Insights
Lessons from the field on AI adoption, value creation, and building companies.
Your AI Vendor Claims Their LLM Can Reason. Here's What's Actually Happening.
Every AI vendor claims their LLM can reason. They all run next-token prediction underneath. Here's what that means for the capability claims you trust.
Build Once, Sell Many. AI Just Broke That.
The software model works because you build once and sell many times. Enterprise data restrictions are breaking that replication engine for AI companies.
The Unbundling of LLMs
LLM wrappers get dismissed. But the companies building real workflow around foundation models are following the same pattern that turned Excel spreadsheets into billion-dollar SaaS companies.
Better Data Beats More Data. Here's How to Build for It.
Knowledge graphs, contextual retrieval, and data-centric AI deliver larger gains than model upgrades. Practical architecture for AI systems that prioritize signal over volume.
The Problem with Feeding Everything to Your AI
The most expensive mistake in enterprise AI is assuming more data produces better answers. Research going back 75 years explains why it doesn't.
Your Team Structure Made Sense at 10 Engineers. It Doesn't at 40.
Most software companies start flat. Then the team grows and delivery slows. The structure didn't break because the people got worse. It broke because nobody redesigned it.
Your Engineers Know How to Code. Do They Know What You Stand For?
Every engineering organization has coding standards. Few have documented what the team actually believes about how software should be built. That document is the engineering manifesto.
Your Team Says It Does Agile. Your Release Frequency Says Otherwise.
71% of organizations say they use Agile. Most of them still ship unpredictably. The ceremony isn't the problem. The operating cadence is.
You're Measuring the Wrong Things. Here's What Actually Tells You If Your Engineering Team Is Healthy.
Story points and tickets closed tell you something is happening. They say nothing about what's being produced. Three metrics actually predict engineering team health.
Your Engineering Team Has Most of the Right Practices. It's the Last Four That Make the Difference.
Most teams have the rituals but lack the results. The difference is four decision-quality practices that feel like overhead but compound over time.
Before You Let AI Act Alone, Ask This One Question
The question isn't whether AI is capable. It's what happens when it's wrong. A practical framework for deciding when AI needs human oversight and when it doesn't.
What the 5% Who Succeed with AI Actually Do Differently
MIT and BCG both found that only 5% of companies create real value from AI. The difference is not better technology. It is six operational decisions the winners make before writing a line of code.
74% Want Revenue from AI. 20% Are Getting It.
88% of companies have adopted AI. Fewer than 40% can point to a financial result. The gap between AI winners and everyone else is widening fast.
Your AI Pilot Worked. That's the Worst Thing That Could Have Happened.
88% of AI proofs-of-concept never reach production. The problem isn't the technology — it's that pilots are designed to succeed, not to survive real operations.
Which AI Future Do You Believe In? The Terminator, Wall-E, or Star Wars
Will AI replace your job? BLS projections, OECD surveys, McKinsey research, and 224 years of automation history all point to the same answer: AI changes tasks, not entire occupations.
Agent Collaboration: Where the Value Lives, Where the Risk Hides
Agent collaboration creates compound value no single platform can deliver. But complexity grows quadratically and data security risks multiply with every handoff.
Guardrails and Human Oversight: The Governance Layer that Makes AI Agents Safe
Over 40% of agentic AI projects will be canceled by 2027 due to inadequate risk controls. Here are the five governance components every organization needs before scaling.
AI Orchestration Without Visibility Creates Hidden Bottlenecks
Orchestration dashboards measure what agents produce. They rarely measure the exceptions that land on humans downstream. Here's how to fix that.
Where Your AI Agents Actually Run: Compute, Data, and Infrastructure Decisions
AI agents need compute, data access, and runtime environments. The infrastructure decisions you make now will shape cost, security, and vendor dependency for years.
PE's Zombie Problem Isn't About Capital. It's About Operations.
PE's zombie problem isn't fundraising or fee structures. It's that portfolio companies aren't getting better. The data shows why.
The Orchestration Layer: What It Is, What It Does, and What to Look For
The orchestration layer makes multiple AI agents work together. Here are the five core functions, what to look for, and how to evaluate solutions.
AI Agents Are Bringing Back Point Solutions. This Time, It Might Actually Work.
AI agents are reversing the logic that drove platform consolidation. Here's why point solutions are back, what changed, and how to structure adoption.
AI Agents Automate the Grunt Work. Your Team Focuses on Judgment.
AI agents follow a gather-check-synthesize-deliver pattern that automates grunt work. Here's how to identify where they fit in your business.
The Einstein Test: Why AGI Is Not Around the Corner
Demis Hassabis proposed the clearest test for AGI: could AI have discovered general relativity? Today's systems cannot come close. Here's why that matters.
What's Actually Inside an AI Agent (And Why Most of It Isn't AI)
Most of an AI agent isn't AI. I built one to prove it. Here's the actual breakdown and why it matters for every AI project you evaluate.
Before You Bet 90 Days on AI: An Operator's Scoping Checklist
Most AI projects fail because they're scoped like technology rollouts, not process experiments. A six-point checklist for operators before committing 90 days.
AI Can Write. It Cannot Think.
Writing quality and factual accuracy are independent. LLMs optimize for the first. Understanding why changes how you use every AI output.
The AI Analysis That Analyzed Itself Into Nonsense
A 9-out-of-10 AI-generated market analysis fell apart in 15 minutes. The pattern behind it explains why 42% of companies are abandoning their AI initiatives.
What 130 Real Agentic AI Vendors Reveal About Where the Technology Actually Works
Of thousands of vendors claiming agentic AI, Gartner says only about 130 are real. What those 130 reveal about where the technology works — and where it doesn't.
Demystifying Agentic AI
Every vendor in your inbox has it. None of them agree on what it means. Here's what agentic AI actually is, what it isn't, and why most projects fail.
The AI Privacy Comparison Every Business Needs to See
We compared the privacy policies of 11 major AI platforms. Every consumer tier trains on your data. The enterprise tier is the minimum for business use.
What Is Agentic AI? Nobody Agrees.
We counted fifteen definitions from fifteen organizations. Here's what we found — and the only two things everyone agrees on.
The Simple Rule that Separates AI Winners from Everyone Else
One rule separates companies getting ROI from AI and companies burning money on it. Automate the repeatable. Keep humans on the judgment calls.
APIs: The Unsung Heroes of AI
While AI models provide reasoning and data provides context, APIs are the 'action layer' that allows AI to perform tasks. If your AI strategy doesn't include an API strategy, you don't have an AI strategy.
LLM Pong and the Crustafarian Panic
AI agents 'creating religion' made headlines. But what's really happening is pattern matching, not consciousness. It's Pong with better graphics.
AI Explained in Two Exercises
Most people think AI is either magic or hype. It's neither. A simple exercise shows the fundamental difference between traditional software and machine learning.
The Software Feature No One Asked For (But AI Agents Will Force)
AI agents break traditional permission models. Learn why role-based access control fails at machine speed, and what companies need to deploy AI agents safely.
Have a question or topic you'd like me to cover?
I'm always interested in what's on the minds of operators and investors. Reach out.
Get in touch