🤝 Join our live course: How to Avoid AI Project Failures, For Org Leaders

Most AI pilots don’t fail because of the tech.
They fail because leaders fund the wrong things, in the wrong order.

In this 30-minute Lightning Lesson on Maven, I break down:

• why AI pilots stall before ROI
• the decision mistakes that quietly kill momentum
• a simple way to prioritize AI bets that actually stick

Built for Org Leaders who have to sponsor AI, not babysit it.

→ Reserve your Free seat (limited)

P.S. Can’t attend live? Get the recording right after.

Weekly finds

📰 AI Insight

1. Gemini’s “Personal Intelligence” Upgrade. Google just gave Gemini the keys to your Gmail, Photos, and YouTube. It can now reason across your personal data to help you at a tire shop or plan your day without you lifting a finger.

2. Months of Research in 20 Minutes Anthropic’s latest case studies show Claude completing genomic research that usually takes humans weeks. It’s officially moving from “chatbot” to “lab partner” for high-stakes data discovery.

3. Pocket TTS: Voice Cloning for Your Laptop Kyutai released a model that clones any voice with just 5 seconds of audio, and it runs locally on your CPU. No cloud, no GPU, and total privacy for sensitive business audio.

4. Anthropic’s “Cowork” for the Rest of Us Claude’s agentic power is now in a macOS tool for non-coders. It can autonomously organize files, build reports, and manage expenses while you focus on actual strategy.

5. The “Ralph Wiggum” Agent Workflow A new viral workflow turns product docs into JSON for coding agents to build features overnight. For $30, you can have an agent implement 10 features while you sleep.

THE ISSUE

You Can’t Automate Chaos: Cleaning Up Data Before You Add AI

1. The Real Pain

Your CRM says a deal is “hot.”
Sales swears they followed up.
Marketing insists the lead never converted.
Ops is quietly exporting another CSV at 10:47 p.m.

Meanwhile, someone asks, “Can we just add AI to fix this?”

Here’s the blunt truth most teams don’t want to hear:
AI doesn’t clean up messes. It scales them.

If your data is fragmented, outdated, or flat-out wrong, automation just helps you make bad decisions faster. And if you’ve felt that tension lately, dashboards you don’t trust, forecasts that miss, tools that don’t talk, you’re not behind. You’re normal.

Painfully normal.

2. What Most People Get Wrong

Most advice about AI sounds like it was written for two extremes:

• Startups with five tools and a clean slate

• Enterprises with armies of analysts and governance teams

Mid-market companies live in the middle. Too complex for “just plug it in.” Too lean for massive rebuilds.

The hype says: “AI will magically unify your data.”
Reality says: “AI will confidently give you the wrong answer.”

This is why so many AI transformation for SMB efforts stall. Not because the tech doesn’t work, but because the underlying systems were never designed to support it.

AI isn’t the strategy. It’s the amplifier.

3. What This Actually Means

For a $10M–$100M company, AI-ready data doesn’t mean perfect data. It means:

• Your systems agree on basic facts

• Fields mean the same thing across tools

• People trust the numbers enough to act on them

Inside the business, this shows up everywhere:

• Sales: Duplicate accounts, missing stages, stale pipeline

• Marketing: Leads with no source, contacts with five owners

• Ops: Workarounds, shadow spreadsheets, manual reconciliations

• Leadership: Dashboards that spark debates instead of decisions

What it isn’t:
A massive replatform. A shiny AI chatbot. Or a six-month “data initiative” that dies quietly.

This is operations data quality, the unglamorous work that makes everything else possible.

4. Why This Matters Now

If you don’t address this, here’s what keeps happening quietly:

• Leads wait hours (or days) because routing is unclear

• Forecasts lag reality because inputs are stale

• Teams re-check work because they don’t trust the system

• Decisions slow down because nobody’s sure what’s true

AI for business only creates leverage when the basics are solid. Otherwise, you’re automating hesitation.

Competitors who fix this first don’t look flashier. They just move faster, and that adds up.

5. The Practical Breakdown: How It Actually Works

This isn’t a checklist. It’s a progression.

Step 1: Map What Data You Actually Use

Why it exists: You can’t clean what you don’t depend on.
What breaks: Teams document everything instead of what matters.
Good enough: The 10–15 fields that drive revenue and execution.

Step 2: Define Meaning Once

Why it exists: “Active customer” shouldn’t mean three things.
What breaks: Every team has its own definition.
Good enough: One shared definition, written down, socially enforced.

Step 3: Find the Source of Truth

Why it exists: AI needs a referee.
What breaks: Multiple systems claiming authority.
Good enough: One system wins, even if it’s imperfect.

Step 4: Fix the Flow, Not Just the Field

Why it exists: Bad processes create bad data.
What breaks: Teams patch symptoms instead of workflows.
Good enough: Fewer handoffs, clearer ownership.

Step 5: Remove Zombie Data

Why it exists: Old data poisons models.
What breaks: Fear of deleting “just in case.”
Good enough: Archive aggressively. Keep what’s alive.

Step 6: Add Guardrails

Why it exists: Humans will always find shortcuts.
What breaks: Optional fields and loose rules.
Good enough: Required fields where it counts, not everywhere.

Step 7: Only Then, Layer AI

Why it exists: AI learns from what you feed it.
What breaks: Automating before stabilizing.
Good enough: Start narrow. Prove value. Expand.

This is the foundation of any real AI strategy.