't catch it

  • The two ways a stat dies: retracted-but-undead, and quietly promoted past what it ever measured (the "21× MIT study" that's neither 21× nor really MIT)

  • The case where the careful, hedged source got locked behind a paywall while the stripped, confident, wrong version stayed free — and AI, predictably, serves the free one

There's a fix, and it's the rare one where cleaning up your content and winning AI citations turn out to be the same move.

This Week's Breakdown

Weekly Blog

You Can Do It All Yourself Now. That's Exactly the Problem.

AI handed you the report, the dashboard, the stat — without the analyst who used to check them. I traced eight of one industry's most-quoted statistics and five led back to two dead sources, including one the author publicly retracted, to no effect. Here's why your own content archive is now a liability, and the audit that turns the cleanup into a citation win.

Pallas Advisory • Pallas Advisory

AI News Roundup for SMB Decision-Makers and Marketing Professional

KPMG fabricated AI case studies in a report designed to sell clients on AI adoption

Total coincidence this came out the week I publish something along these lines! I didn’t even cite them but, I could have.

The Decoder • Matthias Bastian

The martech categories hit hardest by AI agents

Marketo scores 50 out of 100 on agent-readiness. Gainsight scores 47. These aren't obscure tools — they're the platforms half your stack is built around. Turns out "we have an API" and "agents can actually use it" are very different sentences.

MarTech • Pamela Parker

Marketing needs AI outcomes, not more AI pilots

The pilot-to-value gap is now a real liability. The framework here is worth stealing: categorize AI use cases as defend (efficiency), extend (better outcomes), or upend (new capability) and build a portfolio across all three. The trap is over-indexing on efficiency wins while calling it AI strategy.

MarTech • Nicole Greene

AI in the workplace

The best AI deployments I keep hearing about aren't the ambitious ones — they're the embarrassingly specific ones. "Our reps waste 15 minutes on research before every ticket." Fix that. The ClickUp team got 917 hours back monthly not by transforming their business, but by fixing one annoying thing really well.

zapier.com/blog/ai-in-the-workplace

85% of IT teams claim every AI agent is under control

If your organization is heading toward the 46% IT automation threshold Ivanti projects within 18 months, the gap between "we have a policy" and "it's enforced at runtime" will cost you. The practical move: use Q3 vendor renewals to demand live demos, not documentation. Can they revoke agent access in 60 seconds under production load? If not, you don't have governance — you have paperwork.

Venturebeat

Companies Are Using Reddit to Manipulate ChatGPT and Google AI Search

SEO was already a game of trying to manipulate algorithms before humans could read your content. Now there's a version where you spam Reddit comment threads so a chatbot recommends your peptides. We've officially lapped ourselves.

404 Media

Where CMOs are finding AI success and what gaps remain

32% of CMOs are crushing it with agentic AI. 42% are still in pilot purgatory. 26% are somewhere in the messy middle. Three years in, and most of us are still figuring out how to get from "this works in one corner" to "this actually changed how we operate."

adage.com/brand-marketing/aa-where-cmos-ai-success-gaps-remain-bcg

Just Because

  • ⚽ FIFA's broadcast security was apparently one free account away from a global rickroll.

  • 🧠 Finally, a glossary for the language your brain has been dissolving into.

  • 🏢 Meta promising better snacks is a notable admission something went very wrong.

  • 🔐 The CIA has a sculpture with an unsolved code, which is extremely on-brand.

  • 🪙 Turns out "move fast and break things" also applies to your AI budget.

No theory. No slides. Just pipeline.

Most founders know their product. Few know how to get it in front of the right people. In this hands-on session, Clay + HubSpot for Startups walk you through ICP definition, prospect list enrichment, and AI-personalized outreach. You launch your first sequence before the session ends. June 18. 11am ET / 4pm GMT.

Register Before June 18

Interested in starting your own newsletter?
Use this link and get a free 30-day trial + 20% OFF for 3 months.*

*A commission is earned if a purchase is made through this link.

Keep Reading

© 2026 Pallas Advisory | AI Strategist.
beehiivPowered by beehiiv