Where AI Actually Earns Its Keep: Notes From Running a Practice on It Every Day

I run OKC CIO Partners on AI — not as a talking point, but as the actual operating system behind the practice. Client research, meeting-to-deliverable turnaround, content, scheduling, follow-up: I use it daily, the same way I'd expect a client to use it inside their own business. That's a different vantage point than most of what gets written about AI adoption, which tends to come from people selling the tools rather than running a business on them. Here's what that vantage point has actually taught me.

The pattern that keeps repeating

The failures I see aren't about the tool. They're about the missing owner. Years ago, I watched a company keep paying for call-center analytics software nobody was actually reading — the tool worked fine, but there was no human process wrapped around it, no one whose job was to look at the data and change a decision because of it. AI tools fail the same way. A business turns on Copilot or a chatbot, nobody redesigns a single workflow around it, and six months later it's a subscription line item nobody remembers approving. The tool was never the problem. The missing owner was.

Where I've actually seen it pay off

Three patterns show up consistently, in my own practice and in the businesses I advise:

• Turning unstructured input into a finished deliverable, fast. A meeting transcript becomes a written assessment, a risk summary, or a follow-up email within hours instead of days. The value isn't "AI writes for me" — it's that a task that used to sit in a queue for a week now clears the same day, which changes how responsive the whole business feels to a client.
• Compressing research and first-draft work. Competitive scans, vendor comparisons, first drafts of a proposal — the kind of work that used to eat half a day now takes twenty minutes of setup and review. The judgment call still belongs to a person; the blank-page problem doesn't exist anymore.
• Recurring, well-defined tasks running on a schedule. Status reports, follow-up sequences, weekly summaries — anything with a repeatable shape can run automatically, with a human reviewing output rather than producing it from scratch every time.

Where it's a distraction, not a capability

The same tools fail predictably in a few places: tasks where the "review" step takes as long as just doing the work yourself; decisions that need context the AI doesn't have and won't ask for; anything customer-facing where a wrong or off-tone answer costs more than the time saved. And there's a data-handling question underneath all of it — what's actually leaving your network when an employee pastes a client contract into a public chatbot, and who signed off on that. That's a governance question, not a technology question, and it's usually nobody's job to ask it.

The readiness question, before the tool question

Most owners ask "which AI tool should we use." The better first question is "which of our workflows has a clear owner, a repeatable shape, and a defined output" — because that's the shortlist of places AI will actually save time instead of just moving the busywork somewhere new. That assessment takes an afternoon, and it's the difference between AI adoption that shows up on the bottom line and AI adoption that shows up as an unused line item next year.

← All articles