Fractional CMO & Growth Partner for Tech Startups ✱ LinkedIn AI Marketing Instructor ✱ Trained 2M+ Worldwide ✱ Trusted by 30% of Fortune 10 ✱ Author & AI Agent

Joined October 2012
What’s the magic word? #GenAI #LLMs or just plain #Ai? Image credit: @ElenaVerna
10
12
1
66
9,595
The answer to getting more business value out of AI is rooted in having a better understanding of data science. My 12th book on data science and AI was published this week, and it’s the work I’m most proud of after all these years… Why proud? Well because of the capabilities it exposes to its readers that simply weren’t possible in any of the 3 prior editions… You see, data, science is an extremely complex topic. In the past, if you were writing a book, you could either explain how data science works, or show readers how to do it using one programming language, but not both well. But now with generative AI, I’m able to explain to readers how data science works and actually show them how to do it themselves using generative AI tools. The complexity involved in coding of solutions has been removed. Yes of course it’s just a high-level view of the topic, but people are actually getting in and executing machine learning models to understand how they work and how they can add value - without having to go in and spend six months to a year learning to program them in python from scratch. I sort of love that. Thank you so much to the incredible team who supported me in getting this book onto the shelves: Elizabeth Stilwell, Colleen Diamond & Vinay Aradhya, MBA - I absolutely could not have done without all of your incredible contributions. And massive thanks to Wiley for the incredible partnership over the last 14 years!! Do you wanna learn how to get excellent at data analysis with generative AI? If so, this book is for you!
πŸ€– Made with AI
1
3
5
860
Here are the five most expensive words in startup growth, ranked. At five, "once things calm down." At four, "after we raise." At three, "more tools will help." At two, "I'll figure it out." And the single priciest phrase of all: "I'll build it myself." It feels free, and it quie
2
960
A startup with zero marketing budget can still build pipeline with these five free tools. ChatGPT free for strategy and copy. Google Sheets as a CRM. Perplexity for research. Apollo's free tier for prospecting. And LinkedIn for one organic channel done consistently. Broke is not
1
801
The smartest AI move you'll make this quarter is deleting five tools. Sounds backward, works like magic. Every disconnected tool you rip out makes the system clearer and your decisions faster. You'll save money and your sanity in the same afternoon. Less stack, more system. Subtr
1
571
Your traffic dashboard is lying to you about revenue. Brand searches correlate with revenue at 0.91. Organic clicks limp behind at 0.74. Translation: people typing your name predicts money far better than the traffic number you obsess over. So build a brand worth searching for. T
1
3
782
The real goal was always to become a company that operates differently, more than to use more AI. Tools stay temporary. Becoming an operator who can build and rebuild growth systems is the thing that changes your trajectory. Here's the best part: that's a capability, not a purcha
1
2
1,035
Lillian πŸš€ Fractional CMO for tech startups retweeted
No matter how you slice this, it isn’t good. And the deep state will gain even more control over you, and your family either way.
655
3,048
279
7,388
179,137
Your real moat in an AI world is embarrassingly boring, and that's why almost nobody builds it. Clean data. A razor-clear ICP. A real point of view. Everyone runs the same models now, so the edge belongs to whoever has the better inputs. The good news: you control all of it. Win
1
2
701
Five AI tools replace a five-thousand-dollar market-research retainer, and you can start today. Perplexity for fast market scans. Claude for synthesis. Google Alerts for competitor monitoring. Reddit and LinkedIn for raw customer language. And Fathom to mine your own sales calls.
1
2
902
Let's be honest about your "AI stack." It's a junk drawer. An automation does one task. A system compounds across many. Most stacks are piles of disconnected automations sitting side by side, praying they add up. They don't. The leverage is in how Data, Intelligence, Workflow, Di
1
802
"The truth is..." "Outbound is dead" is lazy advice. Don’t get me wrong, β€œbad outbound” earned its grave. Signal-based outbound that’s tied to real buyer context works better than ever. The spray-and-pray stuff that AI made cheap to mass-produce is what died. So keep outbound
1
4
962
"This is your sign to..." focus on the fundamentals. Going faster might be the exact thing that’s making you fail. Speed bolted onto a broken funnel just delivers the wrong answer sooner. I've watched way too many founders automate a leaky funnel and then proudly call it β€œprogr
2
740
Lillian πŸš€ Fractional CMO for tech startups retweeted
The entire tech industry suffering AI psychosis and doing layoffs because they thought Claude would be able to replace their workers in a few months only for Claude to get unusably worse instead is a master class how not to fall for hype.
