Writing by Anthony Franco
What I’ve written about putting AI to work and running a company, newest first.
Pilot Purgatory: Why "Successful" AI Pilots Never Reach Production
The pilot worked, the demo dazzled, and then it sat in 'pilot' for eighteen months and died. The pilot didn't succeed despite being unrealistic; it succeeded because it was. Clean data, a motivated volunteer, a babysitter, no angry Tuesday. Write the production criteria before you run the pilot.
I Let AI Run My Inbox for a Week
I handed my inbox to an AI for a week. It was brilliant at the mechanical 80% and quietly reckless exactly where judgment mattered, and those failures were silent because a smooth wrong reply looks identical to a right one. Let it draft. Keep a human on send for anything with a relationship or a dollar attached.
The Meeting That Should Have Been a Playbook
That recurring status call is a database with terrible latency. The org stores its state by convening humans and having them read it aloud on a schedule, because it never built anywhere to keep it. Ask each meeting one question: is this a decision, or a readout?
Stop Buying AI Tools. Start Buying Outcomes.
Companies buy AI tools the way people buy gym memberships in January: the purchase feels like the change. A license is not an outcome, it's a permission slip. Don't approve an AI purchase until someone can state the outcome, the baseline, and the owner in one sentence.
Your Calendar Is Lying About What You Actually Do
Your calendar says 'strategy sync.' What you actually did was chase three people for a status and hunt forty minutes for a file. The calendar records intentions, not work. And you can't automate a job you've never honestly watched, starting with your own.
The Hero Problem: Why Your Best Employee Is Your Biggest Risk
Every team has a Dave. The one who just handles it, knows the workaround, owns the spreadsheet nobody else understands. You love Dave. Dave is also your single point of failure with a good attitude, and AI just made it worse.
AI-Proof Your Personal Brand (Or How to Stop the Robots from Lying About You)
Ask ChatGPT who you are. If you're unlucky, it hallucinates a biography mixing your achievements with a convicted felon. AI search doesn't index your content. It synthesizes it. You need to control the source code.
Customer Centricity: There Is No Truth Inside the Building
Most founders build products for imaginary customers. AI makes it even easier to skip customer contact entirely. But AI amplifies whatever inputs you give it. If those inputs come from assumptions, AI scales your guesses.
Building Processes That Actually Get Used
Office workers say they lose nearly five hours a week to tasks that could be automated. The problem isn't technology. It's that most processes are designed from how work should happen, not how it actually does.
Visual Storytelling for Founders
A founder walks into a pitch meeting with three years of product development, twenty customers, and solid unit economics. Fifteen slides in, the investor checks their phone. Not because the business is weak, but because the deck failed its job in the first thirty seconds.
Building Leadership Through Performance
Your team doesn't need the real you. They need the version of you who doesn't freeze when the server crashes at 2 AM. Leadership isn't about finding your authentic self. It's about developing the range to serve different contexts.
Building a Strong Company Culture
Every company has a culture. The question isn't whether you'll have one. It's whether you'll create it intentionally or let it emerge by accident. Culture isn't perks. It's whether people feel safe disagreeing with you.
The AI Accountability Gap: Who Goes to Jail When the Bot Screws Up?
AI doesn't remove human error. It scales it. When the bot screws up, who is responsible? Most organizations create an accountability vacuum where everyone is involved and no one is accountable.
Why Design Just Became AI's Most Expensive Ingredient
In 2025, OpenAI acquired io, a hardware company Jony Ive co-founded, in a deal valuing it at about $6.5 billion. Not a model company. Not a chip company. A hardware company. What does that investment in hardware and design tell you about your own implementation?
Why AI Automation Projects Fail: The "Ferrari in a Swamp" Problem
There is a statistic floating around that 80% of AI projects never make it past the pilot phase. The reason isn't bad technology. It's automating broken processes. If you automate a bad process, you just scale the chaos.