Claude, Perplexity, Gemini and me.

AI Wrote This. Sort Of.

Yes, AI slop is everywhere.  I don’t contribute to it.

I use AI to write. I don’t hide it. But like everything I do, I’ve built a system to get results.

AI Has Knowledge, Not Experience

Every post starts with something I notice. A pattern in my work, a question nobody is asking, a gap between what a vendor reports and what’s really happening. That’s 25 years of experience, not a language model.

AI doesn’t know if your backup strategy actually protects your business. AI can’t spot when a vendor relationship goes bad. AI hasn’t sat across from a business owner who just learned they’ve been hacked. I have.

I’ve been in this business for 25 years.   Reading, listening, and watching. Succeeding and failing.   This is where I get my ideas.

The Researcher

I pick the topic. I decide what matters. I outline the argument.

I built a custom agent that mimics how I actually research. I named it Scoop.

The process starts with a simple chat and a bit of back-and-forth to nail down the topic. This is done using Claude. Once we have an idea, Scoop gets to work. First is a quick survey of the landscape using Gemini. I use Gemini because it’s fast, cheap, and can easily surface very recent content.

Once we know a story has legs, Scoop digs in. I use Perplexity to generate a source-backed outline and executive summary. Is there enough data to back up my claims? If not, why not? This may take a few rounds to get right. From that outline, Claude steps back in with a voice guide to transform it into a rough draft.

Then it’s my turn.

Insert Me

My draft isn’t sloppy, but it’s not me either.

I pull the draft into Grammarly to make it my own. I still find Grammarly great for cleaning up my writing, especially flagging passive voice — a habit I picked up doing scientific research ages ago.

Even with style guides and clear processes, AI tends to be verbose and hedges. That’s not me.

I try to write and speak as I would to my clients — simple and direct. I’d rather be clear than clever.

I cut anything that sounds like a consulting firm’s slide deck — which, for whatever reason, nearly every LLM loves to spit out. I connect every claim to something I’ve done, seen, or can back up with a reference.

If I wouldn’t say it to a client, it doesn’t survive the edit.

The Editor

Once I have my draft, I send it to my editor. Yes, another AI agent designed to edit. This agent doesn’t make content changes without approval. Instead, it flags times when I violate my own rules. Did I make the same point in two different places? Do all claims have referenced data, or are they firsthand accounts? Did technobabble slip in?

When you dial in your editing rubric, you’ll be surprised how well it can identify problems and, in some cases, correct them.

The Fact-Check

Before publishing, I fact-check. Any claim with data or technical details gets verified against current sources. I use AI again, but hands-on. NotebookLM is very useful for this — the tool can confirm that my linked data actually supports my claims. Technology changes fast, so I try to keep current. Sure, I may make mistakes. If you see one, let me know.

Stories are different — firsthand accounts of 25 years in this business. Like the time Robert Marsh of EV1 Servers had a fleet of limos, a bright yellow Humvee, and a big rig take nerds from the Rackshack data center to the after-party. That’s not a data point. That’s a memory. I know the difference.

Why Tell You This?

Transparency is how I work.

I tell clients I don’t know everything. I have deep expertise in some areas and solid knowledge in others. I’m not afraid to say, let me find someone who really knows about this. After 25 years, I have good contacts — lawyers, programmers, graphic artists, system admins, SEO experts, business consultants, other founders, and CEOs.

The value isn’t in whether I typed every word. It’s in whether the thinking is solid, the experience is real, and the advice helps you make better decisions.

That’s what I’m accountable for in every post.