Two years ago I made a decision that changed everything at Metadata: we're going AI-first. Not as a marketing buzzword. AI-first in how we build, how we sell, how we operate.
It panned out.
We launched MetadataONE, an AI agency that never sleeps. We shipped Pixel for AI ad creation and multi-channel launch. And we built an MCP server with roughly 100 tools that gives humans and AI agents the ability to run full end-to-end ad campaigns across Meta, LinkedIn, X, Reddit, Google Ads, and Bing. Not prototypes. Production campaigns, at scale.
Most people are still debating whether AI is ready. We're running our business with it.
When it clicked for me
I started experimenting with Manus and its open-source alternatives over a year ago. Even back then, with early versions, I could see what was coming. Agents that plan, execute, and iterate on real tasks.
I built BedtimeMagic.com in 72 hours using Manus. AI bedtime stories for kids in 11 languages. Real users, real revenue. That project taught me more about what's possible than six months of reading papers.
And even with two years of head start, I'm still amazed by the pace. Every week something new shows up that would have been science fiction a year ago. Jensen Huang just told Lex Fridman he thinks we've achieved AGI. The CEO of a $3.4 trillion company. This is real.
AI brought back my love for building
I didn't expect this part. When you run a company for years, you can drift from the hands-on work that made you excited in the first place. AI changed that for me. Especially Claude Code.
I'm back in the terminal. Building things. Shipping fast. It feels like the early startup days again.
I recently built an entire presentation with Claude Code. Research, data, graphics, avatar, voice cloning, a resource library with 132 tools. No engineers. Just me and an AI coding agent.
A 3-person team can now out-execute a team of 30. Not someday. Right now.
Playbooks that are working
Through our work at Metadata and my own experiments, a handful of AI playbooks keep delivering:
AI video at scale. HeyGen avatars plus ElevenLabs voice cloning. High-value video content that used to need a production team. One startup got 18 million views doing this.
Inbound-to-outbound pipeline. Take your best content and use Claude Code to build automated cold email sequences from it. Content that attracts becomes the seed for outbound that converts.
Programmatic SEO. Bulk-create targeted landing pages from keyword research. What took a content team weeks now takes hours.
Multi-agent cold email. A 4-agent chain trained on thousands of real campaigns. Research, personalization, send optimization, all automated.
LLM council. Multi-model debate for strategy decisions. Put Claude, GPT, and Gemini in a room and let them argue it out. The output beats any single model.
I go deep on all of these in my Scaling GTM with AI talk.
The risks nobody wants to talk about
I'd be doing you a disservice if I skipped this part. AI at scale can go wrong, and the damage is real. I've documented cases totaling $75-100M+ in losses from AI failures in enterprise settings.
Six risk categories to manage: hallucination in customer-facing content, brand safety in automated campaigns, data leakage through AI tools, cost overruns from unmonitored API usage, compliance gaps, and over-automation without human checkpoints.
Build fast. Build guardrails at the same time.
Resources
I'm documenting this journey as I go. If you're exploring AI as a founder or marketer, start here:
132 AI tools, templates, and playbooks curated from two years of building.
Scaling GTM with AI is my full talk on how small teams can outperform much larger ones.
AI Tools Night is a 90-minute hands-on session from beginner to advanced.
I post weekly on LinkedIn about what's working and what isn't.
The gap between people who are building with AI and people who are watching keeps getting wider. The learning compounds. Start now.