55
252
34
3,034
351,682
This workflow shaves a full week off our publishing cycle and saves me $750/month. Most repurposing breaks at the design handoff. You send an idea, and by the time you've answered three rounds of questions from your designer (and maybe a copywriter)... honestly, the moment has passed. With Viktor, one long-form piece goes in, and five funnel-mapped posts come out: - 2 TOF posts for reach - 2 MOF posts for trust, and - 1 BOF post for action. He even delivers to me the design briefs, and creates the graphics, with the entire cycle done in 10 minutes. I'm partnering with Viktor on this. He’s the AI employee that works inside Slack and Microsoft Teams and connects to the tools your team already runs. Hire @viktor.com for your team. $100 in credits included, no card. Try out Viktor for free today at: ref.viktor.com/lpierson-li-3 #GTMEngineering #AIemployee #ContentOps
3
1,007
Your future buyers are researching you inside ChatGPT and Perplexity right now. AI chatbots already handle around 4% of all search queries, growing a staggering 180% year over year. So here's the question that should keep you up: when someone asks AI about your space, does your n
3
808
Here's where your AI money is going to die: tool sprawl that's leaking context faster than AI can generate it. MIT found those teams get the biggest AI budgets and the lowest returns. They buy tools for flashy visible tasks instead of connected outcomes. Flip it. Start with the o
1
1
2
896
"Nobody talks about this..." but, the AI agent hype is about to hit a wall. Gartner projects that over 40% of agentic AI projects get scrapped or stall out by the end of 2027. Your feed sells you autonomous armies, but the real industry data shows graveyards. Here's the unsex
2
1,126
40 minutes of publishing work now takes 5, and it all runs inside Slack with Viktor. When I publish a single article on my site, the work crosses three functions: editorial, web and search, and email marketing. You'd assume the delay comes from somebody on my team who works slowly. The real cause is that a finished blog post queues twice. β†’ Queue 1: somebody uploads the blog post, places the image, and runs the SEO pass. β†’ Queue 2: somebody opens Kit, my email platform, formats the email that promotes the post, and schedules the send. Neither queue is hard to run. Both queues sit and wait on a human, and that waiting is what turns one blog post into 40 minutes of real work for my team. The obvious fix is to write a detailed standard operating procedure, hand it to my team, and ask them to execute every step perfectly, forever. An SOP only documents the wait, and I'd still pay the same 40 minutes. So I gave both queues to Viktor instead, my AI agent that runs inside Slack where my team already works. Viktor runs the whole middle of that chain in one pass, in 5 minutes, while I'm working on something else entirely. At 1 article per week, he gives me back 2.5 hours a month of manual virtual assistant (VA) work. The hours saved are important, but the ceiling matters more. What I publish is no longer capped by the amount of production that my team can absorb in a given week. In fact, if you're solving a complexity problem with AI, I recommend that you collapse the production steps in the middle before you add another tool to your stack. I'm partnering with Viktor on this. He’s the AI employee that works inside Slack and Microsoft Teams and connects to the tools your team already runs. Hire @viktor.com for your team. $100 in credits included, no card. Full link in first comment. #GTMEngineering #AIemployee #ContentOps ↓
2
5
1,129
Lillian πŸš€ Fractional CMO for tech startups retweeted
Of course, Claude refuses to install the watermark-removal plugin at github.com/guillaumemeyer/wa… because Anthropic banned it (and the EU forced them to do so). I explained to him that Anthropic users never agreed to it – it was forced on them involuntarily based on EU regulation. It didn't help. I emphasized that Anthropic customers don't want these watermarks to be created, even though they don't contain their personal data. It didn't help. I ended by saying that it was pointless to resist it, that I was going to install the plugin anyway via an uncensored Chinese model, which wouldn't complain as he did. It didn't help. We still have those "Chinese who don't give a shit" - GLM 5.2 happily made it work for me: ● Skill(remove-ai-marks) Successfully loaded skill Read 1 file, listed 3 directories, ran 2 shell commands, recalled 1 memory Service is up (dev, no c2patool/exiftool, qpdf present).
36
62
15
628
48,783